FEMS Microbiology Ecology 45 (2003) 135^145
www.fems-microbiology.org
Microbial community variation in pristine and polluted nearshore
Antarctic sediments
Shane M. Powell
a
a;
, John P. Bowman a , Ian Snape b , Jonathan S. Stark
b
School of Agricultural Science, University of Tasmania, Private Bag 54, Hobart, Tasmania 7001, Australia
b
Human Impacts Research Programme, Australian Antarctic Division, Kingston, Tasmania, Australia
Received 20 November 2002 ; received in revised form 29 April 2003; accepted 29 April 2003
First published online 4 June 2003
Abstract
Two molecular methods were used to investigate the microbial population of Antarctic marine sediments to determine the effects of
petroleum and heavy metal pollution. Sediment samples were collected in a nested design from impacted and non-impacted locations. A
detailed description of the diversity of the microbial population in two samples was obtained using 16S ribosomal DNA clone libraries
constructed from an impacted and a non-impacted location. The clone libraries were very similar with the exception of two sequence
clusters containing clones from only the impacted location. All samples were analysed by denaturing gradient gel electrophoresis. The
band patterns generated were transformed into a presence/absence matrix and a multivariate approach was used to test for differences in
the locations. Statistically significant differences were observed both between and within locations. Impacted locations showed a greater
variability within themselves than the control locations. Correlations between the community patterns and environmental variables
suggested that pollution was one of a number of factors affecting the microbial community composition.
2 2003 Federation of European Microbiological Societies. Published by Elsevier Science B.V. All rights reserved.
Keywords : Denaturing gradient gel electrophoresis ; Antarctic microbial ecology ; Impacted sediment
1. Introduction
Despite our relatively recent arrival in the Antarctic,
human activities have already had a signi¢cant impact
on the environment [1]. Past waste management practices
as well as accidents have been responsible for pollution of
both the marine and terrestrial environments. Although all
waste produced through our activities is now returned to
the country of origin, dumping of waste and disposal to
sea were common practices that left a legacy of pollution.
Petroleum spills have been reported at most scienti¢c research stations including Amundsen^Scott at the South
Pole [2], Palmer Station on the Antarctic Peninsula [3]
and Casey Station in the Windmill Islands [4]. There
have also been major marine spills such as the Bahia Pariso that ran aground on the Antarctic Peninsula losing
over 150 000 gallons of petroleum products [5].
* Corresponding author. Tel. : +612 (3) 62262776;
Fax : +61 (3) 62262642.
E-mail address : [email protected] (S.M. Powell).
Three successive stations have been in operation in the
Windmill Islands region since 1957 and there are welldocumented accounts of contamination associated with
station activities [1,4,6]. Near the locations of Old and
New Casey stations there have been several large spills
of petroleum products as well as many smaller spills associated with day-to-day operations. There is also an abandoned waste disposal site in the area that is known to be
highly contaminated [4]. During the annual melt, water
runs through both the waste disposal site and several hydrocarbon plumes and entrains and transports contaminants into Brown Bay. Elevated levels of hydrocarbons
(up to 200 mg kg31 of total petroleum hydrocarbons)
and heavy metals such as copper, lead and zinc have
been measured in Brown Bay sediments [1]. Sediment
characteristics such as grain size and total organic carbon
(TOC) were found to be variable at all sites, though TOC
was more variable at impacted locations than control locations (see Table 1 for examples).
Previous work on the benthic fauna in the marine environment surrounding Casey station has shown di¡erences
in the infaunal communities that were correlated with the
presence of pollutants, the most important of which were
0168-6496 / 03 / $22.00 2 2003 Federation of European Microbiological Societies. Published by Elsevier Science B.V. All rights reserved.
doi:10.1016/S0168-6496(03)00135-1
FEMSEC 1532 7-7-03
136
S.M. Powell et al. / FEMS Microbiology Ecology 45 (2003) 135^145
heavy metals [7]. However, no single pollutant or environmental variable appeared to be consistently important.
Infaunal communities in control locations were more diverse than Brown Bay communities and several taxa were
not found in any Brown Bay samples. Moreover, some
species that are commonly found associated with pollution, such as capitellid polychaetes, were present in Brown
Bay but not at control locations. The intention of our
study was to extend the ecological investigations of the
region to include some measure of the microbial populations in both contaminated and non-contaminated sites. In
terms of impact assessment, microbial techniques potentially o¡er an alternative method to surveys of infauna
and macrofauna for identifying areas a¡ected by pollution.
Molecular techniques are important in microbial ecology as they are not dependent on the culturability of unknown micro-organisms but instead rely on the extraction
of DNA from environmental samples. As for all methods,
polymerase chain reaction (PCR), which forms the basis of
many molecular techniques, is subject to certain biases and
limitations. These include the existence of multiple 16S
rRNA gene operons, primer annealing speci¢city, di¡erences in the ease of ampli¢cation of DNA from di¡erent
organisms and the need to design primers that will bind to
as many target organisms as possible. Despite this, a relative comparison between samples is still possible if all
samples are treated in the same manner.
The construction of clone libraries and the sequencing
of 16S rRNA gene sequences is a well-established molecular method. It provides high-resolution phylogenetic information on the micro-organisms present in a sample.
The higher the number of clones sequenced, the more
reliable the estimate of diversity and the more complete
the knowledge of community structure. Clone library analysis is, however, very labour-intensive and it is not yet
possible to analyse the large numbers of samples required
to obtain quantitative, statistically signi¢cant data in a
regional survey.
Denaturing gradient gel electrophoresis (DGGE) is a
relatively new, but increasingly popular, technique in microbial ecology [8]. One of the advantages of this method
is that it is possible to analyse many samples simulta-
neously so that statistically testable data can be obtained.
