# MICROBIOLOGY OF THE RAPIDLY CHANGING POLAR ENVIRONMENTS

EDITED BY : Julie Dinasquet, Eva Ortega-Retuerta, Connie Lovejoy and Ingrid Obernosterer PUBLISHED IN : Frontiers in Marine Science and Frontiers in Microbiology

#### Frontiers Copyright Statement

© Copyright 2007-2018 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.

The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.

Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.

Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.

As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.

All copyright, and all rights therein, are protected by national and international copyright laws.

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88945-513-3 DOI 10.3389/978-2-88945-513-3

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

#### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

#### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# MICROBIOLOGY OF THE RAPIDLY CHANGING POLAR ENVIRONMENTS

Topic Editors:

Julie Dinasquet, Scripps Institution of Oceanography, University of California, San Diego, United States Eva Ortega-Retuerta, CNRS, Sorbonne Universities, France Connie Lovejoy, Laval University, Canada Ingrid Obernosterer, CNRS, Sorbonne Universities, France

Marginal ice zone, sea ice and meltwater in the Baffin Bay, Arctic Ocean, June 2016. Image: Julie Dinasquet.

Marine and freshwater polar environments are characterized by intense physical forces and strong seasonal variations. The persistent cold and sometimes inhospitable conditions create unique ecosystems and habitats for microbial life. Polar microbial communities are diverse productive assemblages, which drive biogeochemical cycles and support higher food-webs across the Arctic and over much of the Antarctic. Recent studies on the biogeography of microbial species have revealed phylogenetically diverse polar ecotypes, suggesting adaptation to seasonal darkness, sea-ice coverage and high summer irradiance. Because of the diversity of habitats related to atmospheric and oceanic circulation, and the formation and melting of ice, high latitude oceans and lakes are ideal environments to investigate composition and functionality of microbial communities. In addition, polar regions are responding more dramatically to climate change compared to temperate environments and there is an urgent need to identify sensitive indicators of ecosystem history, that may be sentinels for change or adaptation. For instance, Antarctic lakes provide useful model systems to study microbial evolution and climate history. Hence, it becomes essential and timely to better understand factors controlling the microbes, and how, in turn, they may affect the functioning of these fragile ecosystems.

Polar microbiology is an expanding field of research with exciting possibilities to provide new insights into microbial ecology and evolution. With this Research Topic we seek to bring together polar microbiologists studying different aquatic systems and components of the microbial food web, to stimulate discussion and reflect on these sensitive environments in a changing world perspective.

Citation: Dinasquet, J., Ortega-Retuerta, E., Lovejoy, C., Obernosterer, I., eds. (2018). Microbiology of the Rapidly Changing Polar Environments. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-513-3

# Table of Contents

*07 Editorial: Microbiology of the Rapidly Changing Polar Environments* Julie Dinasquet, Eva Ortega-Retuerta, Connie Lovejoy and Ingrid Obernosterer

#### CHANGES IN POLAR PHYTOPLANKTON COMMUNITIES DYNAMICS


Archana R. Meshram, Anna Vader, Svein Kristiansen and Tove M. Gabrielsen

*50 Atlantic Advection Driven Changes in Glacial Meltwater: Effects on Phytoplankton Chlorophyll-a and Taxonomic Composition in Kongsfjorden, Spitsbergen*

Willem H. van De Poll, Douwe S. Maat, Philipp Fischer, Patrick D. Rozema, Oonagh B. Daly, Sebastiaan Koppelle, Ronald J. W. Visser and Anita G. J. Buma


Maria Vernet, Tammi L. Richardson, Katja Metfies, Eva-Maria Nöthig and Ilka Peeken

*108 Protist Communities in Moored Long-Term Sediment Traps (Fram Strait, Arctic)–Preservation With Mercury Chloride Allows for PCR-Based Molecular Genetic Analyses*

Katja Metfies, Eduard Bauerfeind, Christian Wolf, Pim Sprong, Stephan Frickenhaus, Lars Kaleschke, Anja Nicolaus and Eva-Maria Nöthig

#### CHANGES IN POLAR HETEROTROPHIC MICROBIAL COMMUNITIES DYNAMICS

*121 A Decadal (2002–2014) Analysis for Dynamics of Heterotrophic Bacteria in an Antarctic Coastal Ecosystem: Variability and Physical and Biogeochemical Forcings*

Hyewon Kim and Hugh W. Ducklow

*139 Seasonal Succession of Free-Living Bacterial Communities in Coastal Waters of the Western Antarctic Peninsula*

Catherine M. Luria, Linda A. Amaral-Zettler, Hugh W. Ducklow and Jeremy J. Rich

*152 Changes in Marine Prokaryote Composition With Season and Depth Over an Arctic Polar Year*

