The original publication is available at: Environmental Pollution 203, 165 – 174 (2015)
Metabarcoding of benthic eukaryote communities predicts the ecological condition of estuaries
Anthony A. Charitona, *, Sarah Stephensona, Matthew J. Morganb, Andrew D.L. Stevenc, Matthew J.
Colloffb, Leon N. Courtb, Christopher M. Hardyb
a CSIRO Oceans and Atmosphere, Locked Bag 2007, Kirrawee, NSW 2232, Australia
b CSIRO Land and Water, GPO Box 1700, Canberra, ACT 2601, Australia
c CSIRO Oceans and Atmosphere, GPO Box 2583, Brisbane, QLD 4001, Australia
DNA-derived measurements of biological composition have the potential to produce data covering all of
life, and provide a tantalizing proposition for researchers and managers. We used metabarcoding to
compare benthic eukaryote composition from five estuaries of varying condition. In contrast to traditional
studies, we found biotic richness was greatest in the most disturbed estuary, with this being due to the
large volume of extraneous material (i.e. run-off from aquaculture, agriculture and other catchment
activities) being deposited in the system. In addition, we found strong correlations between composition
and a number of environmental variables, including nutrients, pH and turbidity. A wide range of taxa
responded to these environmental gradients, providing new insights into their sensitivities to natural and
anthropogenic stressors. Metabarcoding has the capacity to bolster current monitoring techniques,
enabling the decisions regarding ecological condition to be based on a more holistic view of biodiversity.
Keywords: biomonitoring, metabarcoding, sediments, DNA, eukaryotes, high-throughput sequencing,
18S rRNA, indicator taxa, threshold analysis
The increasing human population and its activities are having pronounced deleterious effects on the
ecological condition of the world's estuaries (Rabalais et al., 2009). These activities degrade the physical
environment and modify the chemical composition of the water column and sediments and their
associated biota (Davis and Koop, 2006). Ultimately, such activities are expressed as distinct changes in
ecological composition and function (Dauer et al., 2000; Hooper et al., 2012). In Australia, more than
85% of the population of 22 million live within 50 km of the coast (ABS, 2003). With a population
increase of 82% projected by 2056 (ABS, 2003), the pressures on estuarine environments in rapidly
developing coastal regions such as southeast Queensland are likely to increase markedly.
In order to mitigate the pace of environmental degradation, fundamental information on the chemical,
physical and ecological characteristics and components of estuaries is required. However, many of the
variables which drive the ecology of estuaries are difficult to define and vary greatly across space and
time (Morrisey et al., 1992; Wiens, 1989; Ysebaert and Herman, 2002). The most commonly monitored
ecological component of estuaries is the macrobenthos, with many studies demonstrating its
responsiveness to a range of natural and anthropogenic variables (Johnston and Roberts, 2009). This
approach can lead to management decisions being made on the assumption that the macrobenthos
accurately represents overall ecological condition (Chariton et al., 2010a), despite the knowledge that the
meio- and microbiota are far more species-rich, have a greater diversity of life-histories and ecological
niches, and are often more responsive to environmental change (Austen and Warwick, 1989; Kennedy
and Jacoby, 1999). The inclusion of ecological data derived from these elements of the biota would
provide a more representative, informative ecological picture.
In recent years, there have been considerable advances in applying DNA-based diversity methods
using high-throughput sequencing (Baird and Hajibabaei, 2012; Taberlet et al., 2012a), commonly
referred to as metabarcoding (Taberlet et al., 2012b). Metabarcoding provides previously unattainable
insights into communities and ecosystems, aiding our understanding of them. The approach has proven
especially useful in deriving compositional data from samples containing organisms that are difficult to
identify because of small size, cryptic habits, their occurrence in the form of propagules (e.g. spores and
zoocysts) or a lack of traditional identification keys (Medinger et al., 2010; Valentini et al., 2009).
In this study, we used high-throughput sequencing of the 18S rRNA gene to examine sub-macro
benthic biotic composition of five estuaries in eastern Australia. These estuaries have been routinely
sampled since 2000 as part of a larger monitoring program, but due to time and cost constraints the
ecological health of these estuaries is determined using abiotic surrogates rather than ecological data. Our
initial aim was to examine whether DNA-based eukaryotic composition could differentiate between
estuaries. Secondly, we explored the relationships between the eukaryotic communities and
environmental gradients observed among the estuaries. Finally, we examined whether metabarcoded
eukaryotic data has the potential to produce relevant ecological information which can be used to further
develop DNA-based approaches for the routine monitoring of estuarine sedimentary environments.
In February 2010, we sampled five estuaries (Noosa, Maroochydore, Pine, Logan and Currumbin) in
south-east Queensland, Australia (Fig. 1). All estuaries are monitored monthly by the Queensland
Government as part of the Ecosystem Health Monitoring Program (EHMP) (http://www.health-e-
waterways.org). The EHMP includes an Ecological Health Index based on algal productivity derived
from measurement of chlorophyll a, concentrations of dissolved oxygen, major nutrients and turbidity.
The conformation of these variables with national guideline values, together with estimates of seagrass
and riparian vegetation cover is used to develop an annual report card for each estuary (Table S1 in
The five estuaries were located no more than 190 km apart (Fig. 1) and represent a range of ecological
conditions (Table 1). There were large differences in morphology between the five estuaries, with the
Currumbin Creek considerably smaller than the others (Table 1). Within each estuary, five sites were
sampled, and in all but two cases (one in each the Noosa and Currumbin) the sites have been routinely
monitored under the EHMP.
Collection and analysis of environmental variables
Sediment collection was confined to non-sandy substrates. Five sediment samples were collected at each
site from ca. 2 m below low water using a Van Veen grab. Sub-samples were taken from the surficial
layer (1.5-2 cm) of each sample for DNA, grain size and total organic carbon analysis. All samples for
DNA analysis were transferred into clean 50 mL Greiner tubes and placed on ice immediately, then
frozen within 6 h of collection and thawed only just prior to DNA extraction. All materials used for the
collection and storage of DNA samples were pre-rinsed for at least 24 h in 5% sodium hypochlorite, and
rinsed thoroughly five times with Milli-Q water (Millipore, Academic Water Systems, Australia). To
minimise cross contamination, sediments were only sub-sampled from the centre of each grab sample.