DGGE relies on the separation of a mixture of 16S rRNA
gene fragments with di¡erent sequences. Each sample results in a pattern of bands on a gel. Although theoretically
each band represents a unique sequence and therefore a
unique species, this is not always the case. Several sequences may co-migrate to form what appears to be a single
band [9,10] and some organisms generate more than one
band [11,12]. In addition, DNA extraction methods may
select for certain types of organisms [13]. PCR biases are
known to have an e¡ect [14,15] and even sample handling
can change the result [16]. However, di¡erences in the
banding patterns overall are related to the di¡erences in
the microbial community composition. Statistical analysis
of these banding patterns is a potentially powerful tool for
distinguishing signi¢cant di¡erences in microbial communities. Once di¡erences have been identi¢ed it is also possible to excise bands from the gel and sequence them in
order to identify species present. This information o¡ers
insights into species distribution and possible biochemical
processes in the sediment. Several approaches have been
used to transform DGGE banding patterns into a quantitative format. Some authors have incorporated the intensity of the bands as well as their position [17,18]. Alternatively, the banding patterns have been transformed into
a presence/absence matrix [19]. Recently, ordination methods such as principal components analysis [20], multi-dimensional scaling [19] and cluster analysis [21] have been
used to graphically display the similarities making interpretation of the data easier.
There were several aims in this work. Firstly, we wanted
to evaluate DGGE as a tool for following changes in
microbial populations in large-scale studies. This involved
testing the reproducibility of the banding patterns and
establishing statistical methods for analysing them. Another molecular method (the clone libraries) was used to examine broad community composition of an impacted and
non-impacted location in order to verify or refute the conclusions reached from the DGGE analysis. The second
part of the work was to investigate variation in microbial
communities in contaminated marine sediments in comparison to sediment from control locations in the context
of environmental impact.
Table 1
Range of values for some environmental variables
Zinc (mg kg )
Copper (mg kg31 )
Lead (mg kg31 )
Iron (mg kg31 )
Manganese (mg kg31 )
Arsenic (mg kg31 )
Cadmium (mg kg31 )
TOC (g kg31 )
Grain size (mean particle diameter, Wm)
31
Brown Bay
O’Brien Bay
Sparkes Bay
Wharf area
13^65
5^30
18^85
500^5300
1^4
5^35
0.5^2
15^47
24^109
2^18
1^2
ND
50^230
3^5
1^14
0.1^2
10^25
108^363
23^26
2^3
ND
150^220
1^3
4^7
3^4
29^32
34^55
8^35
2^5
1^6
200^400
1^3
3^8
0.5^3
14^42
30^296
Data are summarised from [7]. ND = not detected.
FEMSEC 1532 7-7-03
S.M. Powell et al. / FEMS Microbiology Ecology 45 (2003) 135^145
137
Newcomb Bay
Wharf.
Brown
Bay
BB3
Windmill Islands
CASEY
OB2
OB3
O`Brien Bay
OB1
N
0
1
2
3
4 Kilometres
Sparkes
Bay
Fig. 1. Map of the Windmill Islands region showing the position of sampling locations and Casey Station.
were taken back to the laboratory and frozen to 320‡C
within 6^8 h of collection.
2. Materials and methods
2.1. Design
2.3. DNA extraction
Samples were collected from three locations within
Brown and O’Brien Bays, from one location within
Sparkes Bay and one location at the Casey wharf (see
Fig. 1). The locations were generally separated by kilometres, except for Brown Bay in which the locations
were separated by approximately 300 m. Samples for
this survey were collected in a hierarchical nested design.
That is, samples were taken from two sites within each
location (approximately 100 m apart) and from two plots
within each site (approximately 10 m apart). Four replicate samples were taken from each plot, one of which was
utilised for microbial analysis. The locations were shallow
embayments with a range of sediment characteristics
(muddy to sandy) and physical attributes (sea-ice cover,
depth, aspect etc.). Sampling locations are described in
greater detail in Stark et al. [7].
2.2. Sampling
The samples were collected by diver using a hand-held
corer. Details are described in Stark et al. [7]. Samples
DNA was extracted from sediment samples using a
freeze^thaw method based on that described by Rochelle
et al. [16]. Approximately 1 g or 1 ml of sediment was
suspended in 2 ml of lysis bu¡er (0.15 M NaCl, 0.1 M
EDTA, 4% sodium dodecyl sulfate) with 30 mg lysozyme
and ca. 20 mg polyvinylpolypyrrolidone. Samples were
heated in a 55‡C water bath for 10 min and then subject
to three rounds of freezing at 380‡C for 15 min and heating at 55‡C for 10 min. After the ¢nal thaw, samples were
extracted with an equal volume of Tris-equilibrated phenol
followed by extraction with an equal volume of phenol:
chloroform:isoamyl alcohol (25:24:1). The aqueous phase
was removed to a clean tube and 0.7 volumes of isopropanol added. Extracts were incubated for 1 h at room
temperature, followed by centrifugation for 30 min at
3100Ug. Pellets were air-dried and resuspended in 100 Wl
of sterile milli-Q water overnight at 4‡C. Extracts were
checked on a 1% agarose gel before the ¢nal puri¢cation
step on Chromaspin columns (Clontech) following the
manufacturer’s directions. The amount of DNA present
FEMSEC 1532 7-7-03
138
S.M. Powell et al. / FEMS Microbiology Ecology 45 (2003) 135^145
in extracts was measured using the Hoechst £uorometric
assay and a Bio-Rad £uorimeter.
compare the similarity of the two libraries directly. This
calculation gives a P value that is considered to show a
signi¢cant di¡erence at values less than 0.05.
2.4. Clone libraries
2.5. DGGE
Two clone libraries were generated using DNA extracted from an O’Brien Bay sample (OB311) and a Brown
Bay sample (BB222). A fragment of the 16S rRNA gene
was ampli¢ed using Advantage 2 Taq (Clontech) with the
supplied 10U bu¡er and the primers 519f (CAG CMG
CCG CGG TAA TAC) and 1392r (ACG GGC GGT
GTG GRC). These universal primers are expected to
bind to the majority of bacteria and archaea. Each 100
Wl reaction mix contained 5 Wl of 10U bu¡er, 2 Wl of Taq,
1.25 mM of each deoxynucleoside triphosphate, 20 nmol
of each primer and 80 ng of template DNA. The following
thermal cycling programme was used: initial denaturing at
94‡C for 15 min; 30 cycles of denaturing at 94‡C for 1 min,
annealing at 52‡C for 1 min, extension at 72‡C for 1.5
min; ¢nal extension at 72‡C for 10 min. The reaction
products were puri¢ed using the Prep-a-gene kit (BioRad).