Bryan Wilson, Oliver Müller, Eva-Lena Nordmann, Lena Seuthe, Gunnar Bratbak and Lise Øvreås


Mary Thaler, Warwick F. Vincent, Marie Lionard, Andrew K. Hamilton and Connie Lovejoy

*199 Changes of the Bacterial Abundance and Communities in Shallow Ice Cores from Dunde and Muztagata Glaciers, Western China*

Yong Chen, Xiang-Kai Li, Jing Si, Guang-Jian Wu, Li-De Tian and Shu-Rong Xiang

*215 Uptake of Leucine, Chitin, and Iron by Prokaryotic Groups During Spring Phytoplankton Blooms Induced by Natural Iron Fertilization off Kerguelen Island (Southern Ocean)*

Marion Fourquez, Sara Beier, Elanor Jongmans, Robert Hunter and Ingrid Obernosterer

*228 Viruses and Protists Induced-Mortality of Prokaryotes Around the Antarctic Peninsula During the Austral Summer* Dolors Vaqué, Julia A. Boras, Francesc Torrent-Llagostera, Susana Agustí, Jesús M. Arrieta, Elena Lara, Yaiza M. Castillo, Carlos M. Duarte and Maria M. Sala

# ARCTIC MICROBIOLOGY

*240 Microbial Community Response to Terrestrially Derived Dissolved Organic Matter in the Coastal Arctic*

Rachel E. Sipler, Colleen T. E. Kellogg, Tara L. Connelly, Quinn N. Roberts, Patricia L. Yager and Deborah A. Bronk

*259 Carbon Bioavailability in a High Arctic Fjord Influenced by Glacial Meltwater, NE Greenland*

Maria L. Paulsen, Sophia E. B. Nielsen, Oliver Müller, Eva F. Møller, Colin A. Stedmon, Thomas Juul-Pedersen, Stiig Markager, Mikael K. Sejr, Antonio Delgado Huertas, Aud Larsen and Mathias Middelboe

*278 Upstream Freshwater and Terrestrial Sources are Differentially Reflected in the Bacterial Community Structure Along a Small Arctic River and its Estuary*

Aviaja L. Hauptmann, Thor N. Markussen, Marek Stibal, Nikoline S. Olsen, Bo Elberling, Jacob Bælum, Thomas Sicheritz-Pontén and Carsten S. Jacobsen

#### ANTARCTIC BENTHIC MICROBIOLOGY

*294 Biogeochemical and Microbial Variation Across 5500 km of Antarctic Surface Sediment Implicates Organic Matter as a Driver of Benthic Community Structure*

Deric R. Learman, Michael W. Henson, J. Cameron Thrash, Ben Temperton, Pamela M. Brannock, Scott R. Santos, Andrew R. Mahon and Kenneth M. Halanych

*305 High Prevalence of Gammaproteobacteria in the Sediments of Admiralty Bay and North Bransfield Basin, Northwestern Antarctic Peninsula* Diego C. Franco, Camila N. Signori, Rubens T. D. Duarte, Cristina R. Nakayama, Lúcia S. Campos and Vivian H. Pellizari

# Editorial: Microbiology of the Rapidly Changing Polar Environments

Julie Dinasquet 1,2 \*, Eva Ortega-Retuerta<sup>2</sup> , Connie Lovejoy <sup>3</sup> and Ingrid Obernosterer <sup>2</sup>

<sup>1</sup> Marine Biology Research Division, Scripps Institution of Oceanography, University of California, San Diego, La Jolla, CA, United States, <sup>2</sup> Laboratoire d'Océanographie Microbienne, LOMIC, Observatoire Océanologique de Banyuls sur mer, CNRS, Sorbonne Universités, Banyuls sur mer, France, <sup>3</sup> Département de Biologie, Université Laval, Québec, QC, Canada

Keywords: Arctic, Antarctica, polar microbiology, aquatic microbiology, climate change