The physico-chemical properties of the water column were measured at each sampling site
approximately 0.5 m above the sediment surface using a calibrated YSI 6920 multi-sonde. At all EHMP
sites (excluding one site in both the Noosa and Pine), water samples were collected for nutrient and
chlorophyll a analyses. Water samples for nutrient analysis were filtered upon collection, with the filtrate
stored in clean foil-wrapped containers stored on ice. Total phosphorus, filterable reactive phosphorus,
total nitrogen, organic nitrogen, inorganic nitrates, ammonia and chlorophyll a analyses was performed
using standard methods (Clesceri et al., 1998). Total organic carbon (TOC) and grain size analysis for the
following grain size classes: <63 μm (fines), 0.63 μm- mm (sand), >1 mm (coarse) were performed as
previously described (Chariton et al., 2010c).
DNA extraction, amplification and sequencing
DNA was extracted from 1.5 g of sediment and purified using UltraCleanSoil DNA extraction kits
(MO BIO, Carlsbad, CA) following the manufacturer's protocols. In addition to the sediment samples,
three internal reference samples containing sixteen clones from a range of eukaryotic taxa were also
processed, as previously described (Morgan et al., 2013). Polymerase chain reaction (PCR)
amplification of a 200e500-bp fragment of the 18S rRNA gene was carried out with the ‘universal’
primers All18SF-TGGTGCATGGCCGTTCTTAGT and All18SR-CATCTAAGGGCATCACAGACC
(Hardy et al., 2010), and sample preparation was conducted as previously described (Baldwin et al.,
2013). Sequencing was performed by the Australian Genome Research Facility (St Lucia, Queensland)
using a single plate of Roche 454 GS FLX Titanium. Demultiplexing and the removal of potential PCR
artefacts, sequencing errors and chimeric sequences were performed using the Amplicon Pyrosequence
Denoising Program (APDP) (Morgan et al., 2013). Taxon identification of each unique sequence,
herein referred to as a Molecular Operational Taxonomic Unit (MOTU) was inferred using the RDP
classifier with the SILVA 18S rRNA database (release 113) (www.arb-silva.de).
As there is a weak statistical relationship between the number of sequence reads and organism biomass
or abundance (Egge et al., 2013), all MOTU data were converted to presence or absence prior to
computation (Chariton et al., 2014). Ordination of MOTU data was performed by non-metric
multidimensional scaling (nMDS) using the Jaccard similarity coefficient in the Primer 6 þ statistical
package (Plymouth Marine Laboratory, UK). Statistical differences between estuaries were tested by a
two-factor permutational multivariate analysis of variance (PERMANOVA), with ‘sites’ nested within
‘estuary’. Differences between treatments were identified by pairwise a posteriori tests based on 9999
random permutations. The proportions of explained variation at spatial scales of estuary, site and
residual were calculated using the procedure described by Quinn and Keough (2002). Differences in
richness of total MOTU and dominant taxonomic groups were examined using a two-factor nested
ANOVA. Residuals were assessed for skewness, kurtosis, and omnibus normality using D'Agostino's
tests (D'Agostino et al.,1990) with homogeneity of variances examined using a modified Levene equal
variance test (Levene, 1960). When assumptions of homogeneity were violated, appropriate
transformations were performed (Sokal and Rolf, 1995). In cases in which the data remained
heteroscedastic, the level of statistical significance was set at P < 0.01. All ANOVAs were performed
in NCSS v8 (NCSS, Kaysville, UT). MOTUs indicative of each estuary and combinations of estuaries
were identified using the R package Indispecies, with Indictor Values (IV) reflecting both the
conditional probability of the MOTU as an indicator of a particular estuary and the probability of
finding the MOTU in samples associated with the estuary. In addition to the package's multipatt
function, the signassoc function was used to determine whether the occurrences of each potential
indicator MOTU identified by the multipatt analysis were random and to correct for multiple testing.
The relationships between community composition and environmental variables were examined
using distance-based linear models (DISTLM) (Legendre and Anderson,1999). In order to match the
number of biological and physico-chemical samples, i.e. one sample per site, the similarity matrix for
the biological data was recalculated using the distance between centroids for each site. The two sites
where nutrient data were unavailable were discarded from the analyses. Following the procedure of
Bellchambers et al. (2011), an initial analysis was performed using forward selection of all
environmental variables with the goodness-of-fit examined using Akaike's information criterion. The
most parsimonious model was re-run using only the variables selected for this model and distance-
based redundancy analysis (dbRDA) was performed to visualise the influence of predictor variables
identified by the DISTLM. Threshold Indicator Taxa ANalysis (TITAN) in R was used to estimate
MOTUs whose occurrences declined (z scores) or increased (þz scores) and to examine community
thresholds along the significant environmental gradients identified from the DISTLM's sequential tests
(total phosphorus, mono-nitrogen oxides, turbidity and pH). TITAN is an extension of indicator
analysis which partitions the biological data into two groups at the value of a predictor variable that
maximizes the association of each taxon (or MOTU) with each side of the partition. Using standardized
z-scores, TITAN can distinguish those taxa whose occurrences declined (z scores) or increased (þz
scores) along the environmental variable. Prior to running TITAN, all MOTUs which were observed in
fewer than three samples or and in more than 95% of samples were removed. Peak values in z and þz
scores were used to respectively determine negative and positive community responses to the
environmental variable. Bootstrapping was used to estimate the confidence limits of the change points
(King and Baker, 2010).
The pyrosequencing run produced >1.3 million sequences. The dataset for the sample estuaries contained
2937 MOTUs after all potentially erroneous sequences had been removed and sample rarefaction to 2093
reads. The data are accessible via the CSIRO data portal http://dx.doi.org/10.4225/08/52DF4D8008B99.