The fragment was cloned using the pGEM-T easy vector system (Promega) and transformed into Epicurian coli
XL ultracompetent cells (Stratagene) following the manufacturer’s directions. Transformants were screened using
blue^white screening on Luria agar containing X-gal and
isopropyl-L-D-thiogalactose. Approximately 250 white colonies from each library were sub-cultured.
Ultraclean mini plasmid preps (MoBio) were used to
extract the plasmids from the sub-cultured clones. 3 Wl
of the extracts were run on a 1% agarose gel alongside a
molecular mass marker in order to verify that the plasmid
contained the correct-sized insert.
Positive clones were sequenced with the BigDye Terminator Ready Reaction mix sequencing reactions (Applied
Biosystems). 7 Wl of the plasmid extract was used in a 20
Wl reaction with 5 pmol of either the M13f or M13r primer. This generated sequences of approximately 1000 bp.
These sequences are deposited under GenBank accession
numbers AY133347 to AY133467.
The chimera-check tool of the Ribosomal RNA Database Project (http://www.rdp.cme.msu.edu) [22] was used
to check possible chimeric sequences. Sequences were
aligned against reference sequences obtained from GenBank (http://www.ncbi.nlm.nih.gov/blast) [23]. DNADIST
and NEIGHBOR from the PHYLIP package [24] were
used to generate phylogenetic trees. Cloned sequences
that were more than 98% similar to each other were considered to be the same phylotype [25] for the purposes of
calculating diversity statistics. However, all sequences are
shown in Fig. 3. Simpson’s index (D = gp2i ) and the Shannon^Wiener index (H = 3gpi ln(pi )) were calculated and
the Chao-1 estimator (http://www2.biology.ualberta.ca/
jbrzusto/rarefact.php) was used to calculate species richness. The method of Singleton et al. [26] was used to
Advantage 2 Taq (Clontech) with the supplied 10U
bu¡er was used for ampli¢cation of a fragment of the
16S rRNA gene containing the V3 and V4 regions. Each
50 Wl reaction mix contained 5 Wl of 10U bu¡er, 1 Wl of
Taq, 1.25 mM of each deoxynucleoside triphosphate, 20
nmol of each primer and either 20 ng of sample DNA or
1 ng of the standard control DNA mix. The standard
DNA mix consisted of 5 ng Wl31 each of genomic DNA
preparations from four strains grown routinely in our laboratory and chosen because they denatured at a range of
di¡erent denaturant concentrations. The primers were
907R (CCG TCA ATT CCT TTG AGT TT) and 341F
with a GC clamp (CGC CCG CCG CGC CCC GCG
CCC GGC CCG CCG CCC CCG CCC CCC TAC
GGG AGG CAG CAG). The touchdown thermal cycling
programme consisted of the following steps: initial denaturing step at 94‡C for 5 min; then 10 cycles of denaturing
at 94‡C for 1 min, annealing at 65‡C for 1 min (decreasing
by 1‡C each cycle) and extension at 72‡C for 3 min; followed by 20 cycles of 94‡C for 1 min, 55‡C for 1 min,
72‡C for 2 min; ¢nal extension at 72‡C for 4 min and then
held at 4‡C.
The DGGE was performed using a D-Code Universal
Mutation Detection System (Bio-Rad). Half the volume of
the PCR products were run on 6% acrylamide gels with a
denaturing gradient of 30^65% (where 100% denaturant is
7 M urea and 40% formamide). Gels were run at 80 V for
16 h at 60‡C in 1UTAE (40 mM Tris, 20 mM sodium
acetate, 1 mM EDTA). Standards were run on either side
of the gel and the outside lanes were not used. In order to
obtain even heat distribution throughout the tank, the
entire tank was placed on a magnetic stirring plate.
Gels were stained in 1:1000 Sybergold (Molecular
Probes) in the dark with gentle shaking for approximately
20 min. They were then washed once with deionised water
and destained with deionised water for 20 min before
viewing on a UV transilluminator.
Photos were scanned in and viewed with the UTHSCSA
ImageTool program (developed at the University of Texas
Health Science Center at San Antonio, TX, USA and
available from the Internet by anonymous FTP from
ftp://maxrad6.uthscsa.edu). The best possible banding pattern was obtained by enhancing the contrast and greyscale
of the images. This banding pattern was then transformed
into a presence/absence matrix for statistical analysis. The
standards were used to check for gradient consistency between gels and to assist in comparing the position of
bands between gels.
A multivariate approach using the Primer5 package
(Plymouth Marine Laboratory, UK) was used to investi-
FEMSEC 1532 7-7-03
S.M. Powell et al. / FEMS Microbiology Ecology 45 (2003) 135^145
2.6. Nested samples
In order to reduce the e¡ects of PCR and gel bias on the
banding patterns, all the samples were analysed three
times. All samples were ampli¢ed in the same round of
PCR. This PCR was carried out twice and each sample
was run on a total of three gels. That is, one of the rounds
of PCR was run on two gels (half the volume of a PCR
reaction is loaded onto a gel) and the second round of
PCR was run only once. These three banding patterns
were then added together as described above.
