oEPA
United States
Environmental Protection
Agency
EPA/600/R-17/490 | January 2018 | www.epa.gov/research
Developing Microbial
Community Indicators of
Nutrient Exposure in Southeast
Coastal Plain Streams using a
Molecular Approach
Office of Research and Development
National Health and Environmental Effects Research Laboratory, Gulf Ecology Division
-------
EPA/600/R-17/490
January 2018
Developing Microbial Community Indicators of Nutrient
Exposure in Southeast Coastal Plain Streams using a
Molecular Approach
Contributing Authors
James D. Hagy III, Richard Devereux,
Katelyn A. Houghton, David L. Beddick, Jr., Troy A. Pierce,
and Stephanie D. Friedman
Other Contributors
Jessica Aukamp, Tripp Boone, Jerry Boos, Ryan Boylan, Lael Butler, Rebecca Crosby, Fred
Genthner, Joe James, Brandon Jarvis, Nicholas Thiemann, Sherry Wilkinson, and Diane Yates
Gulf Ecology Division
National Health and Environmental Effects Research Laboratory
Gulf Breeze, FL 32561
-------
Notice/Disclaimer Statement
This document has been reviewed by the U.S. Environmental Protection Agency, Office of Research and
Development, and approved for publication.
Any mention of trade names, products, or services does not imply an endorsement by the U.S. Government or the
U.S. Environmental Protection Agency (EPA). The EPA does not endorse any commercial products, services, or
enterprises.
Citation
Hagy, J. D. Ill, R. Devereux, K. A. Houghton, D. L. Beddick, Jr., T. A. Pierce, and S. D. Friedman. 2018. Developing
Microbial Community Indicators of Nutrient Exposure in Southeast Coastal Plain Streams using a Molecular
Approach. US Environmental Protection Agency, Office of Research and Development, National Health and
Environmental Effects Research Laboratory, Research Triangle Park, NC. EPA 600/R-17/490. 44 pp.
-------
Foreword
The U.S. Environmental Protection Agency (EPA) is charged by Congress with protecting the Nation's land, air, and
water resources. Under a mandate of national environmental laws, the Agency strives to formulate and implement
actions leading to a compatible balance between human activities and the ability of natural systems to support
and nurture life. To meet this mandate, EPA's research program is providing data and technical support for solving
environmental problems today and building a science knowledge base necessary to manage our ecological
resources wisely, understand how pollutants affect our health, and prevent or reduce environmental risks in the
future.
The National Health and Environmental Effects Research Laboratory (NHEERL) conducts systems-based, effects
research needed to achieve sustainable health and wellbeing. Research encompasses both human and
ecosystem health, in that they are inextricably linked. NHEERL's research is aligned to several strategic goals,
which are:
• Lead innovative research and predictive modeling efforts that link environmental condition to the health
and wellbeing of people and society.
• Advance research and tools for achieving sustainable and resilient watersheds and water resources.
• Advance systems-based research to predict the adverse effects of chemicals and other stressors across
species and biological levels of organization through the development and quantification of adverse
outcomes pathways across multiple scales.
• Use integrated research to identify and characterize modifiable factors that respond to environmental
conditions, and through intervention, improve health and wellbeing.
• Translate and communicate integrated environmental and health effects science to impact decisions
positively at all levels.
Wayne Cascio, Acting Director
National Health and Environmental Effects Research Laboratory
About this Report
This report describes the results obtained from a research project entitled "Water Quality and Aquatic
Life Responses to Implementation of Best Management Practices in Gulf of Mexico Initiative Focus Watersheds."
The project was partially supported as a Regional Applied Research Effort (RARE) by US EPA Region 4 as well as
the Safe and Sustainable Water Research Program, Project 4.02, Task A, "Improved Nutrient Indicator
Development." The research was planned and conducted as a collaboration between the US EPA Gulf of Mexico
Program and the US EPA National Health and Environmental Effects Research Laboratory, Gulf Ecology Division.
Due to circumstances uncovered as the project unfolded, the research focus was shifted away from evaluating
Agricultural Best Management Practices (BMPs) and instead focused more broadly on developing indicators of
nutrient exposure in streams located within agricultural landscapes in US Department of Agriculture Gulf of
Mexico Initiative watersheds.
4
-------
Table of Contents
List of Figures 6
List of Tables 8
Acknowledgments 10
Executive Summary 11
Introduction 14
Methods 15
Study Sites 15
Water Quality Analysis 16
Continuous Nitrate Sensors 17
Periphyton collectors 17
Surficial sediment 17
Nutrient Diffusing Substrate Experiments 17
Periphyton Bulk Properties 17
DNA Extraction and Sequencing 19
Analyses of DNA Sequences 19
Periphyton Community Structure Analysis 19
Calculation of Nutrient Response Index 19
Results 20
Site Characteristics 20
Periphyton Bulk Properties 26
Periphyton Community Composition 26
Eukaryote Community Composition - 18S rRNA 26
Prokaryote Community Composition - 16S rRNA 28
Nutrient Diffusing Substrate (NDS) Experiments 29
Characterizing Community Responses to Nutrients 31
Nutrient Response Index 32
Periphyton Community Composition 34
Discussion 35
References 40
5
-------
List of Figures
Figure 1. Location of study watersheds (upper panel) and station locations and land use within each study
watershed (lower panels). Land use data is from the 2011 National Land Cover Database (Homer, Dewitz et
al. 2015). Water quality data was collected at all stations, whereas periphyton (+ periphyton) was also
collected at selected stations and continuous nitrate data (+SUNA) was also collected at selected
periphyton stations (Table 1) 16
Figure 2. A periphyton collector just prior to recovery from Wet Weather Creek on August 21, 2015 17
Figure 3. A nutrient diffusing substrate upon recovery following deployment in the Fish River in August 2015.. 19
Figure 4. Specific conductivity vs. percent agricultural land use in the catchment 20
Figure 5. The relationship between mean total nitrogen (upper panel) or total phosphorus (lower panel) and
percent agriculture in the catchment 21
Figure 6. Time series of nitrate concentration in Corn Branch (Weeks Bay Watershed) measured in situ every 30
minutes using an Satlantic SUNA V2 nitrate sensor (black line) with laboratory nitrate plus nitrite
measurements (red points) collected before and after deployments. Bars indicate daily local rainfall totals.
HM = micromoles per liter 22
Figure 7. Time series of nitrate concentration at Fish River Reference (Weeks Bay Watershed) measured in situ
every 30 minutes using a Satlantic SUNA V2 nitrate sensor (black line) with laboratory nitrate plus nitrite
measurements (red points) collected before and after deployments. Bars indicate daily local rainfall totals.
HM = micromoles per liter 23
Figure 8. Relationships among total organic carbon, total nitrogen, and percent agriculture in the catchments,
with shapes indicating the three different watershed study areas 24
Figure 9. Results of a principal components analysis conducted on key water quality variables and watershed
attributes for the 12 periphytometer sites. Variables were in each case transformed by log(x+l). Variable
loadings (upper panel) for principal components 1 and 2 are scaled by the respective eigenvector. PCA
scores (lower panel) illustrate the distribution water quality on the PC axes for each of the three study
watersheds. %Ag=percent agriculture, DA=drainage area, WT=water temperature, SpC=specific
conductivity, P04=P043", NH4=NH4+, TN=total nitrogen, N0x=N02~+N03~ 25
Figure 10. The relationship between the average rate of periphyton accumulation and measured total nitrogen
(upper panel) and total phosphorus (lower panel) concentrations. The linear relationship (upper panel) is
fitted only to the Weeks Bay stations and is not present at the other stations 27
Figure 11. Multi-dimensional scaling (MDS) ordinations illustrating shifts in eukaryote community structure
associated with nutrient enrichment treatments at Corn Branch (upper panel) and Fish River Reference
(lower panel) on up to 6 different dates. Ordinations were computed using Primer (v7) 28
Figure 12. For 18S libraries in each NDS experiment, the estimated number of eukaryote classes (i.e. OTUs)
present as computed using the Chao index (upper panel) and community diversity computed as the inverse
of the Simpson index. Both metrics were computed using mother software (Schloss, Westcott et al. 2009).
C=Control, +P=Phosphorus amendment, +N=Nitrogen amendment, +NP=Nitrogen and phosphorus
amendment 29
6
-------
Figure 13. Multi-dimensional scaling (MDS) ordinations illustrating shifts in prokaryote community structure
(based on 16S sequences) associated with nutrient enrichment treatments at Corn Branch (upper panel)
and Fish River Reference (lower panel) on up to 6 different dates. Ordinations were computed using Primer
(v7) 30
Figure 14. For 16S libraries in each NDS experiment, the estimated number of prokaryote classes (i.e. OTUs)
present as computed using the Chao index (upper panel) and community diversity computed as the inverse
of the Simpson index. Both metrics were computed using mother software (Schloss, Westcott et al. 2009).
31
Figure 15. Index values representing the degree of community change related to nutrient exposure. Higher
values indicate community responses related to the average change associated with +NP treatments as
compared to controls 33
Figure 16. Non-metric multidimensional scaling ordinations of prokaryote (16S, upper panel) and eukaryote
(18S, lower panel) community responses to +NP nutrient amendments in NDS experiments conducted at 2
sites and 5 seasonal periods, including the average (red) of all responses. Because community response
metrics may be negative, ordinations are based on Euclidean distance rather than the Bray-Curtis
dissimilarity used in this study for ordination of community composition 34
Figure 17. Non-metric multi-dimensional scaling ordinations of prokaryote (16S) sequences from periphyton
collected at each of 12 stream locations. Identical colors and symbols within a panel indicate replicate
analysis at each station and date 37
Figure 18. Non-metric multi-dimensional scaling ordinations of eukaryote (18S) sequences from periphyton
collected at each of 12 stream locations. Identical colors and symbols within a panel indicate replicate
analysis at each station and date 38
Figure 19. Nutrient response index values based on 16S sequences (upper panel) and 18S sequences (lower
panel) applied to periphyton samples and related to percent agriculture in the catchment 39
7
-------
List of Tables
Table 1. Names, abbreviations, and locations of sampling stations and sampling activities. In the sample
column: W=water chemistry, P=periphytometer deployment, S=SUNA nitrate sensor, N=nutrient diffusing
substrate experiment, l=IDEXX fecal indicator bacteria abundance. DA= Drainage Area; Land Use:
%Ag=percent agriculture or pasture, %Nat = percent neither developed nor in agricultural use. Land use
percentages do not sum to 100% (e.g., because there is some "developed" land use). Sites are sorted
within watershed by decreasing agricultural land use. Land use data from 2011 National Land Cover
Database (Homer, Dewitz et al. 2015) 18
Table 2. Mean and standard deviation of total nitrogen (TN) total phosphorus (TP) at each of the periphytometer
stations 24
Table 3. The rate of accumulation of ash-free dry weight (AFDW Accum, mg m"2 d"1) and percent organic matter
(%O.M.) of the periphyton accumulated during nominally 2-week deployments of periphyton collectors
(winter deployment was 28 days) 26
8
-------
Acronyms and Abbreviations
16S
A gene for ribosomal RNA of prokaryotes (Bacteria and Archaea)
18S
A gene for ribosomal RNA of eukaryotes
ANL
Argonne National Laboratory
AFDW
Ash-Free Dry Weight (measure of organic matter content)
ANOSIM
Analysis of Similarity (a multivariate statistical test)
BMPs
Best Management Practices
DNA
Deoxyribonucleic acid
EPA
Environmental Protection Agency
MDS
Multi-dimensional scaling (an ordination method)
MPN
Most probable number, a measure of bacterial abundance
NDS
Nutrient-diffusing substrate
NMDS
Non-metric multi-dimensional scaling (an ordination method)
NHEERL
National Health and Environmental Effects Research Laboratory
ORD
Office of Research and Development
OTU
Operational Taxonomic Unit
RNA
Ribonucleic acid
PC
Particulate carbon
PP
Particulate phosphorus
RARE
Regional Applied Research Effort
SIMPER
Similarity percentage (a statistical method)
TDN
Total dissolved nitrogen
TDP
Total dissolved phosphorus
TN
Total nitrogen
TP
Total phosphorus
9
-------
Acknowledgments
The research described in this report was supported by EPA Gulf Ecology Division staff who made important
contributions to the research. Jessica Aukamp analyzed water samples for total dissolved nitrogen, dissolved
organic carbon, particulate carbon, and particulate nitrogen. Joe James participated in early work and
contributed to early project planning. Sherry Wilkinson participated in field sampling in Alabama. Diane Yates
analyzed water samples for total dissolved phosphorus and provided quality control for some of the analysis.