For all three internal reference samples, the APDP bioinformatics pipeline correctly identified only the
16 clone sequences as valid. As such, we consider our measurement of sample richness to reflect the
‘true’ variant richness of the targeted region. The largest proportion of MOTUs that could be confidently
assigned to a kingdom (59%) belonged to the Metazoa (16%). The eukaryote supergroups Chromista and
Chromalvelolata each contributed 10% to the total taxon richness and fungi, Rhizaria and Viridiplantae
Physico-chemical attributes of the estuaries
A summary of the physico-chemical attributes of the estuaries and relevant Australian water quality
guideline values are presented in Supplementary Material Tables S2 and S3. All estuaries had total
nitrogen (TN) concentrations that exceeded the Australian guideline value (ANZECC/ARMCANZ,
2000), with all sites within the Logan Estuary exceeding this value by at least four-fold. With the
exception of the Noosa Estuary, guideline values for mono-nitrogen oxides (NOx), total reactive
phosphorus (TRP) and total phosphorus (TP) were exceeded in almost all samples. On average, the
Logan Estuary had TRP and TP concentrations 32 and 10 times greater than the guideline values,
respectively. Chlorophyll a concentrations exceeded guideline values in both the Maroochydore and Pine
estuaries. It is emphasised that the physico-chemical attributes of the estuaries are highly variable, with
measurements taken at the time of sampling occasionally deviating from the mean values observed over
of the previous six months (Figs. S1 and S2 in Supplementary material). Most notably, in the Pine and
Currumbin, pH and concentrations of total nitrogen and dissolved organic nitrogen at several sites were
considerably elevated, with conductivity also being substantially below the long term mean in the some
of the Pine Estuary sites. These deviations reflect the influence of a substantial rainfall event which
occurred during sampling. There were some inter-estuarine variation in physicochemical properties
(Tables S2 and S3 in Supplementary material), but the Noosa, Maroochydore, Pine and Currumbin
mostly had similar profiles while those of the Logan Estuary were markedly different: the overlying
waters were more turbid (Secchi depth) and hypoxic. Furthermore, the low conductivity and pH of the
Logan water column indicates that the estuary was being driven by freshwater inputs at the time of
sampling. The environmental data indicates that sites from the Noosa Estuary were relatively
homogeneous and contained the lowest concentrations of nutrients. At the time of sampling, the Pine,
Maroochydore and Currumbin estuaries contained environmentally significant concentrations of
nutrients, i.e. exceeded regional trigger values, with the relative contribution of each nutrient and its
inorganic and organic components varying between estuaries. The water of the Logan Estuary was highly
eutrophic and turbid.
Ecological comparisons between estuaries
As illustrated by the accumulation curve (Fig. S3), 125 replicate samples were sufficient to account for
>99.8% of the 2941 MOTUs estimated (Chao 2) that occur in the five estuaries. The 25 replicates
collected in each estuary accounted for between 79% (Pine) and 90% (Logan) of the estimated eukaryote
richness. Total MOTU richness varied significantly within the estuaries (ANOVA: F = 4.43, P < 0.001).
The Logan Estuary (mean 391 ± 8 SE) had higher MOTU richness (ANOVA: F = 7.69, P < 0.001) than
the other four estuaries, all of which had similar richness (Noosa 294 ± 11; Maroochydore 259 ± 9; Pine
301 ± 6; Currumbin 256 ± 13). The high MOTU richness of the Logan Estuary was due to richer fungal
(ANOVA: F = 20.78, P < 0.001) and protozoan (ANOVA: F = 10.2, P < 0.001) communities.
Chromoalveolata communities were richer in the Logan Estuary (ANOVA: F = 8.0, P < 0.001).
Metazoan richness was lower in the Noosa and Maroochydore estuaries, with the Pine and Currumbin
estuaries containing the richest metazoan communities (ANOVA: F = 9.6, P < 0.001).
The composition of the benthic biota varied within (PERMANOVA: F = 2.90, P = 0.001) and between
the estuaries (PERMANOVA: F = 5.71, P = 0.001), with post-hoc analysis identifying that all five
estuaries contained significantly different assemblages (P < 0.01). Spatial scale was shown to influence
variability, with approximately half of the variation occurring from sample to sample (residuals =
0.52%). Variation in composition at the largest scale of ‘estuary’ (28%) was more important than that at
the intermediate scale of ‘site’ (20%).
As illustrated by the nMDS ordination plot (Fig. 2), the greatest disparity in benthic eukaryote
composition was between the largely unmodified Noosa Estuary and highly eutrophic Logan Estuary
(similarity 5.6%), with both containing markedly different compositions to the other estuaries. While the
compositions of the other three intermediately modified estuaries did differ, their compositions were
more similar to each other (similarity 15.1-15.9%) than to either the Noosa (similarity 7.0-10.8%) or
Logan (similarity 8.1-8.8%).
Indicator analysis identified 426 MOTUs which were characteristic of the estuaries at the time of
sampling (Fig. 3). The largest proportions of these were in the Logan (38%) and Noosa (28%) estuaries.
MOTUs with the ten highest Indicator Values (IV) for each estuary are provided in the Supplementary
material (Table S4).
Relationships between benthic communities and environmental variables
The most parsimonious distance-based linear regression model which used 15 of the original 16 variables
(excluding percentage silt) explained 71.3% of the total variation in benthic community structure. The
first dbRDA coordinate axis explained 19.6% of the total variation in the benthos (Fig. 4), with high
concentrations of nutrients (total phosphorus, total nitrogen, organic nitrogen) and high turbidity clearly
separating communities of the Logan Estuary from the others. The second dbRDA coordinate axis
explained 14.4% of the total variation and indicated that compositional changes within Noosa and
Maroochydore estuaries were correlated with variation in conductivity, pH and temperature. Oxygen
saturation, concentrations of NOx and ammonia, and primary production assisted in distinguishing the
benthic communities from the Pine and Currumbin from the Noosa Estuary. Benthic community
composition was significantly correlated with TP (18.6%, P = 0.01), NOx (7.9%, P = 0.01), turbidity
(8.1%, P = 0.004) and pH (6.12%, P = 0.009). The first three of these variables were symptomatic of
eutrophication and clearly separated the Logan Estuary's benthos from the rest (Fig. 4).
Threshold analysis (TITAN) identified 330 MOTUs whose occurrence (i.e. number of times identified
as present) responded negatively (z scores) to increasing TP concentration (Table S5 in the
Supplementary material), with 40% of those which could be taxonomically assigned belonging to
Bacillariophyta (mostly Bacillariophyceae). The most sensitive indicators of elevated total phosphorus
(i.e. largest z scores) observed across the five estuaries were all bacillariophytes and foraminiferans (Fig.