3. Results
3.1. Clone libraries
After discarding sequences that were suspected of being
chimaeric and sequences that were of poor quality, there
were 98 clones from the Brown Bay library and 85 from
the O’Brien Bay library. Of these, 66 and 64 respectively
were unique phylotypes. Fourteen phylotypes were found
in both libraries, although many of the other phylotypes
18
Number of unique phylotypes
16
14
12
10
8
6
4
2
ba
rs
Fl
a
vo
O
th
e
ct
er
ia
ea
ha
yc
Pl
a
nc
to
m
ba
eo
ro
t
A
rc
et
es
ia
ct
er
te
r
ot
he
rp
C
hr
om
at
ib
ba
ac
ct
er
ia
ia
0
D
el
ta
gate the DGGE banding patterns. For some analyses, the
banding patterns from several runs were pooled and the
total presence/absence of bands recorded. Similarity matrices were generated using the Bray^Curtis measure on
presence/absence of bands. Non-metric multidimensional
scaling plots (nMDS) were used to represent the relative
similarities between the samples. The stress levels of the
nMDS plots were generally between 0.1 and 0.2 and cluster analysis (hierarchical agglomerative clustering with
group average linkage) was used to check the groupings
produced by the nMDS procedure. The analysis of similarity (ANOSIM) test (one-way) was used to compare
groups. ANOSIM R values of 1 indicate that replicates
within a location are more similar to each other than to
any samples from another location whereas an R value of
0 indicates that there is as much variation within a group
as between the two groups being compared. In deciding
whether locations were di¡erent, both the R value and
signi¢cance level were considered. The signi¢cance level
was a¡ected by the small number of samples in the Wharf
and Sparkes Bay locations. Rather than considering a
speci¢c number (e.g. 5%) to be signi¢cant, levels close to
the minimum possible were taken to be signi¢cant.
The BIOENV procedure in the Primer5 package was
used to investigate correlations between environmental
variables as reported in [7] and the microbial community
structure described by the DGGE analysis. TOC, heavy
metals (lead, tin, zinc, copper, iron, antimony, cadmium,
chromium, manganese, mercury, nickel and silver) and
sediment characteristics (skewness, kurtosis, sorted particle
diameter and mean particle diameter) were included in the
analyses.
139
Fig. 2. Comparison of Brown Bay (black bars) and O’Brien Bay (grey
bars) clone libraries by the number of unique phylotypes in various phylogenetic groups.
were closely related despite falling outside our de¢nition of
phylotype. This number of unique sequences represents
33% (Brown Bay) and 25% (O’Brien Bay) coverage where
coverage is considered to be the proportion of clones
found more than once. The species richness was 304 for
Brown Bay and 282 for O’Brien Bay as estimated by the
Chao-1 estimator. Both these ¢gures suggest that the microbial diversity of both bays is much higher than detected
here. The Simpson diversity index was 0.025 for Brown
Bay and 0.022 for O’Brien Bay. The Shannon^Wiener index was 4.05 for both Brown Bay and O’Brien Bay. Again
this indicates a high microbial diversity in these sediments.
The sequences obtained from the clone libraries were
divided into seven groups for ease of handling. The most
numerous sequences belonged to the N and Q proteobacteria followed by the £avobacteria. The two libraries are
compared in Fig. 2 on the basis of the number of unique
phylotypes in each group. Both sites have diverse microbial populations with similar proportions of the seven phylogenetic groups. The P values generated by the method
described in [26] were 0.063 (when BB2 is X and OB3 is Y)
and 0.372 (when OB3 is X and BB2 is Y). Neither of these
values indicates a signi¢cant di¡erence between the two
clone libraries.
When the sequences were aligned onto phylogenetic
trees, there tended to be clusters of very similar phylotypes
that generally contained representatives from both libraries. However, in both the N and Q proteobacteria, there
were clusters in which there were only Brown Bay clones
(Fig. 3). In the N proteobacteria these clones were related
to Desulfobacula toluolica and Desulfobacterium phenolicum, both hydrocarbon-oxidising sulfate reducers. The
cluster of Brown Bay clones in the Q proteobacteria was
related to the genus Pseudoalteromonas.
3.2. Evaluation of DGGE reproducibility: PCR and gel
e¡ects
The ¢rst stage was to evaluate the reproducibility of
DGGE banding patterns in the sediment samples. Ini-
FEMSEC 1532 7-7-03
140
S.M. Powell et al. / FEMS Microbiology Ecology 45 (2003) 135^145
A quif ex aeol icus
B128
O39
Desulf uromonas pal mitati s
B 223
Desulf obul bus medi terraneus
O182
O146
B164
O143
O184
B181
O106
O110
O3
B141
B160
B158
B154
O108
D esulf obacterium catechol icum
Desul fotal ea arctica
Desul focapsa sulf oexi gens
Desul focapsa thi ozymogenes
O190
B136
B127
B144
B115
D esul fobacula toluol ica
D esul fobacterium phenol icum
B109
B41
B201
B94
B162
B183
O87
O136
O162
O134
O35
O43
B215
B111
O233
O207
B97
O40
O142
O176
O123
O50
Desul fobacteri um cetonium
O34
O216
B31
Desul fonema magnum
Desulf ococcus mul ti vorans
Desul fovibrio desul furi cans
B3
O122
O111
A quif ex aeol icus
N itrosococcus hal ophil us
O36
Thi orhodococcus minus
Ri ftia pachypti la endosymbi ont
B230
B193
B140
O189
B131
O55
O221
B135
O62
O217
O57
O138
O140
O172
B 126
O198
O88
B169
B125
Pseudomonas cel lul osa
M icrobul bif er sal ipal udi s
Pseudomonas synxantha
B186
M ethylomonas rubra
B 110
B147
Aeromonas bestari um
Shewanell a putrif aci ens
B 17
B 153
Pseudoal terom onas tetraodoni s
B 149
B15
B178
B69
B177
O96
thial kali vibri o denitri fi cans
O48
B26
B137
OO98
O163
O49
B216
B53
B163
B200
O169
O186
B114
B 219
O75
O117
0.1
0.1
A
B
Fig. 3. Phylogenetic trees of the N proteobacteria (A) and Gamma proteobacteria (B) in the O’Brien Bay (numbers preceded by O) and Brown Bay
(numbers preceded by B) clone libraries. Bracket indicates clusters of Brown Bay clones that are associated with (A) Desulfobacula toluolica and (B) the
genus Pseudoalteromonas.
tially, three samples (one each from Sparkes, Brown and
O’Brien bays) were subject to three simultaneous PCR
reactions that were then run on the same DGGE gel to
provide three PCR replicates for each sample. Each of the
replicates was over 95% similar to each other, but the
three samples were clearly di¡erent from each other. The
PCR was repeated on a di¡erent day and these reactions
run alongside the remaining half reactions from the ¢rst
round of PCR. Once again, replicates that had been subject to the same round of PCR and run on the same gel
were over 95% similar. However, for every sample, there
were di¡erences between reactions either ampli¢ed on different days or run on di¡erent gels (Fig. 4). The ANOSIM
test was used to compare the similarities between each run
(Table 2). In all three samples, replicates run on the same
gel were the most similar whereas those subject to the
same PCR reaction but run on di¡erent gels were the
most dissimilar. This suggests that di¡erences between
gels are the main reason for di¡erences in banding patterns of the same sample.