Ryan Boylan (Student Services Contractor) analyzed water samples for ammonium, organized field datasheets,
and participated in field operations. Fred Genthner (Retired) contributed to early development of the nutrient
diffusing substrate apparatus. Nicholas Thiemann (Volunteer) participated in field operations. Periphytometers
and nutrient diffusing substrate apparatus were constructed by machinists working onsite under contract to the
EPA.
Field work and processing of water samples from Rotten Bayou, MS was supported by EPA Gulf of Mexico
Program staff, including Lael Butler, Jerry Boos, and Tripp Boone. Rebecca Crosby, a Gulf of Mexico Program
volunteer undergraduate student from the University of Southern Mississippi also assisted in Rotten Bayou.
Sequencing of DNA extracts was completed by the Argonne National Laboratory Environmental Sample
Preparation and Sequencing Facility (http://ngs.igsb.anl.gov/).
We acknowledge helpful reviews of early drafts by Yongshan Wan, Eric Stein, Nathan Smucker, and Alison
Watts.
10
-------
Executive Summary
Nutrient (nitrogen and phosphorus) pollution is a major cause of water quality degradation in the US, including
within the states that comprise US EPA Region 4. To address the challenges associated with nutrient
management, Office of Research and Development's (ORD) Safe and Sustainable Water Research (SSWR)
Program has pursued research to inform development and implementation of nutrient management by EPA and
its state and local partners. The Regional Applied Research Effort (RARE) program and ORD jointly support
research of interest to sponsoring regions that is aligned with ORD research priorities. This report describes the
results of a research project initiated under RARE.
This report covers a project period from September 1, 2014 to September 30, 2016 and included subsequent
analysis completed as of September 20, 2017. The objectives of this study were to investigate relationships
between land use and water quality in streams, characterize periphyton1 community responses to water quality,
and develop new sensitive and nutrient-specific periphyton indicators of water quality using a molecular
approach based on analysis of DNA sequences. Although the report addresses each of these objectives, the
most important and novel aspect of the research was use of DNA sequences to characterize microbial
community composition of periphyton and develop indicators, providing an alternative to traditional approaches
using microscopy. We evaluated water quality and sampled periphyton and sediments in streams selected
within 3 coastal plain watersheds in Mississippi and Alabama from June 2015 to June 2016. Sampling sites
spanned a gradient in the proportion of agricultural land use in the catchment.
The data showed that, as expected, stream water quality, particularly nitrogen, was broadly correlated with the
percent of agricultural land use in the catchment. Conductivity and total nitrogen were positively correlated
with percent agriculture and plotted together in the context of multivariate analysis. Phosphorus concentration
was not related in the same way, and was instead related to other variables. The rate of periphyton growth on
artificial periphyton collector plates was correlated with both percent agricultural land use and total nitrogen
concentration, but not total phosphorus. Two-week in situ deployments of continuous nitrate sensors at two
sites showed that concentrations varied dramatically on short time scales associated with rain events. Although
sensors were not available to collect similar data for phosphorus, run-off events mobilize both nitrogen and
phosphorus and thereby drive short term variability in both, albeit via different processes. These processes and
associated patterns of variability increase the number of measurements needed to characterize nutrient
exposures in streams. Combined with uncertainty associated with responses to nutrient exposures, these
observations point to the recognized value of biological indicators. However, existing indicators lack nutrient-
specificity, have a high cost associated with taxonomic enumerations, and additional variability associated with
taxonomic analyst bias. An alternative bioindicator based on gene sequencing could address these problems.
We hypothesized that new and rapidly developing technologies for gene sequencing, bioinformatics, and
multivariate data analysis could enable development of new bioindicators. These indicators could - like many
existing indicators - be based on evaluating stream periphyton community composition. However, unlike
previous indicators that were limited to taxa that could be identified by light microscopy, a gene-based approach
1 Periphyton is the mixture of algae, photosynthetic bacteria, heterotrophic microbes, and detritus that is attached to
substrates within the stream, whether it grows directly on the surface or comes from elsewhere and becomes attached.
11
-------
allows for considering nutrient responses of vastly more taxa. These include algae and other eukaryotes2,
Bacteria and Archaea, and phylogenetically distinct operational taxonomic units (OTUs) that are less well-
characterized taxonomically. Due to the rapidly declining costs associated with sequencing, it was possible to
collect, sequence, and analyze the community composition of more than 500 samples of periphyton and
sediments.
One major challenge identified early in the study was that, whereas traditional approaches to pollution
indicators utilized prior knowledge of the tolerance or sensitivity of taxa to pollutants, this information was not
expected to be available for most of the taxa that would be identified via gene sequences. In a DNA-based
study, taxa resolved genetically to the nominal level of "species" are referred to as "operational taxonomic
units" or "OTUs." More than 200,000 prokaryote OTUs and more than 20,000 eukaryote OTUs were identified
in the sequences periphyton assemblages. To address this problem, we embedded 11 nutrient-enrichment
experiments within the field study by deploying nutrient-diffusing substrates (NDS). In these NDS experiments,
stream periphyton could access added nutrients that slowly diffused through the artificial substrate (glass-fiber
filter) on which they were growing. These experiments showed the effect of nutrient amendments (+P, +N, +N
and P) on the periphyton community and on specific OTUs identified in the study.
The NDS experiments were, in our view, very successful. Nutrient amendments resulted in significant changes
in periphyton community composition. Key results were replicated within experiments and across experiments
conducted at different sites and on different dates. Multivariate tests of community composition showed that
nutrient-amended treatments differed from controls. As expected, nutrient amendments decreased species
richness and diversity among both prokaryotes and eukaryotes in comparison to non-amended controls (C). In
most experiments, phosphorus amendments (+P) or nitrogen and phosphorus amendments (+NP) had a larger
effect on composition, diversity, and richness than amendments that did not include phosphorus (+N). Analysis
of the specific OTUs responsible for community shifts associated with nutrient amendments showed that
thousands of microbial OTUs were potentially "nutrient-sensitive" or "nutrient-indicative." Data quantifying
these responses were used to derive a nutrient-response index, which was strongly reflected in the microbial
communities observed in most of the NDS experiments. Analysis of key taxonomic groups contributing to the
nutrient index showed that diatoms, chrysophytes (golden-brown algae), and ciliates (single cell heterotrophs)
decreased in response to nutrient amendments, whereas abundant groups that became more abundant with
nutrient amendments included chlorophytes (green algae), euglenids, and notably the amoeba genus Naegleria,
of which some species and their variants are pathogenic (De Jonckheere 2014).
Periphyton communities varied distinctly among stations and watersheds, whereas replicate samples had similar
communities. These results suggest that characteristic communities may have developed in response to
ambient water quality, although the nutrient response index developed using the NDS experiments resolved
these differences only to a limited extent. Whereas the nutrient response index values for the prokaryotes (16S
sequences) were associated to some extent with the fraction of agriculture in the watershed and associated
nutrient concentrations, no relationship at all was observed for eukaryotes.
The data collected in this study are extensive and present many analytical possibilities that have not yet been
pursued. Our initial evaluation of the literature related to molecular analysis of microbial communities and
2 Eukarya, Bacteria and Archaea comprise the three domains of life. Eukarya or Eukaryotes are characterized by having a
membrane-bound cell structures and include many single- and all multicellular organisms. Bacteria and Archaea,
comprised entirely of single celled non-nucleated microorganisms, are nonetheless extremely diverse and ecologically
important, being responsible along with the Eukarya for mediating the biogeochemical processes that comprise carbon,
nitrogen and phosphorus cycling in ecosystems.
12
-------
water quality indicators suggests that the former is evolving very rapidly, whereas the latter is more stable but
very well developed. Both fields therefore offer substantial opportunity to further develop this work. Further
research and development is indicated and could reasonably result in a valuable new approach to assessing
water quality and stream condition. The results may be of interest to environmental agencies and stakeholders
responsible for stream water quality in the region, including the Mississippi Department of Environmental
Protection, Alabama Department of Environmental Management, and the Poarch Band of Creek Indians.
Agencies with a broader scale of responsibility such as the US Department of Agriculture and US EPA may also
be interested in the potential for further development and broader application in other watersheds.
13
-------
Introduction
Nutrient (nitrogen and phosphorus) pollution is a major cause of water quality degradation globally - including
the US - and consequently is a significant focus of management actions to reduce pollution, improve biotic
condition and protect against anticipated future degradation (Vitousek, Aber et al. 1997, Millenium Ecosystem
Assessment 2005, Davidson, David et al. 2012). Nutrients entering waterways can come from a variety of
sources. In the case of small streams, sources are often dominated by diffuse inputs associated with shallow
groundwater transport and local surface run-off. In less-developed areas of the southeastern coastal plain of
the US, agricultural land uses including row crops and pasture are significant sources of nutrients (Hoos and
McMahon 2009, Garcia, Hoos et al. 2011), with urban stormwater and waste-water runoff being important
sources of nutrients where developed land uses are more extensive.
Effective management of nutrient pollution requires adequate measures of nutrient exposure and effects and
scientifically defensible methods for developing water quality and biotic thresholds. For management of rivers
and streams, ambient nutrient concentrations are a straightforward indicator of exposure, but can also be highly
variable. For example, nutrient concentrations often depend strongly on streamflow, and these relationships
may change over time as nutrient sources change (Hirsch, Moyer et al. 2010). Biotic indicators potentially offer
the advantage of integrating responses to a temporal average nutrient exposure, and have been developed and
applied widely for assessment of stream condition (Goodnight 1973, Stevenson 2014). The idea of stream
bioindicators is mature (e.g., Goodnight 1973, Karr 1981) and its development has continued for many years
(Barbour, Gerritsen et al. 1996, Stevenson 2014). Periphyton is the mixture of algae, photosynthetic bacteria,
heterotrophic microbes, and detritus that is attached to substrates within a stream, whether it grows directly on
the surface or comes from elsewhere and becomes attached. Periphyton as indicators of nutrient effects in
freshwater systems has some precedent particular in low nutrient systems such as the Florida everglades
(McCormick and Stevenson 1998). Nonetheless, with respect to nutrient effects, there is a need for indicators
that respond specifically to nutrients that can be used to support management of nutrient pollution
(Environmental Protection Agency 2014).
One challenge is that indicators of biotic condition in streams do not necessarily relate to nutrients specifically,
and are therefore less useful for identifying protective nutrient concentration thresholds for management.
Nutrient indicators have been based on the assemblage of diatoms, a class of algae characterized by their
distinctive siliceous cell walls, known as frustules. Although indicators based on stream diatom community
composition are related to nutrient concentrations, some relationships may be confounded with pH or other
aspects of water quality, which along with nutrients also vary with the degree of anthropogenic influence
(Stevenson, Pan et al. 2008, Stevenson and Decker 2012). In the case of a Florida stream diatom index, an
overwhelming association between human disturbance and stream pH limited application of the index to
address phosphorus pollution (Fore 2010). Florida's numeric nutrient criteria that currently apply to natural
streams throughout the state are based on a reference stream approach. Stream condition is also assessed via
bioassessments including the Stream Condition Index and BioRecon (Barbour, Gerritsen et al. 1996, Fore 2004).