5). As indicated by their synchrony in change points and consistently small percentile ranges, declines in
the presence (i.e. occurrence) of phosphorus sensitive taxa were generally abrupt, with the most
pronounced loss of sensitive taxa occurring at a TP concentration of 24 mg L-1 (22-34 mg L-1, 5-95
percentile; Fig. 5). In contrast, 465 MOTUs had significant þ z scores in response to increases in TP
(Table S5). Many of these were identified as potential indicator MOTUs for the Logan Estuary. The
MOTUs with the largest þ z scores included Dinophyceae, Coscinodiscophyceae, Annelida, Gastrotricha,
Rotaliida and Micronuclearia (Fig. 5). The confidence limits for these were broad indicating occurrence
over a wide range of TP concentrations. The largest peak in TP tolerant MOTUs was at 100 mg L-1 (100-
290 mg L-1), with smaller peaks also occurring at 180 and 260 mg P L-1. The community threshold (or
change point), i.e. the point where a maximum change in composition occurred was at TP = P 185 mg L-1
(26-269 mg L-1).
There were some strong similarities between the responses of MOTUs to TP and to NOx and turbidity
(Table S5 in the Supplementary material), with 127 MOTUs responding negatively to all three variables
which are generally elevated in eutrophic estuaries. These MOTUs all shared small tolerance ranges (5-
95 percentiles) and a high representation of MOTUs from Bacillariophyceae, Arthropoda (notably
Crustacea) and Rotaliida (Table S5 in the Supplementary material). There was peak decline in MOTUs
when NOx concentrations reached 5 μg N/L (4-12 μg N/L). No such clear synchronous decline was
observed for turbidity-sensitive MOTUs, with the change point occurring 16 NTU with 9-95 percentiles
encompassing a wide turbidity range (15-68 NTU). Some 265 MOTUs that responded positively to
increases in TP also responded positively to increases in NOx and turbidity (Table S5). They were from
Coscinodiscophyceae, Heliozoa (Heterophryidae), Ciliophora, Ascomycota and Chytridiomycota,
Apusozoa (Rigifilida), Choanozoa and Kinorhyncha (Echinoderes spp). The sum (z+) change points for
NOx and turbidity occurred at 100 μg N/L (79-130 μg N/L) and 134 NTU (49-145 NTU), respectively.
The community threshold for NOx occurred at the same concentration (5 μg N/L, 4-15 μg N/L) as the
sum (z+) change point. The most pronounced shifts in MOTU composition occurred when the overlying
waters reached a turbidity of 101 NTU (49-145 NTU).
The observed pH gradient reflects the transition from fresh to marine waters. Some 433 MOTUs had
significant z scores and occurred less frequently as the water became more marine (Table S5 in the
Supplementary material). Declines began at pH 7 and were maximal at pH 7.06 (6.96-7.31, 5-95
percentile), coinciding with the mean community change-point (pH = 7.06, 6.96-7.32). Some 9% of the
Bacillariophyta MOTUs showed pronounced declines in occurrence as pH increased (Table S5). The
most sensitive taxa were the protistan Micronuclearia, Dinophyceae and Thalassiosirophycidae
(Coscinodiscophyceae) (Fig. 5). The relatively broad percentile ranges of many of the MOTUs indicated
they could persist at a pH 0.5 greater than their mean change points. Some 279 MOTUs responded
positively to an increase in pH (þz scores), with a change point occurring at pH 8.04 (7.91-9.08). MOTUs
that showed increased occurrence with increased pH of waters were generally observed within a
relatively small pH range. Some 33% of bacillariphytes responded positively to increased pH, as did
Apicomplexa, Ciliophora, Cnidaria, Foraminifera and Labyrinthista (Table S5). The most responsive
MOTUs to the increase in pH were Rotaliida, Bacillariophyceae and Arthropoda (Fig. 5).
Comparisons between estuaries
Taxonomic richness is a commonly used index for comparing benthic communities, with the assumption
that endemic taxon richness will be lower in disturbed environments (Dauer, 1993). Previously, we have
cautioned against the use of MOTU richness in DNA-derived ecological studies, as bioinformatic
pipelines generally overly inflate true sequence richness (Chariton et al., 2014). As shown in the present
study, and that by Morgan et al. (2013), an accurate measure of sequence richness (i.e. the number of true
variants of the targeted region) can be obtained using the bioinformatics pipeline APDP, and it is timely
to examine the potential usefulness of MOTU richness as an ecological metric. The underpinning trend
from estuarine macrobenthic studies is that eutrophication leads to a pronounced reduction in richness,
generally in the order of 30% (Johnston and Roberts, 2009). The Logan Estuary is significantly modified,
with the adjacent catchment (3076 km2) now 12% urbanized and the remainder under a variety of
agricultural uses including beef, chicken and lawn farms in the mid reaches; prawn farms, sugar cane
farms and rural residential in the estuarine reaches. There are also two major sewage treatment plants
(STPs) in this system: The Loganholme and Beenleigh STPs, which collectively discharge an average
920 ML/ day into the Logan Estuary. As a consequence the Logan estuary is considered highly
eutrophic. Nevertheless, we observed relatively high biotic richness in the Logan Estuary, with richness
being 25-35 percent greater than that of the other four less disturbed estuaries. We have observed a
similar trend when comparing a heavily contaminated and moderately disturbed estuary (Chariton et al.,
2010b). Protozoan and fungal richness in particular was greater in the Logan than the other estuaries. We
observed no difference in MOTU richness between the other four estuaries. This was surprising given the
marked differences in catchment size, land use, salinity gradients, nutrient concentrations and other
variables commonly shown to alter composition (Archambault and Bourget, 1996; Dauer et al., 2000;
Remane and Schlieper, 1971; Witman et al., 2004).
Many MOTUs specific to the Logan Estuary were associated with the breakdown of detritus material
(Sandgren et al., 1995). These included Oomycetes which have been shown to proliferate in
environments subjected to high organic inputs, aggressively outcompeting fungal species (Newell and
Fell, 1995, 1997) and Order Pythiales, parasites known to affect humans, fish, plants, fungi and insects
(Yu, 2001). The Logan Estuary also contained large volumes of catchment-derived material, as sampling
was performed during an unseasonal high rainfall event (255 mm in February, 2010; median since 1960
¼ 107 mm) (http://www.bom.gov.au/climate/data/; sampling station 040312). Consequently, the high
richness in the Logan estuary was due to MOTUs derived from extraneous sources (e.g. sewage
treatment plants and agriculture) as well as from organisms which were inherently part of the estuary's
biocoenosis. By contrast, in the Noosa Estuary, a large proportion of the richness was associated with
diatoms and a less diverse metazoan fauna. Collectively, our findings suggest that total MOTU richness
is not a sensitive indicator of ecological condition. While a greater breadth of studies is required to
support or refute these patterns, the initial signs suggest that the richness of key taxonomic groups
appears more sensitive than total MOTU richness to environmental conditions.