Fig. 4. nMDS of three samples (one each from Brown Bay, O’Brien
Bay and Sparkes Bay), each replicated three times in two rounds of
PCR and run on a total of three separate gels (squares, triangles and
circles). In some cases, the banding patterns of replicates were identical
and therefore the symbol appears only once or twice instead of three
times (e.g. Sparkes Bay, triangles).
FEMSEC 1532 7-7-03
S.M. Powell et al. / FEMS Microbiology Ecology 45 (2003) 135^145
141
Table 2
Comparison of banding patterns for each sample over two PCR ampli¢cations and two DGGE gel runs
Most dissimilar
Most similar
Sparkes
Brown
O’Brien
A,B (0.889)
A,C (0.556)
B,C (0.519)
A,C (1.0)
A,B (0.704)
B,C (0.704)
A,B (1.0)
A,C (1.0)
B,C (0.556)
A indicates PCR 1/gel 1, B indicates PCR 1/gel 2 and C indicates
PCR2/gel 2. ANOSIM R statistics are shown in parentheses.
3.3. Nested samples
The entire nested survey sample set was analysed three
times. The negative image of one of these gels is shown in
Fig. 5. The controls in the outside lanes were consistent
across all gels used in the analysis. When the results from
individual runs were plotted on an nMDS, some grouping
by location was observed (not shown) although the nMDS
were di¡erent for each run.
The nMDS obtained from pooling the banding patterns
from each batch together is shown in Fig. 6 and the ANOSIM R statistic for each pair of locations is given in Table 3. Generally, the samples group together within their
locations although the Brown Bay locations overlap.
O’Brien Bay, Sparkes Bay and Wharf area locations are
all distinct (ANOSIM values generally over 0.8). In
O’Brien Bay, the two locations from the north side of
the bay are more similar to each other (ANOSIM R value
of 0.448) than to the location from the south side of the
bay (ANOSIM R values of 0.844 and 0.948).
In contrast, the ANOSIM R values for each pair of
Fig. 6. nMDS showing relative similarities between locations: Brown
Bay 2 (black triangles), 3 (black circles), 4 (black squares); O’Brien Bay
1 (grey triangles), 2 (grey circles), 3 (grey squares); Sparkes Bay (crossed
open squares); and Wharf (open down-pointing triangles). All samples
were analysed three times, this plot resulted from pooling the three presence/absence matrices.
Brown Bay locations are negative, indicating as much variation within each location as between them. In Fig. 6 it
can be seen that the spread of the Brown Bay locations is
greater than the spread of the other locations suggesting a
greater heterogeneity in Brown Bay. The two sites within
the wharf location are di¡erent to each other (ANOSIM
R value of 0.8) whereas the two sites within the Sparkes
Bay location are more similar (ANOSIM R value of 0.3).
The BIOENV procedure ¢nds the combinations of environmental variables that best ¢t the microbial community structure patterns. The environmental data were
transformed by several methods including square root,
log and fourth root. The square root transform on the
environmental data resulted in the best separation of the
samples into their locations as seen in nMDS plots (not
shown). The correlations from this transform are presented in Table 4. However, the same variables appeared
in much the same order of importance regardless of which
of the above transforms was used. From Table 4 it can be
seen that the highest correlation between environmental
variables and the microbial community structure occurs
when TOC, arsenic, iron and manganese are combined
(P = 0.411). TOC, arsenic, iron and manganese and cadmium consistently appear in the highest correlations. The
measures of sediment size and sorting are much less important and rarely appear in the higher correlations for
each group of variables.
4. Discussion
Fig. 5. Negative image of a photo of a DGGE gel containing nine of
the nested samples. The two outside lanes are the standard control
DNA mix, lanes 2 and 10 are from Sparkes 2, lane 3 is from the
Wharf, lanes 4, 5, 8 and 9 are Brown Bay 3 and lanes 6 and 7 are
Brown Bay 4.
4.1. Clone library results
The clone libraries described here are the ¢rst recorded
for nearshore Antarctic marine sediments. However, as
FEMSEC 1532 7-7-03
142
S.M. Powell et al. / FEMS Microbiology Ecology 45 (2003) 135^145
Table 3
ANOSIM values comparing the similarity between pairs of locations
based on data pooled from three analyses
Pairs of locations
R statistic
Signi¢cance level (%)
Brown 2, Brown 3
Brown 2, Brown 4
Brown 2, O’Brien 1
Brown 2, O’Brien 2
Brown 2, O’Brien 3
Brown 2, Sparkes
Brown 2, Wharf
Brown 3, Brown 4
Brown 3, O’Brien 1
Brown 3, O’Brien 2
Brown 3, O’Brien 3
Brown 3, Sparkes
Brown 3, Wharf
Brown 4, O’Brien 1
Brown 4, O’Brien 2
Brown 4, O’Brien 3
Brown 4, Sparkes
Brown 4, Wharf
O’Brien 1, O’Brien 2
O’Brien 1, O’Brien 3
O’Brien 1, Sparkes
O’Brien 1, Wharf
O’Brien 2, O’Brien 3
O’Brien 2, Sparkes
O’Brien 2, Wharf
O’Brien 3, Sparkes
O’Brien 3, Wharf
Sparkes, Wharf
30.016
30.286
0.490
0.255
0.635
0.406
0.875
30.107
0.839
0.406
0.646
0.953
0.990
0.607
0.804
0.786
1.000
1.000
0.844
0.948
0.938
0.990
0.448
0.875
0.969
0.943
0.969
0.464
54.3
86.7
2.9
11.4
2.9
2.9
2.9
73.3
2.9
5.7
2.9
2.9
2.9
6.7
6.7
6.7
6.7
6.7
2.9
2.9
2.9
2.9
2.9
2.9
2.9
2.9
2.9
2.9
from the 16S sequence alone whether these clones degrade
hydrocarbons as well. The fact that cultures of these
strains are required to determine metabolic capabilities is
one of the disadvantages of molecular techniques. Given
the random nature of the construction and sequencing of
clone libraries, it is possible that these sequences did occur
in O’Brien Bay, but were overlooked as they were less
numerous. One way of testing this is to use primers designed speci¢cally to amplify these groups and to probe
for them using £uorescence in situ hybridisation.