One way to potentially address some of the shortcomings of existing indicators based on diatoms or
macroinvertebrates is to evaluate microbial communities using a molecular phylogenetic approach. The key
hypothesis is that molecular methods will increase taxonomic resolution, providing more information from
which to develop nutrient-specific response metrics. The concept that microbial communities could be a
sensitive and specific indicator of aquatic ecosystem responses to nutrient enrichment, particularly when
characterized in great detail using DNA sequencing, is suggested by results obtained from soil microbiomes (Leff,
14
-------
Jones et al. 2015) and observed relationships between microbial community structure and nutrient processing in
aquatic ecosystems (Lisa, Song et al. 2015). Paerl et al. (2003) suggested that molecular tools could be applied
to eutrophication indicators, despite the fact that many of the tools available today were new in 2003, yet
undeveloped, or expensive and difficult to implement. Recent studies of microbial communities in aquatic
ecosystems have illustrated community structure responses related to major environmental drivers, such as
water temperature and salinity or conductivity (Ortmann and Santos 2016). Although some community
responses to nutrients have been observed using molecular techniques (Ortmann and Santos 2016),
documentation of nutrient responses is uncommon or reported effects are unclear, perhaps because such
responses are harder to isolate or characterize from other drivers.
The objective of this study was to evaluate periphyton responses to nutrient exposure in southeastern coastal
plain streams using a molecular approach leading to development of an indicator based on changes in microbial
community structure. We did not evaluate abundance or expression of nutrient-relevant functional genes,
although we would agree that this could be a useful approach. In characterizing community responses to
nutrients, we hope to provide information that will lead to improved assessments of stream condition with
respect to nutrient effects, and to new approaches for stream condition bioassessment and associated
management of nutrient pollution in southeastern watersheds.
Methods
Study Sites Water quality variables and periphyton microbial community composition were characterized in
three southeastern coastal plain watersheds from June 2015 to September 2016 (Fig. 1). Two of the
watersheds, Big Escambia Creek and Weeks Bay (Fish River, Magnolia River) are in coastal Alabama, whereas the
third, Rotten Bayou, is in coastal Mississippi. The sampled streams were generally wadable streams. All sites
had partial tree canopy cover, such that sunlight was filtered by tree cover at least part of the day. Catchment
size for sites with periphyton collectors was 6 to 27 km2 in Weeks Bay and 10 to 30 km2 in Rotten Bayou (Table
1). Catchment sizes for periphyton stations in Big Escambia Creek was substantially larger, ranging from 29 to
839 km2. The process of study site selection was initiated via consultation with stakeholders, including the
Mississippi Department of Environmental Quality, Alabama Department of Environmental Management, the
Poarch Band of Creek Indians, and the US Department of Agriculture. Study sites for periphyton collection were
selected to sample watersheds with both relatively large and small fractions of high nutrient intensity land uses
(e.g., agriculture) within each of the watersheds, while also considering local information provided by the
stakeholders. Land use was evaluated based on the 2011 National Land Cover Database (Homer, Dewitz et al.
2015). Land use was strongly dominated by agriculture at 3 of the Weeks Bay sites, with one site having only
11% agriculture identified as a "reference." Agriculture accounted for 71 to 87% of land use in the other 3
catchments. Within the Big Escambia watershed, one site had only 9% agriculture, whereas the other
periphyton sites had 23 to 41% agriculture. One periphyton site in the Rotton Bayou watershed had 34%
agriculture in the catchment, whereas the other sites had 10 to 15% agriculture (Table 1). The extent of
developed land uses was minimal in all the watersheds.
Water quality was evaluated nominally monthly at 30 water quality stations, with 9 to 13 stations per watershed
(Table 1). Periphyton was collected on artificial substrates deployed just below the surface, and from surficial
sediment for a subset of the study sites and dates. Periphyton collectors (see below) were deployed quarterly
for 2 weeks at 12 stations, with 4 stations in each watershed (Table 1). A nutrient-enrichment experiment using
nutrient-diffusing substrates (NDS) was embedded in the overall study design. NDS experiments were
conducted at 2 of the 4 periphytometer deployment sites in the Weeks Bay watershed (Fish River Reference and
15
-------
Corn Branch; Table 1).
Water Quality Analysis Water samples were collected from the main flow-way of the stream using a
Niskin sampler and transferred to opaque HDPE bottles for transportation in an ambient temperature cooler to
the lab for processing within 4-6 hours. Water samples were processed for particulate carbon (PC) and nitrogen
(PN), dissolved organic carbon (DOC), total dissolved nitrogen (TDN), total dissolved phosphorus (TDP),
particulate phosphorus (PP), and dissolved inorganic nutrients (NH4+, NOx, P043 ). For PC, PP, and PN, sample
water was filtered onto a pre-combusted (450°C, 3 h) GF/F filter until nearly clogged. Filters were analyzed for
PC and PN on a CE Elantech Flash EA elemental analyzer. PP filters were analyzed using the ash-hydrolysis
method (Solorzano and Sharp 1980). For dissolved nutrients, GF/F filtrate was dispensed into vials and frozen at
-70°C until analysis. DOC and TDN were analyzed on a Shimadzu TOC-VCSN carbon analyzer with nitrogen
module, utilizing 720 °C combustion catalytic oxidation coupled with nondispersive infrared carbon and
chemiluminescent nitrogen detectors. Total dissolved phosphorus was analyzed as P043" on an Astoria-Pacific
segmented-flow auto-analyzer after persulfate oxidation. NH4+ was analyzed fluorometrically after Holmes et al.
(1999). NOx and P043" were analyzed on an Aquakem 200 discrete analyzer. Samples below limits of
quantitation were re-analyzed using an Astoria-Pacific segmented flow auto-analyzer. The NOx and P043"
analyses with both instruments used standard colorimetric methods (APHA 1989). Reduction of N03" to N02"
prior to analysis was accomplished using an enzymatic reduction method (Patton, Fischer et al. 2002) on the
Aquakem 200, and cadmium reduction on the Astoria instrument. Additional water samples were collected into
sterile sample bottles, stored in the dark on ice for transport and analyzed within 6 hours for Escherichia coll
abundance using the IDEXX Colilert 2000 assay (IDEXX, Westbrook, ME).
Gulf of
Mexico
.Watershed
Mobile
Bay
Rotten Bayou
Watershed
| Agriculture
| Forest
| Water/Wetlands
| Developed
Grassland/Scrub
Atlantic
Ocean
Weeks Bay
0 5.5 11 18.5 km
Big Escambia
Creek
Waler Quality
P + Periphyton
Figure 1. Location of study watersheds (upper panel) and station locations and land use within each study
watershed (lower panels). Land use data is from the 2011 National Land Cover Database (Homer, Dewitz
et al. 2015). Water quality data was collected at all stations, whereas periphyton (+ periphyton) was also
collected at selected stations and continuous nitrate data (+SUNA) was also collected at selected
periphyton stations (Table 1).
16
-------
Continuous Nitrate Sensors A Satlantic (Sea-Bird Scientific, Halifax, NS, Canada) SUNA V2 ultraviolet
sensor that measured nitrate every 30 minutes over two weeks was used to characterize variability of water
quality during two week deployments at two selected stations; Fish River Reference and Corn Branch. A YSI
(Yellow Springs, OH) water quality sonde that measured water temperature and conductivity was deployed
alongside the nitrate sensor instrument.
Periphyton collectors Periphyton was collected on
transluscent textured 95 x 125 mm acrylic plates oriented
vertically just below the water surface in a floating PVC frame (Fig
2). Collectors were deployed for 14-28 days. Each side of 6
periphytometer plates at each station were aseptically scraped
into sterile 50 mL Falcon tubes, resulting in 12 samples per site.
Samples were transported on ice and stored at -80°C until further
processing.
Surficial sediment Sediment was collected from sites in the Big Escambia Creek watershed and Weeks Bay
watershed during retrieval of quarterly periphyton deployments. Sediment was collected beginning in August
2015. Triplicate sediment cores were obtained from a single haphazardly selected site accessible from the
stream bank. Sediment cores were collected from below the water surface using a 25-mm acrylic core tube and
extruded to obtain the top 2 cm of sediment. Sediment samples were collected into new 50 mL Falcon tubes
and transported on ice before freezing at -70 °C.
Nutrient Diffusing Substrate Experiments Nutrient Diffusion Substrates (NDS) were deployed in two
streams, Corn Branch and Fish River, on six different dates and used to test the response of periphyton
communities to addition of nutrients (Fig. 1; Table 1). The experimental design included triplicates of 4
treatments: control (C), phosphorus enrichment (+P), nitrogen addition (+N), and enrichment with both nitrogen
and phosphorus (+NP). NDS were prepared by mixing Nobel Agar (a purified form of agar) to a final
concentration of 2% with equal volumes of 0.5 M solutions of potassium nitrate (+N), sodium phosphate (+P), or
both (+NP), to give a final nutrient concentration of 0.25 M. The control substrate was agar and MilliQ water.
Nutrient solutions and agar were autoclaved separately and cooled to 45-50 °C in a water bath to avoid
formation of potentially toxic substrates (Tanaka, Kawasaki et al. 2014). The still molten solutions were poured
into 130 mL urine specimen containers (Corning 1730-10) and allowed to harden, leaving 10 mL unfilled to
accommodate a 2% agar cap on top of the hardened nutrient substrates. The agar cap prevented excessive
nutrient diffusion at the beginning of the deployment. A sterile 47 mm glass fiber filter was placed onto the agar
surface and held in place by the cap, from which a 38-mm diameter opening was cut. The assembled NDS were
stored in a sterile bag overnight at 4 °C and deployed the following day. The NDS were deployed just below the
water surface for 14 days, exposing the filter to the creek water and providing a surface for periphyton growth
as nutrients diffused through the filter (Fig. 3). Upon recovery, the filters were carefully removed from the agar
surface and placed into sterile petri dishes which were then wrapped in aluminum foil, placed inside a zip lock
bag, and transported on ice before freezing at -70 °C.
Periphyton Bulk Properties Material scraped from periphyton plates were analyzed for periphyton
chlorophyll-a and periphyton biomass (ash-free dry weight). For chlorophyll-a, both sides of one plate were
scraped into a 50 mL Falcon tube for each of 3 replicate samples. The same approach was used to collect
samples for ash free dry weight (AFDW), except that samples were collected in 20 mL pre-combusted glass
scintillation vials. Samples were transported on ice in the dark and stored at -70°C until analysis. Chlorophyll-a
samples were extracted in buffered methanol and analyzed fluorometrically (Welschmeyer 1994). Periphyton
Figure 2. A periphyton collector just prior to
recovery from Wet Weather Creek on August 21,
2015.
17
-------
biomass as AFDW was quantified as the difference in dry mass before and after combustion (APHA 2005, Hauer
and Lamberti 2011). The rate of periphyton accumulation was computed by normalizing accumulate mass to
the deployment duration.
Table 1. Names, abbreviations, and locations of sampling stations and sampling activities. In the
sample column: W=water chemistry, P=periphytometer deployment, S=SUNA nitrate sensor,
N=nutrient diffusing substrate experiment, I=IDEXX fecal indicator bacteria abundance. DA=
Drainage Area; Land Use: %Ag=percent agriculture or pasture, %Nat = percent neither developed nor
in agricultural use. Land use percentages do not sum to 100% (e.g., because there is some
"developed" land use). Sites are sorted within watershed by decreasing agricultural land use. Land
use data from 2011 National Land Cover Database (Homer, Dewitz et al. 2015).