Diatoms along with protozoans, Chromoalveolata and fungal phyla exhibit characteristics which
potentially make them informative indicators of ecological condition. These include their trophic
positioning and responsiveness to a range of environmental contaminants (Harding et al., 2005;
McCormick and Cairns,1994). Although there is a wealth of information describing broad univariate
responses in diatoms to eutrophication, there is currently a paucity of information regarding the structural
attributes of naturally functioning diatom communities and how these may change in response to
eutrophication and other environmental stressors. Pronounced changes in diatom composition have
frequently been shown to correlate with environmental degradation (Desrosiers et al., 2013; Snoeijs,
2013). In our study, diatom composition (derived from presence/absence) was a strong feature of the
largely unmodified Noosa Estuary, with more than 40% of the indicator MOTUs associated with families
within Bacillariophyceae. Palaeoecological evidence capturing 2500 years of history of the Great Lakes
further supports strong correlations between eutrophication and land-use and a decline in diatom richness
(Cooper and Brush, 1993). By contrast, only 7-12 % of the indicator MOTUs for the other estuaries were
diatoms, with these more commonly associated with Coscinodiscophyceae.
While still containing different compositions, the three estuaries (Maroochydore, Pine and Currumbin)
which have been historically shown to be in similar ecological condition (EHMP scores C to C-), were
more similar to each other than either the Noosa or Logan estuaries. As one of the metrics used to
calculate indicator scores is a MOTU's specificity to a particular estuary (De Caceres et al., 2010), the
indicator scores for the MOTUs from these three estuaries were generally lower than those from the
Logan and Noosa estuaries. MOTUs whose increased occurrences aided discrimination between the
Maroochydore, Pine and Currumbin estuaries were derived from a wide breadth of taxonomic groups. In
particular, the Pine estuary had a number of metazoan indicator MOTUs, including nematodes, rotifers
and turbellarians. Indicator MOTUs associated elsewhere for bacillariophytes and ciliophorans were
common to all three estuaries.
We found that the largest proportion of variance occurred at the smallest spatial scale, that is, between
samples. This suggests that benthic community composition is driven by small-scale localised processes,
e.g. competitive interactions, habitat heterogeneity and disturbance events (Archambault and Bourget,
1996; Thrush et al., 1996; Whitlatch and Zajac, 1985). While Anderson et al. (2005) proposed that
similar patterns are likely to occur in other biological assemblages, the present study is the first to
demonstrate the preservation of this pattern using presence/absence compositional data simultaneously
derived from numerous phyla. The low level of variation which occurred at the site scale (20%) reflects
the lack of true independence between sampling sites due to the ebb and flow of the overlying waters.
Variation at the estuary scale was more pronounced than at the site scale. While all five estuaries were
located within the southeast Queensland biogeographical zone and shared the same oceanic input (Pacific
Ocean) e although the Pine and Logan estuaries are buffered by Moreton Bay e marked differences in
catchment size, channel morphology, land use, salinity profiles and nutrient concentrations undoubtedly
contributed to biological variation at the estuary scale. Biological connectivity within estuaries can occur
via active migration, passive movement via the water column, or adherence to other organisms, sediment
and organic material. Unique to DNA surveys is that patterns in intra-estuary connectivity are
undoubtedly distorted by the inadvertent sequencing of DNA which is either adhered or retained within
the gut contents of the targeted biota (Chariton et al., 2010b).
Relationships between biological composition and environmental variables
Whilst nutrient inputs are critical for estuarine functioning, extensive changes in the land use of south-
east Australia have profoundly altered the timing of riverine flows and their nutrient loadings into the
region's estuaries, altering the ecological compositions of environments and potentially creating
conditions ideal for cyanobacterial blooms (Davis and Koop, 2006; Quigg et al., 2010; Roy et al., 2001).
In the present study, elevated concentrations of total nitrogen were reported in all five estuaries, however,
elevated concentrations of nitrogen (total, organic, ammonia) and phosphorus (total and filterable
reactive) were clearly greater in the Logan Estuary, reflecting its current land uses. A majority of the
variables which explained a significant proportion of the variation in benthic community data were either
directly (e.g. total phosphorus and mono-nitrogen oxides) or indirectly (turbidity) associated with
excessive nutrient inputs (Fig. 4). Consequently, it was unsurprising that the constrained analysis clearly
separated the Logan Estuary's benthic communities from those of the other four estuaries (Fig. 4). To a
lesser degree, differences in NOx concentrations differentiated the Currumbin and the Pine estuaries
from the Maroochydore and Noosa estuaries.
Threshold analysis (TITAN) indicated that there was a pronounced decline in TP sensitive taxa
(change point in z scores) when the mean overlying water reached 24 μg P/L (23-34 μg P/L, 5 and 95
percentiles), which is close to the Australian default water quality guideline for south-eastern Australian
estuaries of 30 mg P/L (ANZECC/ARMCANZ, 2000). As measurements of TP and other environmental
variables, including their co-variates, can vary dramatically across a range of temporal scales (e.g. tides,
seasons and run-off events) it is highly unlikely that there is absolute synchrony between water column
concentrations of nutrients and the turnover rates of benthic communities. As such, in the current study
the derived threshold values are considered to be notional, with the additional information regarding the
tolerances of MOTUs being obtained from the width of the percentiles created from the threshold
Diatoms were the most responsive group to TP, with the occurrences of a large proportion of MOTUs
declining with increasing concentrations of TP. With the exception of Thalassiosira, a common attribute
of the most sensitive MOTUs was their relatively small tolerance ranges (Fig. 5), with the occurrences of
the taxa declining rapidly as indicated by their narrow percentile range. In contrast, the MOTUs which
responded favourably to TP, NOx and turbidity were generally present across a wide range of
concentrations (Fig. 5). While nutrient addition has been shown to stimulate algal biomass (Anderson et
al., 2002; Hecky and Kilham, 1988), our findings emphasize that compositional change also occurs, with
diatom richness being substantially reduced by elevated concentrations of TP. The ecological
ramifications of a compositional shift in diatoms is difficult to define, however, because of their rapid
generation time (days to hours), evidence of a shift in diatom composition may be a precursor for
subsequent changes in ecological integrity, e.g. trophic bottle necks and cyanobacteria blooms
(Desrosiers et al., 2013; Logan and Taffs, 2013; Snoeijs, 2013). The latter can be determined by the
inclusion of the chloroplast 16S rRNA gene in subsequent studies.