**Editorial on the Research Topic**

#### **Microbiology of the Rapidly Changing Polar Environments**

Polar environments are warming at alarming rates (Tingley and Huybers, 2013; Schmidtko et al., 2014). The rapid warming is a threat to the integrity of these ice-influenced ecosystems, where microbes are the dominant life form (Boetius et al., 2015; Cavicchioli, 2015; Pedros-Alio et al., 2015; Mohit et al., 2017). Polar regions play a crucial role in regulating the climate system, and have acted as an anthropogenic CO<sup>2</sup> sink due to combinations of low temperatures and high gas solubility, high winds, extensive winter sea ice cover, and intense biological productivity (Arrigo et al., 2008; Bates and Mathis, 2009). Despite being geographically distant, the Arctic and the Antarctic aquatic ecosystems have much in common. Both regions are subjected to harsh environmental conditions such as low temperatures and darkness in winter, extreme seasonal shifts in solar radiation, high UV exposure in summer, and seasonal changes in ice cover on both lakes and seas. In spite of this apparent inhospitality, polar aquatic ecosystems host diverse and active microbial communities that drive biogeochemical cycles and support higher trophic levels (Cavicchioli, 2015; Pedros-Alio et al., 2015). Since polar regions are more strongly impacted by global change (Yoshimori et al., 2017) than temperate regions, there is an urgent need to understand how these perturbations will affect microbial functions and major elemental cycles, including carbon. Marine and freshwater high-latitude aquatic environments offer an ideal variety of habitats to investigate changes in the community composition and functionality of aquatic microbes, who may be true sentinels of global change. The microbiology across many habitats in these vulnerable ecosystems are the subject of this special topic in Frontiers in Marine Science and Microbiology (Aquatic Microbiology).

The goal of this topic was to bring together contributions that provide a "bi-polar view" of aquatic microbiology. The range of contributions highlights both similar and distinct microbial processes that take place within the two high-latitude extreme environments. This research topic touches upon a wide range of interests to polar microbiologists, and provides a collection of 20 articles, including: a review, in situ observations, experiments, technical notes and modeling studies. While most of the articles focus on the response of bacterial communities to environmental changes, several look into the response of microbial eukaryotes, and others take a more biogeochemical approach that includes examining carbon and nutrient dynamics. Finally, we are proud of the high contribution by women to this research topic (the four scientific editors, 75% of first authors, 47% of all authors, 32% of reviewers), which highlights the important contribution of women to polar science in general and microbial ecology in particular.

#### Edited by:

Jonathan P. Zehr, University of California, Santa Cruz, United States

#### Reviewed by:

Patricia Lynn Yager, University of Georgia, United States

> \*Correspondence: Julie Dinasquet jdinasquet@ucsd.edu

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Marine Science

Received: 31 December 2017 Accepted: 13 April 2018 Published: 04 May 2018

#### Citation:

Dinasquet J, Ortega-Retuerta E, Lovejoy C and Obernosterer I (2018) Editorial: Microbiology of the Rapidly Changing Polar Environments. Front. Mar. Sci. 5:154. doi: 10.3389/fmars.2018.00154

In contrast to lower latitude regions, annual pelagic primary production in polar ecosystems is constrained by relatively short summers. It is during this brief summer period that light conditions are favorable for the phytoplankton growth, but this only can occur if nutrients are available. Recently, earlier spring-retreat of sea ice, along with the decrease in seasonal sea ice cover especially in the Arctic has led to increased light availability, which could increase primary production. However, increased stratification from meltwater may result in decreased nutrient supply to the euphotic zone. Earlier studies suggested that meltwater itself can sometimes increase nutrient supply to the phytoplankton in the Southern Ocean, and that higher winds across the open water in the Arctic could promote fall blooms. Such changes in phytoplankton phenology could have a cascading effect on other biological activities, potentially perturbing the polar carbon cycle. In this special topic, the review by Deppeler and Davidson concluded that long-term changes in Antarctic phytoplankton dynamics will depend on the magnitude and timing of the climate change induced stressors. In the Arctic, Meshram et al. show that under-ice bloom composition is related to both water mass mixing and local processes. Warming also influences melt water inputs and water column stratification, governing bloom intensities and composition, as reported by van De Poll et al. Current and future environmental changes may shift dominant phytoplankton groups toward smaller species (Paulsen et al.; Onda et al.; Vernet et al.). As changes in the relative size of the dominant primary producers appears to favor small mixotrophic species, tools to assess their trophic interactions and carbon fluxes should be developed. The need to establish baselines was raised by Metfies et al. who propose exploiting preserved sediment trap samples using molecular markers to increase the length of records and facilitate comparisons to more recent data. Looking toward future trends, Vernet et al. used a modeling approach to suggest that community changes could be neutral and perhaps may not impact carbon export budgets.