Site Name
Lat (°N)
Lon (°W)
DA
(km2)
Sample
%Ag
%Nat
Weeks Bay Watershed
Magnolia Tributary (MAGT)
30.43605
87.73289
11.54
WP
87%
12%
Baker Branch (BKBR)
30.47548
87.75055
9.95
WPI
80%
19%
Corn Branch (COBR)
30.61854
87.78465
5.73
WPNSI
71%
25%
Turkey Branch (TRKB)
30.42187
87.84390
15.4
W
70%
27%
Magnolia River (MAGR)
30.43602
87.69860
11.08
W
69%
24%
Waterhole Branch (WATB)
30.44548
87.85227
12.42
W
61%
31%
Green Branch (GRBR)
30.44954
87.83552
8.37
W
61%
33%
Pensacola Branch (PENB)
30.52370
87.81231
12.52
W
57%
41%
Polecat Creek (PLCC)
30.49834
87.75087
37.69
w
54%
40%
Perone Branch (PERB)
30.54552
87.78839
22.31
w
47%
48%
Cowpen Creek Weeks (CCWB)
30.48300
87.81905
30.81
w
31%
51%
Fish River Middle (FRMD)
30.54549
87.79821
130.6
w
28%
66%
Fish River Reference (FRRF)
30.65321
87.79210
26.67
WPNSI
11%
87%
Big Escambia Creek Watershed
Sizemore Creek (SZMC)
31.08019
87.48171
29.07
W
55%
38%
Big Escambia Creek Upper West (BEUW)
31.31144
87.37438
159.83
WP
41%
57%
Wet Weather Creek (WTWC)
31.12337
87.46646
76.36
WPI
36%
63%
Cowpen Creek BEC (CCBE)
31.05924
87.34316
32.92
w
33%
66%
Big Escambia Creek Lower (BECL)
31.01042
87.26341
838.79
WPI
23%
75%
Escambia Creek Northeast (ECNE)
31.30414
87.33372
122.28
WPI
9%
90%
Rotten Bayou Watershed
Boardpile Branch (BRPB)
30.51688
89.56593
21.56
WP
34%
65%
Davis Branch (DAVB)
30.51681
89.53621
83.14
w
23%
77%
Mill Creek (MLCR)
30.40945
89.34435
13.5
w
16%
80%
Bayou Lasalle (BYLS)
30.45924
89.34584
9.76
WP
15%
83%
Bacon Branch (BABR)
30.46873
89.49010
38.13
w
15%
84%
Orphan Creek (ORCR)
30.45493
89.47452
29.68
WP
15%
83%
Bayou La Terre (BYLT)
30.44713
89.40477
39.67
w
15%
84%
Kiln Delisle (KLND)
30.42134
89.33987
12.47
WP
10%
89%
18
-------
DNA ExtrdCtion and Sequencing DNA was extracted from periphyton and sediment samples using a
KingFisher Duo (Thermo Scientific™, Waltham, Massachusetts) with the PowerMag Soil DNA Isolation Kit (MoBio
Laboratories, Carlsbad, CA). The same extraction method was used for filters obtained from NDS experiments.
Partial 16S ribosomal RNAgene sequences (251x251 paired end) were obtained on an lllumina (San Diego, CA)
MiSeq sequencer by the Argonne National Laboratory (ANL) Environmental Sample Preparation and Sequencing
Facility (http://ngs.igsb.anl.gov/) using the forward primer 515fB and the reverse primer 806rB targeting the V4
region of the 16S SSU rRNA in Bacteria and Archaea (Caporaso, Lauber et al. 2012, Apprill, McNally et a! 2015).
Partial 18S ribosomal sequences (151x151 paired end) targeting Eukaryotes were obtained in the same way
using forward primer 1391f and reverse primer 1510R as modified by ANL (Amaral-Zettler, McCliment et al.
2009). The sequencing procedure for periphyton was benchmarked by sequencing DNA from a mock community
composed of 20 known bacterial strains (Microbial Mock Community B, HM-783D. BEI Resources, National
Institute of Allergy and Infectious Disease, National Institutes of Health, as part of the Human Microbiome
Project). An identical, but higher concentration, mock community (Mock HM-277-D, lot 60257284) was used for
sediments.
Analyses of DNA Sequences dna sequences were
processed using the mothur software version 1.39.5 (Schloss,
Westcott et al. 2009) following the mothur MiSeq standard
operating procedure. The SILVA non-redundant sequence data
base release 128 was used as the template for aligning and
identifying both 16S and 18S rRNA gene sequences. Sequences
were clustered into Operational Taxonomic Units (OTUs) based on
97% nucleotide sequence identity. 16S contigs assembled from
forward and reverse sequences had an average length of 254 base
pairs. Libraries were sub-sampled to 10,000 sequences, with
libraries having fewer than 10,000 sequences excluded from
further analysis. The 18S contigs assembled from paired end
sequences were around 125 nucleotides in length. The number of
sequences used for each 18S sample was rarified to 2,495.
Identifications based on the 18S rRNA partial gene sequences was
limited by the length of the sequences and incomplete - but rapidly expanding - coverage of eukaryote taxa in
the SILVA database (Quast, Pruesse et al. 2013). Relative abundances of sequences were assumed to indicate
relative abundance of the associated cells, or more precisely differences among samples in relative abundance
for each OTU. While we acknowledge the potential for primer bias, variations in gene copy number, and other
factors to introduce differences between the measured and true community composition, our focus on
differences in microbial communities that were analyzed the same way makes this assumption more defensible
for our study (e.g., Klindworth, Pruesse et al. 2013, Tremblay, Singh et al. 2015).
Periphyton Community Structure Analysis Resulting 16S and 18S abundances (number of sequences
in an OTU divided by the rarified number of sequences in the sample) for OTUs were analyzed separately for
NDS experiments, natural periphyton, and sediments. Visualization and statistical tests for changes in microbial
communities was accomplished using ordination and other statistical methods commonly used to analyze
species abundance data (e.g., NMDS, ANOSIM, PERMANOVA, SIMPER) as implemented in PRIMER, version 7
(Quest Research Limited, Auckland, New Zealand) and the "vegan" Community Ecology package in R (Oksanen,
Blanchet et al. 2016).
Calculation Of Nutrient Response Index A nutrient response indicator was developed by contrasting
OTU abundances in controls vs. +NP treatments in each NDS experiment as follows. Let C be an / x j matrix of
19
Figure 3. A nutrient diffusing substrate upon
recovery following deployment in the Fish River in
August 2015.
-------
coefficients quantifying the contributions of i OTUs to dissimilarity (computed using SIMPER) between control
communities and +NP communities in each NDS experiment j. The matrix D is the same size as C and quantifies
the direction of change (-1 = decreasing, 0=unchanged, +l=increasing) in treatment-average abundances for
each OTU in each experiment. The column vector of nutrient-response metrics, R, was calculated as the row-
means of the elementwise product of C and D and the expected direction and magnitude of nutrient response of
each OTU based on the aggregated results of all the nutrient experiments. Alternative estimates for R were also
obtained by averaging responses from experiments at individual sites or seasons. Nutrient response index
values for individual libraries were calculated as RTN where N is a column vector of OTU abundances in a library.
Index values for each of many libraries were similarly calculated as RTNij, where Ny is instead the matrix of OTU
abundances for all j libraries in the study. The analysis was implemented identically for 16S and 18S data.
Site Characteristics Agricultural land use in the Weeks Bay watershed ranged from 11 to 87% and was the
highest among the three watershed areas (Table 1). Three sites designated as periphyton sites each had >70%
agriculture, whereas a contrasting site (Fish River Reference) had only 11% agriculture. The other two
watersheds had a narrower range of agricultural land use. Agricultural land use ranged from 9 to 55% in the Big
Escambia Watershed and 10-34% in Rotten Bayou Watershed with periphyton collectors spanning a range of
land use in each (Table 1). Across all three watersheds, average specific conductivity was significantly correlated
with percent agriculture (r2=0.72, p<0.01). Relationships within individual watersheds were present for 2 of the
3 watersheds, the exception being Rotten Bayou (p>0.05; Fig. 4).
Results
120 -
Watershed ® Big Escambia 0 Rotten Bayou • Weeks Bay
E. coli abundances (MPN, most probable
number) varied from 5 to 100 MPN/100
ml, were highly variable within sites and
thus not significantly different among sites.
Median abundance varied from 71
MPN/100 ml at Big Escambia Creek Lower
to 361 MPN/100 ml at Corn Branch.
Median E. coli abundance was not
correlated with land-use.
Figure 4. Specific conductivity vs. percent agricultural land use in the
catchment.
0-
25
50
Percent Agriculture
75
Average (±s.d.) total nitrogen (TN)
concentration at the 12 periphyton
stations varied from 0.44±0.28 mg L1 at
Fish River Reference (Weeks Bay) to
3.74±1.0 mg L1 at Magnolia Tributary, but
was generally in the range of 0.5 to 1 mg L
(Table 2). No seasonal pattern was
evident in the TN concentration data.
Mean TN was related to agricultural land
use (N=12, r2=56, p<0.01), increasing
0.24±0.07 mg L1 for every 10% increase in
agriculture in the watershed (Fig. 5).
Ammonium generally accounted for only 1
20
-------
to 3% of TN. Ammonium concentrations were nearly two-fold higher in the Weeks Bay (p<0.01) watershed
compared to the other two watersheds. This reflects much higher concentrations at Magnolia Tributary and
Corn Branch, where the median concentrations were 41 and 70 ng L1, respectively, and higher concentrations at
Baker Branch as well. Concentrations were lower in Rotten Bayou (average=ll ng L"1) and Big Escambia Creek
(average=14 ng L"1).
Nitrate plus nitrite (NOx) accounted for a
variable fraction of TN when TN
concentrations were low, but accounted for a
larger fraction of higher TN concentrations. At
TN > 1 mg L_1, NOx accounted for about 50-
70% of TN. The relationship between percent
agriculture and NOx was similar to the
relationship with total nitrogen.
Continuous nitrate measurements
demonstrated that N03 increased rapidly in
association with rain events (Fig. 6, Fig. 7) and
similarly declined rapidly following the events.
With few exceptions, NOx measured via grab
samples underestimated average exposure to
NOx by sampling baseflow conditions instead
of storm flows and thereby not characterizing
event mean concentrations.
Average total phosphorus (TP) at periphyton
stations varied from 19 ng L"1 at Escambia
Creek Northeast to 121 ng L"1 at Corn Branch.
TP was 2-fold higher (p<0.01) in the Weeks
Bay watershed compared to the other
watersheds. Mean TP was not significantly
correlated with percent agriculture (Fig. 5).
Total organic carbon, which was dominated by
the dissolved fraction differed significantly
among all three watersheds, with TOC highest
at Rotten Bayou and the lowest in the Weeks
Bay Watershed (p<0.01). Where % agriculture
was relatively low in the catchment, there was
a modest positive relationship between TN
and TOC, which was associated with increasing
organic N. On the other hand, the highest TN
was associated with high % agriculture and
lower TOC (Figure 8), with high NOx
accounting for a high fraction of TN.
a-
£
Watershed
Big Escambia 9 Rotten Bayou 9 Weeks Bay
£
e
-------
Com Branch
Summer 2015
40-
-40
-30
- 10
-0
08/07 08/09 08/11 08/13 08/15 08/17 08/19 08/21
40"
20-
1 ZVL J
10/21 10/23 10/25 10/27 10/29 10/31 11/01 11/03
- 125
- 100
-75
¦50
"25 g.
-0
Winter 2016
100-
50-
-60 §
-0
01/15 01/19 01/23 01/27 01/31 02/04 02/08 02/12
Summer 2016
200-
150 "
100-
50-
o-
- 10
-0
0629
07/01
07/03 07/05 07/07 07/09 07/11 07/13
Figure 6. Time series of nitrate concentration in Corn Branch (Weeks Bay Watershed) measured in
situ every 30 minutes using an Satlantic SUN A V2 nitrate sensor (black line) with laboratory nitrate
plus nitrite measurements (red points) collected before and after deployments. Bars indicate daily
local rainfall totals, fj.M = micromoles per liter
22
-------
Fish River Reference
Summer 2015
15 -
10-
0-
-30
-20
- 10
-0
08/06 08/08 08/10 08/12 08/14 08/16 08/18 08/20
Fall 2015
75"
25"
0-
-75
-25
-0
10/21 10/23 10/25 10/27 10/29 10/31 11/01 11/03
£ Winter 2016
20-
-40
I
\ -10
01/15 01/19 01/23 01/27 01/31 02/04 02/08 02/12
Summer 2016
5-
0 -
- 10
-o
06/29 07/01
07/03 07/05
07/07
07 09
07/11
07 13
Figure 7. Time series of nitrate concentration at Fish River Reference (Weeks Bay Watershed)
measured in situ every 30 minutes using a Satlantic SUN A V2 nitrate sensor (black line) with
iaboratory nitrate plus nitrite measurements (red points) collected before and after deployments.