4.2. DGGE technique
they were constructed from only two samples, it is di⁄cult
to say how representative they are of the two bays. The
sequences obtained are only a fraction of those present in
the sediment, estimated at 33 and 25% coverage for Brown
Bay and O’Brien Bay respectively. Both libraries contain a
diverse array of sequences as indicated by both the Shannon^Wiener and Simpson’s indices. It was expected that
only very large di¡erences between the microbial communities would be seen in the clone libraries. They both have a
similar diversity, both contain the same major phylogenetic groups and the large number of closely related phylotypes (see for example Fig. 3) suggests that they are
quite similar. Other studies comparing marine sediment
clone libraries have found similarities between communities from apparently diverse locations. For example,
N proteobacteria, a large group in both the Brown and
O’Brien Bay clone libraries, were also a dominant group
in clone libraries from lakes in the Vestfold Hills in Eastern Antarctica [27] and Arctic marine sediment [28]. Both
of these studies also concluded that the diversity of cold
marine sediments was surprisingly high.
The two most interesting phylogenetic groups are shown
in more detail in Fig. 3. In both the Q and N proteobacteria, there are clusters of sequences that only contain
Brown Bay clones. Although phylogenetically related to
hydrocarbon-degrading strains, it is not possible to tell
DGGE banding patterns are known to be in£uenced by
many things: the DNA extraction method used; minor
variations in the PCR reaction; the gradient used; gradient and acrylamide variations between gels ; the complexity of the microbial community present in a sample
and the actual species present. Attempts have been made
previously to determine the reproducibility of banding patterns [10,29]. However, these mainly compared di¡erent
rounds of PCR on the same gel. Unfortunately, ecological
studies will generate more samples than can be analysed
on one gel. In addition, most work attempting to de¢ne
the sensitivity and reproducibility of DGGE used constructed assemblages [8,10]. Whilst this is an e¡ective, controlled approach, these assemblages are much more simple
than sediment samples and care should be taken in extrapolating results from DNA mixes to environmental samples. Generally we had between 20 and 30 distinguishable
bands per sample and it is clear from the clone libraries
that there are at least 60 di¡erent phylotypes present. It is
assumed that only numerically dominant sequences will
appear in a banding pattern, although this is also dependent on the DNA extraction method used and PCR biases.
We took a step-by-step approach to analysing the contribution of these biases to the reproducibility of DGGE
Table 4
Correlations between environmental variables and microbial community
structure patterns
All environmental data were subject to a square root transformation.
The environmental variables giving the best results when taken k at a
time are presented. Correlation coe⁄cients are given in parentheses,
bold indicates the highest overall value of P.
a
Total organic carbon.
b
Mean sorted particle size.
FEMSEC 1532 7-7-03
S.M. Powell et al. / FEMS Microbiology Ecology 45 (2003) 135^145
banding patterns. Our results show that both gel and PCR
e¡ects are important (Fig. 4) but that di¡erences between
gels are slightly more important (Table 2). Ferrari and
Hollibaugh [29] found a similar phenomenon. When the
same sample was subject to several rounds of PCR and
run on the same gel, the patterns were over 90% similar.
However, when they repeated a gel, depending on the
method used to compare the banding patterns, di¡erent
samples from the same gel could be more alike than the
same sample run on two di¡erent gels. They concluded
that the ‘gel signature’ could not be completely removed
in the image-processing step. Despite this, we have shown
that the banding patterns generated by repeated analysis
of the same sample are more similar to each other than
those from other samples analysed at the same time. That
is, at least for the samples used in this study, the greatest
overall in£uence on banding patterns was di¡erences within the samples themselves. In order to minimise the e¡ect
of other factors and allow the sample di¡erences to become the dominant factor, we found pooling data from
multiple runs to be an e¡ective solution. Our results suggest that analysing a small number of samples only once
is not su⁄cient to determine the relative di¡erences (or
similarities) between samples or between sampling locations. Furthermore, without determining the extent of variation in banding patterns caused by gel and PCR e¡ects,
banding patterns from di¡erent gels should not be compared.
4.3. DGGE results
The ANOSIM tests (Table 3) as well as the nMDS plot
(Fig. 6) showed that the O’Brien Bay locations were all
separate, signi¢cantly di¡erent groups. Within each location, however, the sites and plots are very similar to each
other. It is also interesting that the two O’Brien Bay locations that are geographically closer to each other are more
similar. This suggests that natural di¡erences in the environment at a scale of kilometres are enough to in£uence
the microbial community structure. However, on a smaller
scale (e.g. hundreds of metres), the communities are much
the same, perhaps as a result of homogeneous environmental conditions.
The spatial variation in microbial communities in the
Brown Bay locations is very di¡erent from the control
locations. The Brown Bay locations overlap to a greater
extent than the O’Brien Bay locations, but this is probably
due to the fact that they are much closer together (300 m
apart rather than kilometres apart). The variation within
each Brown Bay location is greater than the variation
within the control locations. Brown Bay 2, the closest to
the tip site, has two points close together and two outlying
points. We interpret this to be the result of heterogeneous
environmental conditions possibly caused by ‘hot-spots’ of
contamination. A sample taken next to a battery fragment
for example will have high levels of heavy metals and this
143
will probably in£uence the microbial community structure.