There were many commonalities between those MOTUs which responded positively to TP, NOx and
turbidity (Table S4 in the Supplementary material). In addition to the Apusozoa and dinoflagellate
MOTUs, these included relatively large proportions of the MOTUs representative of Heliozoa,
Ciliophora and Choanozoa.
Highly diverse protozoan communities are commonly associated with sewage treatment plants
(Madoni, 1994), and consequently their observed increase in occurrence is most likely to be associated
with the sewage treatment plants within the Logan Estuary catchment, a primary source of the estuary's
nutrient enrichment. Further evidence of this was the increase in the occurrence of fungal MOTUs from
the Phyla Ascomycota and Chytridiomycota. Increases in the abundance of chytrid fungi have been
shown to occur in organic materials derived from sewage plants (Novinscak et al., 2009). Interestingly,
some chytrids are parasitic and can cause dramatic shifts in diatom populations, with this phenomenon
potentially contributing to the observed decline in diatoms (Bruning et al., 1992). Collectively, the
findings from the current study show a strong correlation between anthropogenic induced changes to the
estuaries and benthic composition, with the communities shifting from primary producing taxa (e.g.
diatoms) to ones dominated by protozoans and fungi, and other taxa associated with the consumption of
bacteria and the breakdown of organic material.
In those estuaries that were less influenced by nutrients, e.g. Noosa and Maroochydore, compositional
changes were more strongly correlated with pH, with an increase in pH reflecting an increase in the
influence of marine waters (Bianchi, 2012). In general, taxa which preferred low pH waters were found
at a pH below 7.06, with most taxa persisting across a relatively broad pH range. It should be noted that
the Logan Estuary had the lowest pH, and as such, other variables which contributed to the composition
of this estuary (i.e. nutrients and turbidity) may also be contributing to the perceived composition of
lower pH environments. The pH levels can vary dramatically within estuaries, however, as indicated by
the synchrony between the mean z scores and community change point (from TITAN analysis), and the
relatively narrow tolerances of the taxa associated with the higher pH waters, even relatively small
changes can substantially alter biotic composition.
For many researchers, the ability to obtain compositional data from the amplification of DNA/RNA
provides an exciting prospect, increasing the pool of taxa which can be included in biological surveys
and monitoring; enabling the identification of cryptic or decomposed organisms (e.g. in gut contents); as
well as potentially reducing the costs, latency and identification issues associated with traditional
surveys. DNA-based monitoring is in its infancy and as such considerable research is required to further
develop, refine, and evaluate the utility of the approach. One major limitation is the unreliability of
proportional data derived from PCR-based approaches, and hence our use of presence/absence data. To
address this limitation we have recently resequenced three of the estuaries system using a PCR-free
metagenomic approach (Chariton in prep.). Although we have demonstrated the capacity for
metabarcoding to discriminate between the five estuaries, show correlative patterns between composition
and the major environmental variables, and identify those taxa which responded positively and
negatively to the key environmental variables; it is emphasised that the present study was from a single
sampling event and flows can greatly vary. Consequently, repetitive temporal sampling is required to add
credence or refute the observed trends. To address this knowledge gap we resampled the five estuaries in
2012 using an expanded program (additional sites, particulate metals and organic contaminants,
additional genes, and a large sediment volume for DNA extraction). Although there were differences in
estuary specific MOTU indicators, the broad differences among the estuaries and the relationships
between biotic assemblages and the key environmental variables (e.g. total phosphorus, turbidity and pH)
remained unchanged (Chariton in prep.).
The current approach used by the EHMP to establish the ecological condition of estuaries uses data
solely derived from water quality parameters and habitat condition, and excludes ecological data. The
aim of the presented approach is not to replace traditional sampling programs but rather, to add an
additional line of ecological evidence which encompasses a greater breadth of diversity. As with all
monitoring programs, a considerable wealth of data is required to identify predictable patterns and to
understand the ecological ramifications of any observed changes in community composition. Only when
additional data has been obtained and methodological issues refined can the utility of this approach be
Financial support for the project was provided by the CSIRO Oceans and Atmosphere. The authors wish
to thank John Ferris and James Fells (Queensland Environmental Protection Agency) for their assistance
in the collection of samples, Ray Williams (Queensland Department of Environment and Heritage
Protection) for facilitating access to the physico-chemical data, and Chris Moeseneder (CSIRO) for his
ABS, 2003. Regional Population Growth, Australia and New Zealand, 2001e02. Australian Bureau of Statistics,
Anderson, D.M., Glibert, P.M., Burkholder, J.M., 2002. Harmful algal blooms and eutrophication: nutrient sources,
composition, and consequences. Estuaries 25, 704–726.
Anderson, M.J., Connell, S.D., Gillanders, B.M., Diebel, C.E., Blom, W.M., Saunders, J.E., Landers, T.J., 2005.
Relationships between taxonomic resolution and spatial scales of multivariate variation. J. Animal Ecol. 74,
ANZECC/ARMCANZ, 2000. Australian and New Zealand Guidelines for Freshwater and Marine Water Quality.
Vol. 1. Australian and New Zealand Environment and Conservation Council and Agriculture and Resource
Management of Australia and New Zealand, Canberra.
Archambault, P., Bourget, E., 1996. Scales of coastal heterogeneity and benthic intertidal species richness, diversity
and abundance. Mar. Ecol. Prog. Ser. 136, 111–121.
Austen, M.C., Warwick, R.M., 1989. Comparison of univariate and multivariate aspects of estuarine meiobenthic
community structure. Estuarine Coast. Shelf Sci. 29, 23–42.
Baird, D.J., Hajibabaei, M., 2012. Biomonitoring 2.0: a new paradigm in ecosystem assessment made possible by
next-generation DNA sequencing. Mol. Ecol. 21, 2039–2044.
Baldwin, D.S., Colloff, M.J., Rees, G.N., Chariton, A.A., Watson, G.O., Court, L.N., Hartley, D.M., Morgan, M.J.,
King, A.J., Wilson, J.S., Hodda, M., Hardy, C.M., 2013. Impacts of inundation and drought on eukaryote
biodiversity in semi-arid floodplain soils. Mol. Ecol. 22, 1746–1758.
Bellchambers, L., Meeuwig, J., Evans, S., Legendre, P., 2011. Modelling habitat associations of 14 species of
holothurians from an unfished coral atoll: implications for fisheries management. Aquat. Biol. 14, 57–66.