Bars indicate daily local rainfall totals. iaIVI = micromoles per liter
23
-------
0-
5_l
z
CD
c
0
CD
O
Z
ro
o
1-
Figure 8. Relationships among total organic carbon, total nitrogen, and percent
agriculture in the catchments, with shapes indicating the three different watershed
study areas.
Table 2. Mean and standard deviation of total nitrogen (TN) total phosphorus (TP) at each of the
eriphytometer stations.
Total Nitrogen (mg L"1) Total Phosphorus (ng L")
Watershed
Station
Mean
SD
N
Mean
SD
N
Big Escambia
Big Escambia Creek Lower
0.84
0.25
9
36.5
12.52
14
Big Escambia
Big Escambia Creek Upper
1.01
0.49
11
24.57
10.47
12
Big Escambia
Escambia Creek Northeast
0.59
0.18
10
19.12
10.98
12
Big Escambia
Wet Weather Creek
1.38
0.47
11
60.31
79.55
12
Rotten Bayou
Bayou Lasalle
0.68
0.32
9
40.05
23.37
12
Rotten Bayou
Boardpile Branch
1.09
0.74
11
28.47
23.33
12
Rotten Bayou
Kiln Delisle
0.48
0.13
11
34.59
20.01
13
Rotten Bayou
Orphan Creek
0.54
0.17
11
44.41
14.78
13
Weeks Bay
Baker Branch
1.42
0.49
10
73.29
56.52
12
Weeks Bay
Corn Branch
0.81
0.50
11
121.5
88.88
12
Weeks Bay
Fish River Reference
0.44
0.28
10
114.26
178.6
14
Weeks Bay
Magnolia Tributary
3.74
1.00
11
73.64
70.88
13
#
< "m
¦
¦
\ • .
V""-
1 T J ILu _J
A
k
1
•
#
a a
•
a
0
&
! + A A
Watershed
~ Big Escambia
Rotten Bayou
I Weeks Bay
PercentAg
10.8
:
0.6
0.4
0.2
Total Organic Carbon (mgC L [J
24
-------
A principle components analysis
illustrated multivariate relationships
among the water quality variables and
land use drivers (e.g., Wan, Qian et al.
2014), providing a more synthetic
perspective on water quality in the study
watersheds (Fig. 9) . The PCA showed
that % agriculture in the watershed
(%Ag), specific conductivity (SpC), total
nitrogen (TN), and nitrate plus nitrite
(NOx) were all associated with high
values of the first principle component
(PCI), which explained 40% of the
variation in the included variables (Fig.
9, upper panel). Total organic carbon
was associated with negative PCI.
These PC loadings provide a perspective
on the multivariate relationships like Fig.
8. Phosphorus and water temperature
(also indicative of season) were
associated with negative values for PC2.
Drainage area (DA) was associated with
positive values for PC2, which could
represent some biogeochemical
dynamic or, alternatively, simply that Big
Escambia Creek, which had several
particularly large catchments (Table 1)
also had relatively low phosphate.
The three sites appeared to plot
somewhat differently in the principal
components space (Fig. 9, lower panel),
wherein Weeks Bay plotted to the right,
except for the low agriculture Fish River
Reference site. Rotten Bayou and Big
Escambia plotted to the left, with Rotten
Bayou having lower values on PC2 than
Big Escambia. Phosphorus was
associated to some degree with both
PCI and PC2. Accordingly, the Weeks
Bay sites, which tended to have higher P
in addition to higher N, plotted to the
lower right quadrant.
PCA Loadings
*
C\l
O
CL
O
o
O
CL
<
Q
C\l
to
<
O
CL
>
'
DA
X
O
h
)
Ul I
NH4
%Ag
TOC
WT
P04
1
2
0 1
PC1 * 3.93
PCA Scores
„v
-¦2
-2
0 2 4
PCA Axis 1 (+%Ag,+SpC,+TN
Watershed Big Escambia W Rotten Bayou Weeks Bay
Figure 9. Results of a principal components analysis conducted on key water
quality variables and watershed attributes for the 12 periphytometer sites.
Variables were in each case transformed by log(x+l). Variable loadings
(upper panel) for principal components 1 and 2 are scaled by the respective
eigenvector. PCA scores (lower panel) illustrate the distribution water
quality on the PC axes for each of the three study watersheds. %Ag=percent
agriculture, DA=drainage area, WT=water temperature, SpC=specific
conductivity, P04=P0a3~, NH4=NH4+, TN=total nitrogen, N0x=N02+N03.
25
-------
Table 3. The rate of accumulation of ash-free dry weight (AFDW Accum, mg m"2 d"1) and percent
organic matter (%O.M.) of the periphyton accumulated during nominally 2-week deployments of
periphyton collectors (winter deployment was 28 days). Means are for all sample dates.
Watershed
Station
AFDW Accum.
Mean S.D.
% O.M.
Mean S.D.
N
Big Escambia
Big Escambia Creek Lower
27
22
24
6
3
Big Escambia
Big Escambia Creek Upper West
51
47
16
10
4
Big Escambia
Escambia Creek Northeast
37
39
6
3
2
Big Escambia
Wet Weather Creek
130
83
12
5
4
Rotten Bayou
Bayou Lasalle
42
34
14
8
3
Rotten Bayou
Boardpile Branch
22
24
4
4
2
Rotten Bayou
Kiln Del isle
16
5.6
31
15
2
Rotten Bayou
Orphan Creek
82
36
14
3
3
Weeks Bay
Baker Branch
87
21
46
41
4
Weeks Bay
Corn Branch
58
28
20
4
4
Weeks Bay
Fish River Reference
36
36
38
14
4
Weeks Bay
Magnolia Tributary
140
28
29
17
4
Periphyton Bulk Properties Periphyton plates accumulated from 37 to 3,170 mg m"2 of ash-free dry
weight (AFDW; median 843 mg m"2), which for most samples was 11 to 24% (interquartile range) of the
periphyton dry mass (median 20%). Average periphyton biomass varied from (meanis.d.) 230±78 at Kiln Del isle
(Rotten Bayou) to 2,400±580 mg m"2 at Magnolia Tributary (Weeks Bay) with associated periphyton accumulate
rates of 16 to 140 mg m"2 d 1 (Table 3). The chlorophyll-a content of the periphyton averaged 196 ng g wet
weight1 and was not correlated with the quantity of periphyton accumulated. The periphyton accumulation
rate was correlated with total nitrogen concentration at the Weeks Bay sites (r2=0.72, p<0.01, Fig. 10), but no
relationship was present for the other sites. The rate of periphyton accumulation was not significantly
correlated with total phosphorus.
Periphyton Community Composition Periphyton plates were recovered in 6 nominal recovery periods
(July 2015, August 2015, November 2015, February 2016, April 2016, and July 2016) at 12 sites. Separate
samples were collected for 6 replicates per site, providing 432 possible samples. Due to collector loss, 354
periphyton samples were obtained, providing 3 to 6 sample dates at each site. NDS experiments resulted in an
additional 124 samples, while 71 sediment samples were collected. Some 16S and 18S sequence libraries (the
collection of sequences obtained from a sample) had fewer than the desired number of sequences following
sub-sampling and were eliminated. Therefore, fewer libraries than the original number sequenced ultimately
were included in each analysis.
Eukaryote Community Composition - 18S rRNA A total of 7.8 million sequences were obtained,
which were clustered into 460,945 OTUs. Comparison with the SILVA gene database (Quast, Pruesse et al. 2013)
partially classified 61% of the OTUs, leaving 39% as unclassified Eukaryotes. Among the classified taxa, the most
abundant OTUs (18% of sequences) were classified as Ochrophyta, a clade that includes a variety algal groups.
About 50% of these were further classified as diatoms (Diatomea), while most of the remaining were classified
as Chrysophyceae (golden-brown or golden algae). Diatoms have been a focus of indicator development
26
-------
Watershed • Big Escambia A Rotten Bayou ¦ Week:
60
P
JS
60
Q
250-
200-
150 -
100-
50-
0-
Total Nitrogen (mg L 1 )
i
-a
60
E
-C
60
a
-------
Com Branch
i
•<
Date
• 2015-08-20
A 2015-11-04
¦ 2016-02-10
-|- 2016-04-05
El 2016-07-13
Treatment
• +N
• +NP
• +P
• C
To ensure comparability across samples, the
number of 18S sequences was reduced by sub-
sampling to 2,495 sequences per library, for a total
of 1.37 million sequences included in the
abundance matrix for the study. Sub-sampling
reduced the number of OTUs to 18% of the
original number, or 26,094. OTUs retained after
sub-sampling generally had an abundance of >10
prior to sub-sampling. A sufficient number and
quality of 18S sequences were obtained to support
further analysis of 489 18S libraries, of which 297
were periphyton samples, 123 were from nutrient-
diffusing substrate (NDS) experiments and 69 were
from sediments.
Fish River Reference
*
Treatment
• +NP
Date
# 2015-07-10
A 2015-08-20
¦ 2015-11-04
4" 2016-02-10
2016-04-05
^ 2016-07-13
Figure 11. Multi-dimensional scaling (MDS) ordinations
illustrating shifts in eukaryote community structure associated
with nutrient enrichment treatments at Corn Branch (upper
panel) and Fish River Reference (lower panel) on up to 6 different
dates. Ordinations were computed using Primer (v7).
Prokaryote Community Composition -
16S rRNA A sufficient number and quality of
16S sequences was obtained to support further
analysis of 483 libraries, of which 299 were from
periphyton samples, 116 were from nutrient-
diffusing substrate (NDS) experiments, and 68
were from sediments. A total of 10.5 million
sequences were obtained after initial quality
screens, which were clustered into 322,779 OTUs.
Comparison with the SILVA gene database (Quast,
Pruesse et al. 2013) classified all of the OTUs to
some degree. The sub-sampled libraries contained
240,606 OTUs, with bacteria accounting for >99%
of the sequences (the remaining being Archaea).
Abundant bacterial phyla include the
Proteobacteria (58%), Bacteriodetes (14%),
Acidobacteria (5%), Verrucomicrobia (5%), and
Actinobacteria (5%). 16S communities differed
significantly among the periphyton collectors, NDS
experiments, and sediments (ANOSIM, p<0.01).
For example, Archaea accounted for a greater
fraction of sequences (4%) in sediments, while
within the sediment bacteria the phylum
Acidobacteria were more than twice as abundant
(13%) and Bacteriodetes were half as abundant
(6%).
28
-------
4000
3000
O 1000-
2 4000-
Oh 3000-
cto
Summer 2015
i'
Fall 2015
Winter 2016
Spring 2016
JL
O
i=b
1
£
i
* ^ $ ^
*P — -*¦
C +P +N +NP C +P +N +NP C +P +N +NP C +P +N +NP C +P +N +NP
Summer 2015
>, 40 - [
'co
Q 300 -
Winter 2016
J
L,
Summer 2016
£ i
ii i
—
*
-1-
5
if
?-
Lb
#3
a
i i i i i i i i i i i i i i i i ii
C +P +N +NP C +P +N +NP C +P +N +NP C +P +N +NP C +P +N +NP
Treatment
Figure 12. For 18S libraries in each NDS experiment, the estimated number of
eukaryote classes (i.e. OTUs) present as computed using the Chao index (upper
panel) and community diversity computed as the inverse of the Simpson index.
Both metrics were computed using mother software (Schloss, Westcott et al.
2009). C=Control, +P=Phosphorus amendment, +N=Nitrogen amendment,
+NP=Nitrogen and phosphorus amendment.