Environmental factors are more variable in Brown Bay
(Table 1) and these too could be contributing to the variation in microbial community structure.
The Sparkes and Wharf locations both form distinct
groups (Fig. 6). Despite being a control location, Sparkes
Bay is slightly more similar to the Brown Bay and Wharf
locations than the O’Brien Bay locations. This is possibly
due to the occurrence of naturally elevated levels of heavy
metals in this bay.
Although the locations generally form distinct groups,
there is some separation between impacted (Brown Bay
and Wharf) locations and non-impacted locations
(O’Brien and Sparkes Bays) with Sparkes Bay (which
has elevated metal levels) grouping with the impacted locations. The chemistry of the impacted sediments is quite
di¡erent in each location, Sparkes Bay has some elevated
levels of metals, Brown Bay has high levels of heavy metals and some petroleum hydrocarbons and the Wharf location has very high levels of petroleum hydrocarbons.
The e¡ect on the microbial communities is not as simple
as impacted and non-impacted but perhaps re£ects speci¢c
components of the pollution and the complexity of heterogeneous environmental conditions. Sampling of more control locations for further comparison of variability between control locations as well as between control and
impacted locations may assist in resolving this. In order
to con¢rm the e¡ect that human activity has had on the
benthic microbial populations at Casey, it would be necessary to compare samples taken from the same locations
(e.g. Brown Bay) before they became contaminated. As
this is not possible, in situ experiments with arti¢cially
contaminated sediment are currently under way.
4.4. Correlation of environmental variables with microbial
community patterns
In this study TOC levels, in combination with heavy
metals, are correlated (P = 0.4) with the microbial community structure patterns (Table 4). The type and availability
of carbon is one of the most important factors in£uencing
the development of a microbial community. Unfortunately
data which distinguish between petroleum hydrocarbons
and other organic carbons are not available for these samples. However, petroleum hydrocarbons contribute to the
TOC level which is higher for the impacted locations than
non-impacted locations (Table 1). Heavy metals also appear to have some in£uence on the microbial communities,
particularly iron, cadmium, manganese, zinc and arsenic.
Naturally elevated levels of some of these metals (e.g. cadmium) are found in Sparkes Bay, but others such as iron,
zinc and arsenic are far higher at impacted locations (Table 1) and are most likely anthropogenic in origin. However, the correlation between these factors and the microbial community structure was not very high. Other factors
which we have not measured (for example depth of oxygen
FEMSEC 1532 7-7-03
144
S.M. Powell et al. / FEMS Microbiology Ecology 45 (2003) 135^145
penetration into the sediment, hydrocarbon concentrations) may also be just as important.
[8]
4.5. Conclusions
In this work we attempted to de¢ne the extent to which
gel and PCR biases a¡ected the DGGE banding patterns.
With this knowledge, we were able to take steps to reduce
these problems when analysing our samples. In order to
obtain reliable results from DGGE, we suggest that multiple analyses of large numbers of samples should be undertaken and the banding patterns pooled. However, when
used in conjunction with large numbers of samples and
a statistical analysis of the results, DGGE can be a very
useful tool in microbial ecology. By using a second technique we were able to verify the similarity between the
microbial communities found in O’Brien Bay and Brown
Bay. The presence of two sequence clusters of Brown Bay
only clones suggests that there are also di¡erences.
Sediment microbial communities are very diverse and
are subject to a complex array of environmental factors
resulting in complex patterns of community variability. In
this study, variation was observed between control locations as well as between control and impacted locations.
However, as the microbial community structure patterns
corresponded to environmental variables which are
anthropogenic, it is likely that this variation is in part
due to human activities in the region.
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
Acknowledgements
This study was supported by funding from the Antarctic
Science Advisory Committee.
[17]
References
[18]
[1] Snape, I., Riddle, M.J., Stark, J.S., Cole, C.M., King, C.K., Duquesne, S. and Gore, D.B. (2001) Management and remediation of
contaminated sites at Casey Station, Antarctica. Polar Rec. 37, 119^
214.
[2] Wilkniss, P.E. (1990) Fuel spill clean up in the Antarctic. Antarct. J.
USA 25, 3^10.
[3] Kennicutt, M.C., McDonald, T.J., Denoux, G.J. and McDonald,
S.J. (1992) Hydrocarbon contamination on the Antarctic peninsula:
Arthur Harbour ^ subtidal sediments. Mar. Pollut. Bull. 24, 499^
506.
[4] Deprez, P.P., Arens, M. and Locher, H. (1999) Identi¢cation and
assessment of contaminated sites at Casey. Polar Rec. 195, 299^316.
[5] Kennicutt, M.C. (1990) Oil spillage in Antarctica. Environ. Sci. Technol. 24, 620^624.
[6] Snape, I., Cole, C.M., Gore, D.B., Riddle, M.J. and Yarnall, M.
(1998) A Preliminary Assessment of Contaminants at the Abandoned
Wilkes Station, East Antarctica, with Recommendations for Establishing an Environmental Management Strategy. Australian Antarctic
Division, Kingston, Tasmania.
[7] Stark, J.S., Riddle, M.J., Snape, I. and Scouller, R.C. (2003) Human
impacts in Antarctic marine soft-sediment assemblages: correlations
[19]
[20]
[21]
[22]
between multivariate biological patterns and environmental variables.
Estuar. Coast. Shelf Sci. 56, 717^734.
Muyzer, G., DeWaal, E., Uitterlinden, G. and Ingeny, B. (1993)
Pro¢ling of complex microbial populations by denaturing gradient
gel electrophoresis analysis of polymerase chain reaction ampli¢ed
genes coding for 16S rRNA. Appl. Environ. Microbiol. 59, 695^700.