Bianchi, T.S., 2012. Estuarine chemistry. In: Day, J.W.J., Crump, B.C., Kemp, W.M., Yáñez‐Arancibia, A. (eds.),
Estuarine Ecology. Second edition. Wiley, Hoboken, NJ, pp. 39–83.
Bruning, K., Lingeman, R., Ringelberg, J., 1992. Estimating the impact of fungal parasites on phytoplankton
populations. Limnol. Oceanogr. 37, 252–260.
Chariton, A.A., Batley, G.E., Hardy, C.M., 2010a. Improving aquatic ecological assessment. Integr. Environ.
Assessment Manag. 6, 187–189.
Chariton, A.A., Court, L.N., Hartley, D.M., Colloff, M.J., Hardy, C.M., 2010b. Ecological assessment of estuarine
sediments by pyrosequencing eukaryotic ribosomal DNA. Front. Ecol. Environ. 8, 233–238.
Chariton,A.A., Ho, K.T., Proestou, D., Bik,H., Simpson, S.L.,Portis, L.M., Cantwell, M.G., Baguley, J.G., Burgess,
R.M., Pelletier, M.M., Perron, M., Gunsch, C., Matthews, R.A., 2014. A molecular-based approach for
examining responses of eukaryotes in microcosms to contaminant-spiked estuarine sediments. Environ. Toxicol.
Chem. 33, 359–369.
Chariton, A.A., Roach, A.C., Simpson, S.L., Batley, G.E., 2010c. Influence of the choice of physical and chemistry
variables on interpreting patterns of sediment contaminants and their relationships with estuarine macrobenthic
communities. Mar. Freshw. Res. 61, 1109–1122.
Clesceri, L.S., Greenberg, A.E., Eaton, A.D. (eds.), 1998. Standard Methods for the Examination of Water and
Wastewater. 20th edition. American Public Health Association, Washington, DC.
Cooper, S.R., Brush, G.S., 1993. A 2,500-year history of anoxia and eutrophication in Chesapeake Bay. Estuaries
D'Agostino, R., Belanger, A., D'Agostino, R., 1990. A suggestion for using powerful and informative tests of
normality. Am. Statistician 44, 316–321.
Dauer, D.M., 1993. Biological criteria, environmental health and estuarine macrobenthic community structure.
Mar. Pollut. Bull. 26, 249–257.
Dauer, D.M., Ranasinghe, J.A., Weisberg, S.B., 2000. Relationships between benthic community condition, water
quality, sediment quality, nutrient loads, and land use patterns in Chesapeake Bay. Estuaries 23, 80–96.
Davis, J.R., Koop, K., 2006. Eutrophication in Australian rivers, reservoirs and estuaries: a southern hemisphere
perspective on the science and its implications. Hydrobiologia 559, 23–76.
De Chttp://refhub.elsevier.com/S0269-7491(15)00177-3/sref22aceres, M., Legendre, P., Moretti, M., 2010. Improving
indicator species analysis by combining groups of sites. Oikos 119, 1674–1684.
Desrosiers, C., Leflaive, J., Eulin, A., Ten-Hage, L., 2013. Bioindicators in marine waters: benthic diatoms as a tool
to assess water quality from eutrophic to oligotrophic coastal ecosystems. Ecol. Indic. 32, 25–34.
Egge, E., Bittner, L., Andersen, T., Audic, S., de Vargas, C., Edvardsen, B., 2013. Pyrosequencing to describe
microbial eukaryotic community composition, diversity and relative abundance: a test for marine haptophytes.
PLoS One 8, e74371.
Harding, W., Archibald, C., Taylor, J., 2005. The relevance of diatoms for water quality assessment in South
Africa: a position paper. Water SA 31, 41–46.
Hardy, C.M., Krull, E.S., Hartley, D.M., Oliver, R.L., 2010. Carbon source accounting for fish using combined
DNA and stable isotope analyses in a regulated lowland river weir pool. Mol. Ecol. 19, 197–212.
Hecky, R., Kilham, P., 1988. Nutrient limitation of phytoplankton in freshwater and marine environments: a review
of recent evidence on the effects of enrichment. Limnol. Oceanogr. 33, 796–822.
Hooper, D.U., Adair, E.C., Cardinale, B.J., Byrnes, J.E., Hungate, B.A., Matulich, K.L., Gonzalez, A., Duffy, J.E.,
Gamfeldt, L., O'Connor, M.I., 2012. A global synthesis reveals biodiversity loss as a major driver of ecosystem
change. Nature 486, 105–108.
Johnston, E.L., Roberts, D.A., 2009. Contaminants reduce the richness and evenness of marine communities: a
review and meta-analysis. Environ. Pollut. 157, 1745–1752.
Kennedy, A.D., Jacoby, C.A., 1999. Biological indicators of marine environmental health: meiofauna – a neglected
benthic component? Environ. Monit. Assess. 54, 47–68.
King, R.S., Baker, M.E., 2010. Considerations for analyzing ecological community thresholds in response to
anthropogenic environmental gradients. J. North Am. Benthol. Soc. 29, 998–1008.
Legendre, P., Anderson, M.J., 1999. Distance-based redundancy analysis: testing multispecies responses in
multifactorial ecological experiments. Ecol. Monogr. 69, 1–24.
Levene, H., 1960. Robust tests for equality of variances. In: Olkin, I., Ghurye, SG, Hoeffding, W., Madow, W.G.,
Mann, H.B. (eds.) Contributions to Probability and Statistics. Essays in Honor of Harold Hotelling. Stanford
University Press, Stanford, CA, pp. 278–292.
Logan, B., Taffs, K.H., 2013. Relationship between diatoms and water quality (TN, TP) in sub-tropical east
Australian Estuaries. J. Paleolimnol. 50, 123–137.
Madoni, P., 1994. A sludge biotic index (SBI) for the evaluation of the biological performance of activated sludge
plants based on the microfauna analysis. Water Res. 28, 67–75.
McCormick, P.V., Cairns, J. Jr., 1994. Algae as indicators of environmental change. J. Appl. Phycol. 6, 509–526.
Medinger, R., Nolte, V., Pandey, R.V., Jost, S., Ottenwälder, B., Schlötterer, C., Boenigk, J., 2010. Diversity in a
hidden world: potential and limitation of next generation sequencing for surveys of molecular diversity of
eukaryotic microorganisms. Mol. Ecol. 19, 32–40.