Nutrient Diffusing Substrate
(NDS) Experiments Analysis of
18S sequences from NDS
experiments showed that eukaryote
communities changed in response
to nutrient treatments (i.e, Control,
+N, +P, +NP). Community
differences were especially evident
at the Fish River site (Fig. 11).
Overall, statistically significant
differences from controls resulted
from all three of the nutrient
treatments (PERMANOVA, p<0.01).
Eukaryote community richness (i.e.,
number of OTUs) and diversity
appeared to vary by site, season,
and nutrient treatment, with
apparently larger nutrient effects on
some dates (Fig. 12). Effect
estimates and statistical tests of
nutrient effects on richness and
diversity across all the experiments
were obtained using mixed effects
models, wherein site and sample
date were modeled as random
effects and nutrient treatment was
included as a fixed effect. The Chao
index, which estimates the number
of taxonomic classes (Chao 1984),
varied from 640 to 4700 and was
significantly lower in nutrient
treatments (p<0.01). Although
Chao richness was reduced due to
added N (-442, p<0.01), the effect
was nearly twice as large with
added P (-708, p<0.01) or added N
and P (-779, pc.01). Similarly,
diversity as quantified by the
inverse of the Simpson index (Fig.
12, lower panel) also changed in
response to nutrient amendments.
The inverse-Simpson index varied
from 3.7 to 297, with a median of
29
-------
41. In this context, the effects of added P (-26,
p<0.03) and added N and P (-36, p<0.01) were both
quantitatively and statistically significant, whereas
nitrogen amendments did not have a significant
effect on the inverse Simpson diversity index.
Like the eukaryotes, prokaryote communities also
shifted significantly in response to nutrient
amendments (Fig. 13, p<0.01). Analysis of Similarity
(ANOSIM) tests showed that across all NDS
experiments, communities changed due to addition
of P (p<0.05) and addition of N and P (p<0.01).
Community structure did not change significantly in
response to nitrogen-only amendments, although
shifts in response to N appeared to be larger on
certain dates (e.g., 8/20/2015). Shifts due to
nutrient amendments were apparent at both Corn
Branch and Fish River (Fig. 13), although changes
appeared to be relatively larger at Fish River
Reference compared to Corn Branch, like the
pattern observed for eukaryote communities.
Also like the eukaryotes, shifts in prokaryote
community structure (Fig. 13) were strongly
reflected in measures of species richness and
diversity (Fig. 14). Compared to controls, richness
(Chao index) decreased 1202 (31%) with added P
(p<0.01), 476 (12%) with added N (p<0.01), and
1320 (34%) with added N and P (p<0.01).
Community diversity measured via the inverse
Simpson index decreased by 58 (59%, p<0.01) with
added P, by 48 (49%) with added N and P, and by a
marginally significant 22% (p=0.06) N enrichment
only.
Corn Branch
0
Kl
a
••
it s
«
~ •
0
•
* +
1
A A
[X]
0
0
Fish River Reference
0
eF®
¦ 4
* H -
¦
f ¦ -
¦ ¦ >
Date
• 2015-08-20
A 2015-11-04
¦ 2016-02-10
-)- 2016-04-05
0 2016-07-13
Treatment
• +N
• +NP
• +P
• C
Treatment
• +N
• +NP
• +P
• C
Date
• 2015-07-10
A 2015-08-20
¦ 2015-11-04
-|- 2016-02-10
[X] 2016-04-05
2016-07-13
Figure 13. Multi-dimensional scaling (MDS) ordinations
illustrating shifts in prokaryote community structure (based on
16S sequences) associated with nutrient enrichment
treatments at Corn Branch (upper panel) and Fish River
Reference (lower panel) on up to 6 different dates.
Ordinations were computed using Primer (v7).
30
-------
* 4000-
2
Summer 2015
—
o
FaD 2015
Winter 2016
Spring 2016
Summer 2016
1
J
n
o
<=P __
ia
=p
T"
g
o.
6000-
2000
- $
9
?
C +P +N +NP C +P +N +NP C +P +N +NP C +P +N +NP C +P +N +NP
Summer 2015
300"
o
Jl
a.
Fall 2015
Winter 2016
Spring 2016
P
3
E
|3_
-
Summer 2016
£
£ 6
$ -
£
9
2?.
< *
=
ft
et3
o
o
C +P +N+NP C +P +N+NP
C +P +N+NP
Treatment
C +P +N +NP C +P +N +NP
Figure 14. For 16S libraries in each NDS experiment, the estimated number of
prokaryote classes (i.e. OTUs) present as computed using the Chao index (upper
panel) and community diversity computed as the inverse of the Simpson index.
Both metrics were computed using mother software (Schloss, Westcott et al.
2009).
Characterizing Community
Responses to Nutrients
Analysis of the eukaryote OTUs
contributing to community
dissimilarity between controls and
nutrient-amended treatments that
6,799 OTUs decreased with nutrient
amendment and were considered
"nutrient sensitive/' while 4,069
increased and were classified as
"nutrient indicative." Many larger
clades included both OTUs that
increased in response to nutrients
and other OTUs that decreased, but
some clades were dominated by
one type of response. For example,
among the Ochrophytes, an algal
clade that includes diatoms and
Chrysophytes, among other groups,
50% of the sequences were
associated with nutrient-sensitive
OTUs while only 14% were
associated with nutrient-indicative
OTUs. In contrast, 46% of the
sequences classified as
Chlorophytes, or green algae, were
associated with nutrient-indicative
OTUs, while only 23% were nutrient
sensitive. Among the Ochrophytes,
the Chrysophytes were 62%
nutrient-sensitive and only 8%
nutrient-indicative, suggesting that
soft-bodied algae like the
Chrysophytes and Chlorophytes if
classified using molecular methods
could be as useful for developing an
indicator as diatoms, which have
been utilized previously because of
the relative ease of microscopic
analysis.
For sequences classified as diatoms,
nearly twice as many (32%) were
nutrient sensitive OTUs compared
with nutrient indicative (18%), which was consistent with the literature showing most diatoms being pollution
sensitive. Within the diatoms, those genera responding most positively to nutrient amendments included the
genus Nitzschia, which is well recognized as pollution tolerant (e.g., Hill, Stevenson et al. 2001). Some diverse
genera such as Pinnularia included both nutrient-sensitive and nutrient-indicative OTUs, making them less useful
for analysis as a clade. Interestingly, three of the genera that responded most positively to nutrient
amendments include Surirella, Nitzschia and Navicula, all which Hill et al. (2001) noted are motile.
31
-------
Among the prokaryotes, approximately 30,000 OTUs were characterized as either nutrient-sensitive or nutrient-
indicative based on the NDS results. Consistent with the effect of nutrient amendments on diversity,
approximately twice as many OTUs were nutrient-sensitive rather than nutrient-indicative.
The abundance of nutrient-sensitive Archaea was nearly 10-fold greater than nutrient-indicative Archaea. In
particular, 82% of the abundance of Thaumarcheota was associated with nutrient-sensitive OTUs, while only 4%
was associated with nutrient-indicative OTUs. Other Archaea were also very unlikely to be nutrient indicative,
but were more likely to be non-responsive, or to have a relatively weak response to nutrient-amendments.
Among the bacteria, consistent nutrient responses were observed within the phyla Bacteriodetes,
Verricomicrobia, Nitrospirae, Fusobacteria, and Acidobacteria. Among the Bacteriodetes, the class Flavobacteria
was dominated by three genera (Flavobactrerium, Cloacibacterium, and Cryseobacterium) among which 90% or
more of the sequences were associated with nutrient-indicative OTUs. Flavobacteria grow in diverse habitats
and some members of the Flavobacteria are well-recognized as important fish pathogens, with well-
documented adverse effects on both natural and aquaculture (Loch and Faisal 2015). The class
Sphingobacteriia, also within the phylum Bacteriodetes, also included an abundance of nutrient-responsive
OTUs, however, this class included an abundance of both indicative and sensitive OTUs, with as many as 20
genera responding consistently (>90%) either positively or negatively to nutrient amendments. The genus
Pedobacter was abundant among the nutrient-indicative OTUs, with >99% of the OTUs classified to Pedobacter
being nutrient-indicative. A more diverse group of approximately 10 genera within the Sphingobacteriia were
consistently nutrient-sensitive, with the largest contributions from Sediminibacterium, Dinghuibacter, and an
unnamed genus-level group (env.OPS_17) collectively accounting for 43% of the abundance of nutrient-sensitive
OTUs. Similar to the Bacteriodetes, the phylum Verricomicrobia included both nutrient-sensitive and nutrient-
indicative groups, including the class Verrucomicrobiae, which was nutrient indicative and the class-level group
OPB35_soil_group, which was consistently nutrient sensitive. Together, these two clades also accounted for
nearly 70% of the sequences within the phylum. The phyla Acidobacteria and Nitrospirae were almost
universally nutrient sensitive. Although the ecology of the Acidobacteria is not especially well characterized,
higher abundance of Acidobacteria has been associated low resource availability and they are recognized at the
phylum level to be sensitive to inorganic and organic nutrients (Fierer, Lauber et al. 2012, Kielak, Barreto et al.
2016). Similarly, the Nitrospirae were overwhelmingly represented by nutrient-sensitive OTUs. Nitrospirae are
nitrite-oxidizing bacteria (Lucker, Wagner et al. 2010) and have been found to have lower abundance at sites
with higher nitrogen (Fierer, Lauber et al. 2012).
Nutrient Response Index Analysis of the prokaryote OTUs contributing to community dissimilarity
showed that 87% (210,349) had no nutrient-related effect, most likely due to many OTUs having exactly zero or
one sequence in all 6 libraries being contrasted. Among the remaining OTUs, 4% (9,963) increased on average in
nutrient-amended treatments and might be considered "nutrient indicative," whereas 8% (20,294) decreased
and might be considered "nutrient sensitive." This imbalance in numbers is reflected in the estimated decrease
in species richness and diversity associated with nutrient amendments.
The resulting 16S-based nutrient-response metric was applied to the NDS experimental treatments. The
resulting nutrient-response index values varied from -31 (lowest nutrient-impact profile) to 100 (highest
nutrient-impact profile; Fig. 15 upper panel). Nutrient index values were significantly higher than controls for all
nutrient amended treatments. The pattern of nutrient response index values varied seasonally and by station.
At Fish River, +P index values were higher than controls, like +NP, whereas +N treatments were not elevated
compared to controls except in summer. Positive nutrient index responses were observed for summer for all
three nutrient-amended treatments at both sites.
32
-------
Strong nutrient-related responses as quantified by the nutrient response index were not seen in all experiments.
For example, the Corn Branch NDS experiment in Spring 2016 produced unclear results, with controls and +NP
having similar values, whereas +P and +N responses were greater than and less than the control, respectively.
16S Nutrient Response Index
Summer 2015
£
Eukaryote community responses
(18S data), were stronger at Fish
River than at Corn Branch (Fig. 15,
lower panel). As with the 16S
data, the winter-spring nutrient
response was stronger for +P and
+NP than for +N, whereas
increases for the nutrient
response index were apparent for
+N, +P, and +NP in summer.
if
$
I8S Nutrient Response Index
Summer 2015
]i
FaU 2015
Winter 2016
Spring 2016
Summer 2016
p
CD
3
3
a
The magnitude of response to
nutrient amendments as
quantified by the nutrient
response index may reflect both
the nature of the actual NDS
nutrient response, and the index
used to quantify it. Nutrient
responses based on the average of
responses from all the NDS
experiments (Fig. 13) provide one
perspective on the magnitude of
microbial community responses to
nutrients, but community
responses observed in the NDS
experiments differed by site and
season (Fig. 14). A nutrient
response index derived from NDS
experiments at a single site -
rather than the average - is more
sensitive to changes at that site,
and therefore shows larger
responses (not shown). Similarly,
when the nutrient response index
is derived from summer NDS
experiments, it is more sensitive
to and shows larger summer
nutrient responses. Community
responses to nutrient
amendments in NDS experiments
can be contrasted via non-metric
ordinations of nutrient response
vectors. Spatial separation on the
nMDS plane (Fig. 16) points to differences in community responses. For example, prokaryote community
responses were different (e.g., involved different taxa or different magnitudes of changes) at the two sites
OS 40-
•£ 30-
1
6
-E±3
$ X
£
_ —^ cb
C +P +N +NP
Treatment
I
T
$
Figure 15. Index values representing the degree of community change related to
nutrient exposure. Higher values indicate community responses related to the
average change associated with +NP treatments as compared to controls.