Buchholz-Cleven, B.E.E., Rattunde, B. and Straub, K.L. (1997)
Screening for genetic diversity of isolates of anaerobic Fe(II) oxidising bacteria using DGGE and whole cell hybridisation. Syst. Appl.
Microbiol. 20, 301^309.
Murray, A.E., Hollibaugh, J.T. and Orrego, C. (1996) Phylogenetic
compositions of bacterioplankton from two California estuaries compared by denaturing gradient gel electrophoresis of 16S rDNA fragments. Appl. Environ. Microbiol. 62, 2676^2680.
Nubel, U., Engelen, B., Felske, A., Snaidr, J., Wieshuber, A.,
Amann, R.I., Wolfgang, L. and Backhaus, H. (1996) Sequence heterogeneities of genes encoding 16S rRNAs in Paenibacillus polymyxa
detected by temperature gradient gel electrophoresis. J. Bacteriol.
178, 5636^5643.
Satokari, R.M., Vaughan, E.E., Akkermans, A.D.L., Saarela, M. and
de Vos, W.M. (2001) Bi¢dobacterial diversity in human faeces detected by genus speci¢c PCR and denaturing gradient gel electrophoresis. Appl. Environ. Microbiol. 67, 504^513.
Martin-Laurent, F., Philippot, L., Hallet, S., Chaisson, R., Germon,
J.C., Soulas, G. and Catroux, G. (2001) DNA extraction from soils:
old bias for new microbial diversity analysis methods. Appl. Environ.
Microbiol. 67, 2354^2359.
Schmalenberger, A., Schwieger, F. and Tebbe, C.C. (2001) E¡ect of
primers hybridising to di¡erent evolutionarily conserved regions of
the small subunit rRNA gene in PCR based microbial community
analyses and genetic pro¢ling. Appl. Environ. Microbiol. 67, 3557^
3563.
Suzuki, M.T. and Giovannoni, S.J. (1996) Bias caused by template
annealing in the ampli¢cation of mixtures of 16S rRNA genes by
PCR. Appl. Environ. Microbiol. 62, 625^630.
Rochelle, P.A., Cragg, B.A., Fry, J.C., Parkes, R.J. and Weightman,
A.J. (1994) E¡ect of sample handling on estimation of bacterial diversity in marine sediments by 16S rRNA gene sequence analysis.
FEMS Microbiol. Ecol. 15, 215^226.
Roling, W.F.M., van Breukelen, B.M., Braster, M., Goeltom, M.T.,
Groen, J. and van Verseveld, H.W. (2000) Analysis of microbial
communities in a land¢ll leachate polluted aquifer using a new method for anaerobic physiological pro¢ling and 16S rDNA based ¢ngerprinting. Microb. Ecol. 40, 177^188.
Schauer, M., Massana, R. and Pedros-Alio, C. (2000) Spatial di¡erences in bacterioplankton composition along the Catalan coast (NW
Mediterranean) assessed by molecular ¢ngerprinting. FEMS Microbiol. Ecol. 33, 51^59.
van Hannen, E.J., Mooij, W., van Agterveld, M.P., Gons, H.J. and
Laanbroek, H.J. (1999) Detritus dependent development of the microbial community in an experimental system: qualitative analysis by
denaturing gradient gel electrophoresis. Appl. Environ. Microbiol.
65, 2478^2484.
Boon, N., De Windt, W., Verstraete, W. and Top, E.M. (2002) Evaluation of nested PCR-DGGE (denaturing gradient gel electrophoresis) with group-speci¢c 16S rRNA primers for the analysis of bacterial communities from di¡erent wastewater treatment plants. FEMS
Microbiol. Ecol. 39, 101^112.
Moeseneder, M.M., Arrieta, J.M., Muyzer, G., Winter, C. and
Herndl, G.J. (1999) Optimization of terminal restriction fragment
length polymorphism analysis for complex marine bacterioplankton
communities and comparison with denaturing gradient gel electrophoresis. Appl. Environ. Microbiol. 65, 3518^3525.
Maidak, B.L., Cole, J.R., Lilburn, T.G., Parker, C.T.Jr., Saxman,
P.R., Farris, R.J., Garrity, G.M., Olsen, G.J., Schmidt, T.M. and
Tiedje, J.M. (2001) The RDP-II (Ribosomal Database Project). Nucleic Acids Res. 29, 173^174.
FEMSEC 1532 7-7-03
S.M. Powell et al. / FEMS Microbiology Ecology 45 (2003) 135^145
[23] Altschul, S.F., Gish, W., Miller, W., Myers, E.W. and Lipman, D.J.
(1990) Basic local alignment search tool. J. Mol. Biol. 215, 403^410.
[24] Felsenstein, J. (1995) PHYLIP v.3.57c (phylogenetic inference program package). University of Washington, Seattle, WA.
[25] Keswani, J. and Whitman, W.B. (2001) Relationship of 16S rRNA
sequence similarity to DAN hybridization in prokaryotes. Int. J. Syst.
Evol. Microbiol. 51, 667^678.
[26] Singleton, D.R., Furlong, M.A., Rathbun, S.L. and Whitman, W.B.
(2001) Quantitative comparisons of 16S rRNA gene sequence libraries from environmental samples. Appl. Environ. Microbiol. 67,
4374^4376.
145
[27] Bowman, J.P., Rea, S.M., McCammonm, S.A. and McMeekin, T.A.
(2000) Diversity and community structure within anoxic sediment
from marine salinity meromictic lakes and a coastal meromictic marine basin, Vestfold Hills, Eastern Antarctica. Environ. Microbiol. 2,
227^237.
[28] Ravenschlag, K., Sahm, K., Pernthaler, J. and Amann, R. (1999)
High bacterial diversity in permanently cold marine sediments.
Appl. Environ. Microbiol. 65, 3982^3989.
[29] Ferrari, V.C. and Hollibaugh, J.T. (1999) Distribution of microbial
assemblages in the Central Arctic Ocean Basin studied by PCR/
DGGE: analysis of a large data set. Hydrobiology 401, 55^68.
FEMSEC 1532 7-7-03