Morgan, M.J., Chariton, A.A., Hartley, D.M., Court, L.N., Hardy, C.M., 2013. Improved inference of taxonomic
richness from environmental DNA. PLoS One 8, e71974.
Morrisey, D.J., Underwood, A.J., Howitt, L., Stark, J.S., 1992. Temporal variation in soft-sediment benthos. J. Exp.
Mar. Biol. Ecol. 164, 233–245.
Newell, S., Fell, J., 1995. Do halophythopthoras (marine Pythiaceae) rapidly occupy fallen leaves by intraleaf
mycelial growth? Can. J. Bot. 73, 761–765.
Newell, S., Fell, J., 1997. Competition among mangrove oomycetes, and between oomycetes and other microbes.
Aquat. Microb. Ecol. 12, 21–28.
Novinscak, A., DeCoste, N.J., Surette, C., Filion, M., 2009. Characterization of bacterial and fungal communities in
composted biosolids over a 2 year period using denaturing gradient gel electrophoresis. Can. J. Microbiol. 55,
Quigg, A., Litherland, S., Phillips, J.A., Kevekordes, K., 2010. Phytoplankton productivity across Moreton Bay,
Queensland, Australia: the impact of water quality, light and nutrients on spatial patterns. Memoirs Qld. Mus.
Quinn, G.P., Keough, M.J., 2002. Experimental Design and Data Analysis for Biologists. Cambridge University
Rabalais, N.N., Turner, R.E., Díaz, R.J., Justihttp://refhub.elsevier.com/S0269-7491(15)00177-3/sref45c, D., 2009. Global
change and eutrophication of coastal waters. ICES J. Mar. Sci. J. du Conseil 66, 1528–1537.
Remane, A., Schlieper, C., 1971. Biology of Brackish Water. E. Schweizerbart, Stuttgart.
Roy, P.S., Williams, R.J., Jones, A.R., Yassini, I., Gibbs, P.J., Coates, B., West, R.J., Scanes, P.R., Hudson, J.P.,
Nichol, S., 2001. Structure and function of south-east Australian Estuaries. Estuarine Coast. Shelf Sci. 53, 351–
Sandgren, C.D., Smol, J.P., Kristiansen, J., 1995. Chrysophyte Algae: Ecology, Phylogeny and Development.
Cambridge University Press, Cambridge.
Snoeijs, P., 2013. Diatoms and environmental change in brackish waters. In: Stoermer, E.F., Smol, J.P. (eds.), The
Diatoms: Applications for the Environmental and Earth Sciences. Cambridge University Press, Cambridge, pp.
Sokal, R.R., Rolf, F.J., 1995. Biometry: the Principles and Practice of Statistics in Biological Research. Third
edition. W.H. Freeman, San Francisco, CA.
Taberlet, P., Coissac, E., Pompanon, F., Brochmann, C., Willerslev, E., 2012a. Towards next-generation
biodiversity assessment using DNA metabarcoding. Mol. Ecol. 21, 2045–2050.
Taberlet, P., Coissac, E., Pompanon, F., Brochmann, C., Willerslev, E., 2012b. Towards next-generation
biodiversity assessment using DNA metabarcoding. Mol. Ecol. 21, 2045–2050.
Thrush, S., Whitlatch, R., Pridmore, R., Hewitt, J., Cummings, V., Wilkinson, M., 1996. Scale-dependent
recolonization: the role of sediment stability in a dynamic sandflat habitat. Ecology 77, 2472–2487.
Valentini, A., Miquel, C., Nawaz, M.A., Bellemain, E., Coissac, E., Pompanon, F., Gielly, L., Cruaud, C., Nascetti,
G., Wincker, P., Swenson, J.E., Taberlet, P., 2009. New perspectives in diet analysis based on DNA barcoding
and parallel pyrosequencing: the trnL approach. Mol. Ecol. Resour. 9, 51–60.
Whitlatch, R.B., Zajac, R.N., 1985. Biotic interactions among estuarine infaunal opportunistic species. Mar. Ecol.
Prog. Ser. 21, 299–311.
Wiens, J.A., 1989. Spatial scaling in ecology. Funct. Ecol. 3, 385–397.
Witman, J.D., Etter, R.J., Smith, F., 2004. The relationship between regional and local species diversity in marine
benthic communities: a global perspective. Proc. Natl. Acad. Sci. U.S.A. 101, 15664–15669.
Ysebaert, T., Herman, P.M., 2002. Spatial and temporal variation in benthic macrofauna and relationships with
environmental variables in an estuarine, intertidal soft-sediment environment. Mar. Ecol. Prog. Ser. 244, 105–
Yu, Y., 2001. Pythium and fungi. Jiangxi Sci. 19, 55–60.
Table 1 Morphological characteristics and environmental condition of the five estuaries
Urban land use, modified entrance.
Urban and rural land use, treated sewage
Sewage treatment plant, dam,
residential, urban use and dredging
Canal estates, rural residential, urban use
Sewage treatment plant, rural residential,
urban use and dredging
Information compiled from www.ozcoasts.gov.au and www.health-e-waterways.org. a Indicates ecological condition derived
under the Ecosystem Health Monitoring Program (EHMP)
Figure 1. Locations of the five estuaries (Noosa, Maroochydore, Pine, Logan and Currumbin) sampled
within south-east Queensland (Australia). Bold letters represent EMPH Report Card scores for 2010 (see
Table 1 for details). Shaded areas indicate the catchments for the estuaries
Figure 2. nMDS plot illustrating the similarities and differences in the compositions of benthic eukaryotic
communities from the five estuaries. The shading of site markers indicates their position from upstream
(light) to downstream (dark)
Figure 3. Summary of the indicator analysis illustrating the relative proportion of MOTUs associated
with each taxonomic group for each estuary. Bracketed values after estuary names
represent the total number of potential indicator MOTUs identified in each estuary
Figure 4. A dbRDA ordination plot illustrating the relationships between benthic community structure
and the measured environmental variables. Sites are derived from their distances among centroids
Figure 5. Summary of Threshold Indicator Taxa ANalysis (TITAN) results illustrating the change points
and 95% confidence limits for the top 20 (highestz and þz scores) significant MOTUs for key
environmental gradients: Total phosphorus (TP); Mono-nitrogen oxides (NOx); turbidity; and pH. Black
and orange points indicate MOTUs with z and þz scores, respectively. The size of the points is scaled to
reflect the magnitude of their response (z scores)