33
-------
where NDS experiments were conducted (Fig. 16, upper panel). On the other hand, NDS responses from
summer experiments in 2015 and 2016 at Corn Branch were nearly coincident, indicating a similar and
characteristic site-specific summer response even though the experiments were conducted nearly a year apart.
The same results were found for summer experiments at Fish River. Ordinations for eukaryotes showed some
differences by site, but a less apparent separation by site or season. Nutrient responses for the spring 2016 Fish
River NDS experiments were very different for both 16S and 18S data, an unsurprising result given that the
richness and diversity analysis and the community ordinations both showed unusually strong community
responses to nutrient treatments.
Periphyton Community Composition The
taxonomic composition of the periphyton
community varied significantly among sites within
seasonal sampling events (Fig. 17, Fig. 18), while
multivariate similarity indicates replicable
characterizations of community composition for
replicate periphyton samples. With few
exceptions, periphyton communities at stations
within the same watershed were more similar than
at stations in different watersheds. A consistent
exception was the Fish River Reference site, which
differed in community composition from other
sites in the Weeks Bay watershed (Fig. 15, Fig. 16).
Community differences could potentially reflect a
variety of factors associated with watershed,
location, or water quality.
As with the NDS experiments, we examined how
species richness (Chao index) and diversity
(1/Simpson index), as overall measures of
community composition, could indicate nutrient-
related community responses. Whereas both
decreased with nutrient amendments in the NDS
experiments, similar results were not found across
the periphyton samples. Richness and diversity
measures in the periphyton samples were, instead,
significantly and positively related with the fraction
of agriculture in the watershed. Similar results
were obtained for total nitrogen concentration,
which, along with total phosphorus, was correlated
with agricultural land use (Fig. 5).
16S Nutrient Response
Season
£ All Date*
A I all 2015
| Spring 2016
—|— Summer 2015
fXl Summer 2016
Winter 2016
Formal. Station
<^3 Average
0 Com Branch
w tish River Referaice
18S Nutrient Response
Season
0 All Dales
A M 2015
| Spring
—{— Summer 2015
fXl Summer 2016
-7^- Winter 2016
Formal. Station
$ Average
^ Cum Bnirich
^ I 'ish River Reference
Figure 16. Non-metric multidimensional scaling ordinations of
prokaryote (16S, upper panel) and eukaryote (18S, lower panel)
community responses to +NP nutrient amendments in NDS
experiments conducted at 2 sites and 5 seasonal periods, including
the average (red) of all responses. Because community response
metrics may be negative, ordinations are based on Euclidean
distance rather than the Bray-Curtis dissimilarity used in this study
for ordination of community composition.
We applied the nutrient response indices
developed using results from the NDS experiments
at the Corn Branch and Fish River Reference sites
to all the periphyton samples (Figure 17). The range of 16S nutrient response index values across the 12
periphyton sites was smaller than was observed for experimentally manipulated nutrient exposures at the 2 NDS
test sites, but still bore some relationship to landscape and nutrient exposure. Although the 16S nutrient
response index values were not related globally with percent agriculture, values increased significantly with
34
-------
agriculture in Weeks Bay (p<0.03) and the mean value was higher in Big Escambia Creek (Fig. 16). In contrast,
the index based on 18S sequences was relatively invariant across all the periphyton sites, suggesting a degree of
orthogonality between changes in the 18S community composition in the NDS experiments and the same for the
periphyton samples.
Discussion
This study showed that nutrient indicators of stream condition based on periphyton microbial community
composition determined using a molecular approach could be a promising alternative or addition to existing
indicators for southeastern streams. With further development, these indicators could address some of the
known weaknesses of existing indicators, providing an improved scientific basis for water quality assessment
and management, including an ability to characterize and possibly predict changes in stream condition in
response to changes in nutrient loading over time. An ability to characterize biological stream condition could
reduce the impact of scientific uncertainty regarding responses to pollutants by enabling adaptive management
of stream water quality.
Several key observations comprise the basis for our optimism regarding the potential for further development
and validation of nutrient response indicators derived from DNA sequences. First, results were consistently
well-replicated. Multivariate characterizations of community composition in both periphyton and NDS
experiments showed similarity among replicate gene libraries despite several potential sources of random
variability. For example, "founder effects" on development of periphyton communities on plates could cause
divergent communities where instead similar communities are expected (Kelly, Minalt et al. 2014). Similarly,
variation associated with DNA extraction, amplification, or other artifacts of sample processing could have
introduced variability that exceeded environment-related responses. Instead, the data suggested that these
issues were not of overriding importance and that nutrient treatment or site-related effects were replicable and
easily detectible.
Desired water quality or pollution indicator characteristics include sensitivity and specificity in the response to
nutrients (or other pollutant). Nutrient diffusing substrate (NDS) experiments showed that it was possible to
stimulate and detect a strong microbial community response to nutrients in situ in the targeted streams. The
observed community responses such as decreased species richness and diversity associated with nutrient
amendments were expected based on theory and prior research. Given the strong and consistent response to
nutrients in the NDS experiments, we characterized and quantified nutrient responses (i.e., "nutrient-sensitive"
or "nutrient-indicative") for thousands of taxa and utilized these responses to compute a response index
applicable to new samples (or "gene libraries.").
These favorable developments did not come without some significant challenges. One problem relates to the
specificity and generality of the nutrient response index. Although nutrient-specificity is desired in a nutrient
indicator, the ideal index will also respond to nutrients with sufficient generality to permit application across
different sites, seasons, or perhaps different methods of periphyton collection. The nutrient response measures
derived in this study appeared to be overly station- and season-specific (Fig. 16). This would have been ideal if
the objective was to derive season-specific indices (e.g., summer or winter), rather than a more general index.
Combining responses from all 11 NDS experiments contributed to greater generality of the index, but did not
fully solve the problem. Most dramatically, the eukaryote response index, which reflected NDS amendments
very effectively (Fig. 15), detected almost no community differences among the periphyton samples (Fig. 19). In
35
-------
contrast, the 16S response measure did detect some response for prokaryotes (Fig. 19). One factor that may
have contributed to the difference is that the NDS experiments utilized upward-facing filters as a substrate,
whereas the periphyton plates were oriented horizontally and did not have a similar exposure to light. This
difference could alter the community response, potentially to a greater degree for the eukaryotes which
included a significant contribution from of algal taxa.
The prospects for detailed analysis of the data set from this study are substantial and we have left many such
possibilities to future work. For example, the SILVA database provided extensive taxonomic identifications of
the OTUs resolved in the study. The 18S sequences resolved more than 20 genera of diatoms, only a few of
which have been used extensively for indicators. Although some effort to relate diatom indicators to the gene-
based indicators could be worthwhile, it was not clear from our study that diatoms would necessarily be an
outstanding choice as indicator taxa if not for their relative suitability for microscopic identification. Several
intriguing alternatives appear possible given the genetic dataset. Other algal classes, such as the chlorophytes
are abundant and may shift in response to nutrients. The fungi are also abundant and well-resolved and are
expected to be linked to degradation pathways in soils and sediments in the catchment and stream channel.
Although nutrient responses are often presumed to begin via the "green" pathways of stimulating primary
production, gene-based indicators could as easily resolve principally "brown" pathways linked to degradation of
complex organic substrates.
Shifts in the abundance of major groups of Bacteria and Archaea, such as whole phyla, classes, or groups of
genera were generally consistent with the limited information about the ecology of such groups. While a focus
on such groups could have the negative effect of obscuring ecological function within functionally diverse
taxonomic groups, there are two likely benefits. First, evaluation of responses by such identified groups could
provide a pathway to the generality that the OTU-based index lacked. As importantly, taxonomic classification
links the results of this study to future studies or to implementation in environmental assessment. Thus, an
important objective in future work is to establish and validate an index of nutrient response based on relative
abundance of taxonomic groups not on the matrix of OTU abundance.
Another question of interest is whether environmental agencies or similar groups charged with water quality
assessment could expect to be able to implement DNA-based indicators. While some care must be taken to
properly collect and process the samples, the required sampling equipment and field protocols were not
exceptionally challenging. Costs associated with preparation and sequencing have been falling rapidly and are
likely less than microscopic analysis. Procedures for bioinformatic analysis and subsequent multivariate analysis
are not routine and require both specialized skills and powerful computational resources. It is reasonable,
however, to expect that these will become more routine over time. Nonetheless, further development and
validation will be required, including testing in streams affected to a greater degree by other sources of nutrient
pollution such as urban stormwater runoff and wastewater discharges.
There is little or no precedent for using the community composition of Archaea or Bacteria as an indicator of
biotic responses to nutrients or other pollutants, especially in an environmental management context.
Indicators based on eukaryotes, such as algae have precedent, but were very limited. Today, as Forney et al.
(2004) note, we are in the "land of the one-eyed king," where all the subjects are blind, and the one with a
single eye to see is therefore "king." Molecular methods, though imperfect, give us one eye to see the rich
diversity of the microbial community, and the important signals that may provide about environmental
responses to pollutants and other stressors.
36
-------
July 2015
August 2015
4?
*
November 2015
A
April 2016
A
] A
*
X
February 2016
Q
July 2016
ffl
Watershed
. 5 BigEscambia
Rotten Bayou
$ Weeks Bay
Formal. Station
(3 Baker Branch
/\ Bayou Lasallc
—|— Big Escambia Creek Lower
X Big Escambia Creek Upper West
<^/» Boardpile Branch
Corn Branch
[x] Hscambia Creek Northeast
Fish River Reference
'\[/> Kiln Delisle
(J) Magnolia Tributary
X^X Orphan Creek
\-\-\ Wet Weather Creek
*
Figure 17. Non-metric multi-dimensional scaling ordinations of prokaryote (16S) sequences from
periphyton collected at each of 12 stream locations. Identical colors and symbols within a panel indicate
replicate analysis at each station and date.
37
-------
July 2015
August 2015
-f
November 2015
e
February 2016
If
1
We
o
0
April 2016
July 2016
©
Watershed
Big Escambia
£ Rotten Bayou
£ Weeks Bay
Formal. Station
(3 Baker Branch
/\ Bayou Lasalle
—|— Big Escambia Creek Lower
Big Escambia Creek Upper West
<^> Boardpile Branch
Com Branch
|^| Escambia Creek Northeast
Fish River Reference
^ Kiln Delisle
£ ^ MagnoUa Tributary
y(j(, Orphan Creek
[-f-| Wet Weather Creek
Figure 18. Non-metric multi-dimensionai scaling ordinations of eukaryote (18S) sequences from
periphyton collected at each of 12 stream locations. Identical colors and symbols within a panel
indicate replicate analysis at each station and date.
38
-------
Watershed Big Escambia Rotten Bayou . Weeks Bay
5-
d -5 -
-o
c
o
CL
Cfl
¦4—>
c
«D
*c
3
z:
si7
<~
i
5-
M * 1"
oc
01
-5-
25 50 75
Percent Agriculture
Figure 19. Nutrient response index values based on 16S sequences (upper panel) and 18S
sequences (lower panel) applied to periphyton samples and related to percent agriculture in
the catchment.
39
-------
References
Amaral-Zettler, L. A., E. A. McCliment, H. W. Ducklow and S. M. Huse (2009). "Correction: A Method for Studying
Protistan Diversity Using Massively Parallel Sequencing of V9 Hypervariable Regions of Small-Subunit Ribosomal
RNA Genes." PLoS ONE 4(12).
APHA (1989). Standard Methods for the Examination of Water and Wastewater. Washington, DC, APHA
(American Public Health Association).
APHA (2005). 10300 C 6 Periphyton, Sample Analysis, Dry and Ash-Free Weight. Standard methods for the
examination of water and wastewater. A. D. Eaton. Washington, DC, APHA-AWWA-WEF.
Apprill, A., S. McNally, R. Parsons and L. Weber (2015). "Minor revision to V4 region SSU rRNA 806R gene primer
greatly increases detection of SAR11 bacterioplankton." Aquatic Microbial Ecology 75(2): 129-137.
Barbour, M. T., J. Gerritsen, G. E. Griffith, R. Frydenborg, E. McCarron, J. S. White and M. L. Bastian (1996). "A
framework for biological criteria for Florida streams using benthic macroinvertebrates." Journal of the North
American Benthological Society 15(2): 185-211.
Caporaso, J. G., C. L. Lauber, W. A. Walters, D. Berg-Lyons, J. Huntley, N. Fierer, S. M. Owens, J. Betley, L. Fraser,
M. Bauer, N. Gormley, J. A. Gilbert, G. Smith and R. Knight (2012). "Ultra-high-throughput microbial community
analysis on the lllumina HiSeq and MiSeq platforms." ISME J 6(8): 1621-1624.
Chao, A. (1984). "Nonparametric-Estimation of the Number of Classes in a Population." Scandinavian Journal of
Statistics 11(4): 265-270.
Davidson, E., M. David, J. Galloway, C. Goodale, R. Haeuber, J. Harrison, R. Howarth, D. Jaynes, R. Lowrance, B.
Nolan, J. Peel, R. Pinder, E. Porter, C. Snyder, A. Townsend and M. Ward (2012). Excess nitrogen in the U.S.
Environment: Trends, Risks, and Solutions. Issues in Ecology. T. E. S. o. America. 15: 17.
De Jonckheere, J. F. (2014). "What do we know by now about the genus Naegleria?" Exp Parasitol 145 Suppl: S2-
9.
Environmental Protection Agency (2014). U.S. EPA Expert Workshop: Nutrient Enrichment Indicators in Streams.
Proceedings April 16-18, 2013. EPA-822-R-14-004. Washington, DC, US Environmental Protection Agency: 61 pp.
Fierer, N., C. L. Lauber, K. S. Ramirez, J. Zaneveld, M. A. Bradford and R. Knight (2012). "Comparative
metagenomic, phylogenetic and physiological analyses of soil microbial communities across nitrogen gradients."
ISME J 6(5): 1007-1017.
Fore, L. S. (2004). Decelopment and Testing of Biomonitoring Tools for Macroinvertebrates in Florida Streams.
Tallahassee, FL, Florida Department of Environmental Protection: 74 pp.
Fore, L. S. (2010). Evaluation of Stream Periphyton as Indicators of Biological Condition for Florida Streams.
Tallahassee, FL, Florida Department of Environmental Protection: 48 pp.
Forney, L. J., X. Zhou and C. J. Brown (2004). "Molecular microbial ecology: land of the one-eyed king." Curr Opin
Microbiol 7(3): 210-220.
Garcia, A., A. Hoos and S. Terziotti (2011). "A regional modeling framework of phosphorus sources and transport
in streams of the southeastern United States." Journal of the American Water Resources Association 47(5): 991-
1010.
Goodnight, C. J. (1973). "The Use of Aquatic Macroinvertebrates as Indicators of Stream Pollution." Transactions
of the American Microscopical Society 92(1): 1-13.
40
-------
Hauer, F. R. and G. A. Lamberti (2011). Methods in Stream Ecology, 2nd Edition. New York, Academic Press.
Hill, B. H., R. J. Stevenson, Y. D. Pan, A. T. Herlihy, P. R. Kaufmann and C. B. Johnson (2001). "Comparison of
correlations between environmental characteristics and stream diatom assemblages characterized at genus and
species levels." Journal of the North American Benthological Society 20(2): 299-310.
Hirsch, R. M., D. L. Moyer and S. A. Archfield (2010). "Weighted Regression on Time, Discharge and Seaon
(WRTDS), with An Application to Chesapeake Bay River Inputs." Journal of the American Water Resources
Association 46(5): 857-880.
Holmes, R. M., A. Aminot, R. Kerouel, B. A. Hooker and B. J. Peterson (1999). "A simple and precise method for
measuring ammonium in marine and freshwater ecosystems." Canadian Journal of Fisheries and Aquatic
Sciences 56(10): 1801-1808.
Homer, C. G., J. A. Dewitz, L. Yang, S. Jin, P. Danielson, G. Xian, J. Coulston, N. D. Herold, J. D. Wickham and K.
Megown (2015). "Completion of the 2011 National Land Cover Database for the conterminous United States-
Representing a decade of land cover change information." Photogrammetric Engineering and Remote Sensing
81(5): 345-354.
Hoos, A. and G. McMahon (2009). "Spatial analysis of instream nitrogen loads and factors controlling nitrogen
delivery to streams in the southeastern United States using spatially referenced regression on watershed
attributes (SPARROW) and regional classification frameworks." Hvdrological Processes 23: 2275-2294.
Karr, J. R. (1981). "Assessment of Biotic Integrity Using Fish Communities." Fisheries 6(6): 21-27.
Kelly, J. J., N. Minalt, A. Culotti, M. Pryor and A. Packman (2014). "Temporal variations in the abundance and
composition of biofilm communities colonizing drinking water distribution pipes." PLoS One 9(5): e98542.
Kielak, A. M., C. C. Barreto, G. A. Kowalchuk, J. A. van Veen and E. E. Kuramae (2016). "The Ecology of
Acidobacteria: Moving beyond Genes and Genomes." Front Microbiol 7: 744.
Klindworth, A., E. Pruesse, T. Schweer, J. Peplies, C. Quast, M. Horn and F. O. Glockner (2013). "Evaluation of
general 16S ribosomal RNA gene PCR primers for classical and next-generation sequencing-based diversity
studies." Nucleic Acids Res 41(1): el.
Leff, J. W., S. E. Jones, S. M. Prober, A. Barberan, E. T. Borer, J. L. Firn, W. S. Harpole, S. E. Hobbie, K. S.
Hofmockel, J. M. Knops, R. L. McCulley, K. La Pierre, A. C. Risch, E. W. Seabloom, M. Schutz, C. Steenbock, C. J.
Stevens and N. Fierer (2015). "Consistent responses of soil microbial communities to elevated nutrient inputs in
grasslands across the globe." Proc Natl Acad Sci U S A 112(35): 10967-10972.
Lisa, J. A., B. Song, C. R. Tobias and D. E. Hines (2015). "Genetic and biogeochemical investigation of sedimentary
nitrogen cycling communities responding to tidal and seasonal dynamics in Cape Fear River Estuary." Estuarine.
Coastal and Shelf Science 167: A313-A323.
Loch, T. P. and M. Faisal (2015). "Emerging flavobacterial infections in fish: A review." J Adv Res 6(3): 283-300.
Lucker, S., M. Wagner, F. Maixner, E. Pelletier, H. Koch, B. Vacherie, T. Rattei, J. S. Damste, E. Spieck, D. Le Paslier
and H. Daims (2010). "A Nitrospira metagenome illuminates the physiology and evolution of globally important
nitrite-oxidizing bacteria." Proc Natl Acad Sci U S A 107(30): 13479-13484.
McCormick, P. V. and R. J. Stevenson (1998). "Periphyton as a tool for ecological assessment and management in
the Florida Everglades." Journal of Phvcology 34(5): 726-733.
Millenium Ecosystem Assessment (2005). Ecosystems and Human Well-being: Synthesis. Washington, DC, World
Resources Institute: 155.
41
-------
Oksanen, J., F. G. Blanchet, M. Friendly, R. Kindt., P. Legendre, D. McGlinn, P. R. Minchin, R. B. O'Hara, G. L.
Simpson, P. Solymos, M. H. H. Stevens, E. Szoecs and H. Wagner (2016) "vegan: community ecology package. R.
package version 2.4-0.".
Ortmann, A. C. and T. T. Santos (2016). "Spatial and temporal patterns in the Pelagibacteraceae across an
estuarine gradient." FEMS Microbiol Ecol 92(9): 1-9.
Paerl, H. W., J. Dyble, P. H. Moisander, R. T. Noble, M. F. Piehler, J. L. Pinckney, T. F. Steppe, L. Twomey and L. M.
Valdes (2003). "Microbial indicators of aquatic ecosystem change: current applications to eutrophication
studies." FEMS Microbiol Ecol 46(3): 233-246.
Patton, C. J., A. E. Fischer, W. H. Campbell and E. R. Campbell (2002). "Corn leaf nitrate reductase-a nontoxic
alternative to cadmium for photometric nitrate determinations in water samples by air-segmented continuous-
flow analysis." Environ Sci Technol 36(4): 729-735.
Quast, C., E. Pruesse, P. Yilmaz, J. Gerken, T. Schweer, P. Yarza, J. Peplies and F. O. Glockner (2013). "The SILVA
ribosomal RNA gene database project: improved data processing and web-based tools." Nucleic Acids Res
41(Database issue): D590-596.
Schloss, P. D., S. L. Westcott, T. Ryabin, J. R. Hall, M. Hartmann, E. B. Hollister, R. A. Lesniewski, B. B. Oakley, D.
H. Parks, C. J. Robinson, J. W. Sahl, B. Stres, G. G. Thallinger, D. J. Van Horn and C. F. Weber (2009). "Introducing
mothur: open-source, platform-independent, community-supported software for describing and comparing
microbial communities." Appl Environ Microbiol 75(23): 7537-7541.
Solorzano, L. and J. H. Sharp (1980). "Determination of total dissolved phosphorus and particulate phosphorus in
natural watersl." Limnology and Oceanography 25(4): 754-758.
Stevenson, J. (2014). "Ecological assessments with algae: a review and synthesis." J Phycol 50(3): 437-461.
Stevenson, R. J. and C. Decker (2012). Diatom Responses to Nutrient Conditions in Region 4 Streams. Atlanta,
GA, US EPA Region 4: 74 pp.
Stevenson, R. J., Y. Pan, K. M. Manoylov, C. A. Parker, D. P. Larsen and A. T. Herlihy (2008). "Development of
diatom indicators of ecological conditions for streams of the western US." Journal of the North American
Benthological Society 27(4): 1000-1016.
Tanaka, T., K. Kawasaki, S. Daimon, W. Kitagawa, K. Yamamoto, H. Tamaki, M. Tanaka, C. H. Nakatsu and Y.
Kamagata (2014). "A hidden pitfall in the preparation of agar media undermines microorganism cultivability."
Appl Environ Microbiol 80(24): 7659-7666.
Tremblay, J., K. Singh, A. Fern, E. S. Kirton, S. He, T. Woyke, J. Lee, F. Chen, J. L. Dangl and S. G. Tringe (2015).
"Primer and platform effects on 16S rRNA tag sequencing." Front Microbiol 6: 771.
Vitousek, P. M., J. D. Aber, R. W. Howarth, G. E. Likens, P. A. Matson, D. W. Schindler, W. H. Schlesinger and D.
Tilman (1997). "Human alteration of the global nitrogen cycle: Sources and consequences." Ecological
Applications 7(3): 737-750.
Wan, Y., Y. Qian, K. W. Migliaccio, Y. Li and C. Conrad (2014). "Linking Spatial Variations in Water Quality with
Water and Land Management using Multivariate Techniques." J Environ Qual 43(2): 599-610.
Welschmeyer, N. A. (1994). "Fluorometric Analysis of chlorophyll a in the presence of chlorophyll b and
pheopigments." Limnology and Oceanography 39(8): 1985-1992.
42
-------
*>EPA
United States
Environmental Protection
Agency
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35
Office of Research and
Development (8101R)
Washington, DC 20460
Offal Business
Penalty for Private Use
$300
------- |