In polar aquatic ecosystems, the seasonal dynamics of heterotrophic microbial communities are coupled with phytoplankton blooms that produce pulses of labile organic matter. As part of this research topic, several studies report bacterial responses to seasonal and inter-annual variability. Around the western Antarctic Peninsula, seasonal variation in bacterial activities and communities indicate strong coupling with phytoplankton blooms (Kim and Ducklow; Luria et al.). But bacterial heterotrophic production, key to food web functioning, may be fueled by additional sources of carbon (Kim and Ducklow). In the Arctic Svalbard Archipelago, epipelagic microbial communities are also tightly coupled to phytoplankton blooms while seasonal variations do not appear to affect mesopelagic communities (Wilson et al.). Community changes following seasonal ice cover variability were also observed in epishelf and inland lakes where climate change may have major impacts on biogeochemical cycles (Schütte et al.; Thaler et al.). Similar seasonal responses are also observed in alpine glacier ecosystems (Chen et al.). The studies mentioned above, suggest that polar and cold climate microbial communities should be closely monitored as means to anticipate future effects and potential feedback loops on polar ecosystem functioning.

Shifts in heterotrophic microbial communities may also reflect specific community capacities to use seasonal pulses of resources, such as organic carbon and iron, and induce microbial interactions (Fourquez et al.). Vaqué et al. demonstrated that viral infection is more sensitive to climate change and warming compared to grazing, implying that not only variation in community composition and activity, but also bacterial mortality agents will affect the fate of the carbon fluxes through microbial food webs.

Whereas, similar responses to seasonal variability are observed between the Arctic and the Antarctic, future changes may amplify differences between these regions. Compared to the Southern Ocean, the Arctic is largely surrounded by land masses and with large river inputs (∼4,000 km<sup>3</sup> yr−<sup>1</sup> ). These rivers are a source of low-salinity waters, organic matter, and nutrients to the Arctic Ocean and surrounding seas. Moreover, as temperature rises, terrestrial permafrost thaws, which mobilizes ancient carbon that enters the ocean through freshwater discharge. Here, Sipler et al. show that this terrestrial and riverine organic matter is rapidly degraded by bacterial activity even before reaching open waters suggesting a need to revisit the notion that this carbon should be considered as mainly refractory (Kirchman et al., 2009). Nevertheless, phytoplankton derived organic matter remains a major source of carbon fuelling Arctic bacterial production in spring and summer (Paulsen et al.). But as the Arctic warms and freshwater discharge increases, associated pulses of organic matter supplementing bacterial carbon demand (Paulsen et al.; Sipler et al.) may push Arctic microbial communities toward heterotrophy (Paulsen et al.). Increasing riverine input is also associated with changes in bacterial community structure, as specific organisms respond to the terrestrial derived organic matter (Sipler et al.). Hauptmann et al. point out that the transport of freshwater communities into the coastal marine waters could also affect the functioning of the arctic carbon cycle.

Compared to the Southern Ocean, the Arctic Ocean is a semi-closed system, with limited exchanges, through inflow and outflow gateways, with neighboring oceans. Of concern is the recent increase in Atlantic inflow into the European Arctic Ocean (Polyakov et al., 2017), which has been attributed to climate change. The northward intrusion of warmer waters may affect the spring bloom dynamics (van De Poll et al.; Vernet et al.) and favor the establishment of organisms previously absent from these waters, such as the cyanobacterium Synechoccocus (Paulsen et al.). Hence, changes in ocean circulation can have major effects on the biology of the pelagic ecosystem.

Lastly, polar benthic ecosystems are not likely to be spared from climate changes as seasonal shifts in surface communities are associated with microbial communities in the sediments (Learman et al.; Franco et al.). An increase in phytoplankton biomass and subsequent carbon export could cause microbial communities in polar sediments to switch from predominantly lithotrophs to organic matter degraders, which will in turn affect benthic elemental cycles (Learman et al.).

While Arctic sea ice extent breaks record lows, and Arctic and Antarctic ice shelves collapse, it is urgent to understand how these dramatic changes will impact the microbial ecology and potentially disrupt the capacity for carbon storage in polar regions. Overall, the studies published in this research topic confirm that major changes already occurring in aquatic polar ecosystems will impact microbial communities, and further suggest that some microbes are potential indicators of these changes. Because most of these studies focus on biodiversity, however, it remains a challenge to unravel how these community changes will alter ecosystem function. Some of the present results hint toward major changes in carbon fluxes and trophic interactions, stressing the need for increased spatial and seasonal coverage, including winter sampling of microbial communities, and continued focus on top-down and bottom up controls of microbial processes. Long-term monitoring is needed to facilitate reliable predictions on how polar aquatic systems will interact with climate and better assess the resilience and adaptation of these fragile ecosystems. Moreover, as polar microbial communities are fundamental to carbon cycling, informed modeling approaches that integrate polar microbial

#### REFERENCES


complexity are needed to understand the implications of climate change.

# AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

#### FUNDING

This work was supported by the Marie Curie Actions-International Outgoing Fellowship (PIOF-GA-2013-629378) to JD and Intra-European Fellowship (H2020-MSCA-IF-2015- 703991) to EO-R. CL acknowledges support from the Canadian Natural Science and Engineering Council (NSERC) discovery program and the Canada First Research Excellence Fund supporting Sentinel North.

#### ACKNOWLEDGMENTS

We would like to thank the editorial staff at Frontiers in marine sciences and in aquatic microbiology for their initial invitation and support throughout.

doi: 10.1126/science.aai8204


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer PLY declared a past co-authorship with one of the authors, JD, to the handling Editor.

Copyright © 2018 Dinasquet, Ortega-Retuerta, Lovejoy and Obernosterer. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Southern Ocean Phytoplankton in a Changing Climate

Stacy L. Deppeler <sup>1</sup> \* and Andrew T. Davidson2, 3

1 Institute for Marine and Antarctic Studies, University of Tasmania, Hobart, TAS, Australia, <sup>2</sup> Australian Antarctic Division, Department of the Environment and Energy, Kingston, TAS, Australia, <sup>3</sup> Antarctic Climate and Ecosystem Cooperative Research Centre (ACE CRC), University of Tasmania, Hobart, TAS, Australia

Phytoplankton are the base of the Antarctic food web, sustain the wealth and diversity of life for which Antarctica is renowned, and play a critical role in biogeochemical cycles that mediate global climate. Over the vast expanse of the Southern Ocean (SO), the climate is variously predicted to experience increased warming, strengthening wind, acidification, shallowing mixed layer depths, increased light (and UV), changes in upwelling and nutrient replenishment, declining sea ice, reduced salinity, and the southward migration of ocean fronts. These changes are expected to alter the structure and function of phytoplankton communities in the SO. The diverse environments contained within the vast expanse of the SO will be impacted differently by climate change; causing the identity and the magnitude of environmental factors driving biotic change to vary within and among bioregions. Predicting the net effect of multiple climate-induced stressors over a range of environments is complex. Yet understanding the response of SO phytoplankton to climate change is vital if we are to predict the future state/s of the ecosystem, estimate the impacts on fisheries and endangered species, and accurately predict the effects of physical and biotic change in the SO on global climate. This review looks at the major environmental factors that define the structure and function of phytoplankton communities in the SO, examines the forecast changes in the SO environment, predicts the likely effect of these changes on phytoplankton, and considers the ramifications for trophodynamics and feedbacks to global climate change. Predictions strongly suggest that all regions of the SO will experience changes in phytoplankton productivity and community composition with climate change. The nature, and even the sign, of these changes varies within and among regions and will depend upon the magnitude and sequence in which these environmental changes are imposed. It is likely that predicted changes to phytoplankton communities will affect SO biogeochemistry, carbon export, and nutrition for higher trophic levels.

Keywords: Southern Ocean, phytoplankton, climate change, primary productivity, Antarctica

# 1. INTRODUCTION

Iconic Antarctic wildlife from krill to whales, seals, penguins, and seabirds, ultimately depend on single-celled marine plants (phytoplankton) for their food. More than 500 species of protist have been identified in Antarctic waters, ∼350 of which are phytoplankton and ∼150 microheterotrophs (Scott and Marchant, 2005, http://taxonomic.aad.gov.au). These organisms coexist with untold numbers of heterotrophic prokaryotes (bacteria and Archaea) and viruses. Together they comprise the microbial food web (**Figure 1**), through which much of the carbon sequestered by phytoplankton is consumed, respired, and/or remineralized (Azam et al., 1983,

#### Edited by:

Julie Dinasquet, University of California, San Diego, USA

#### Reviewed by:

Ian Salter, Alfred Wegener Institute for Polar and Marine Research, Germany Bernard Quéguiner, Aix-Marseille University, France

> \*Correspondence: Stacy L. Deppeler stacy.deppeler@utas.edu.au

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Marine Science

Received: 30 September 2016 Accepted: 02 February 2017 Published: 16 February 2017

#### Citation:

Deppeler SL and Davidson AT (2017) Southern Ocean Phytoplankton in a Changing Climate. Front. Mar. Sci. 4:40. doi: 10.3389/fmars.2017.00040

1991; Fenchel, 2008; Kirchman, 2008). This food web includes the microbial loop in which dissolved carbon substrates fuel the growth of bacteria and Archaea, which are subsequently consumed by protists, returning carbon to the microbial food web that is otherwise lost to the dissolved pool (Azam et al., 1983). Phytoplankton are the base of the Southern Ocean (SO) food web. In nutrient rich Antarctic coastal waters their blooms can reach concentrations approaching 10<sup>8</sup> cells l−<sup>1</sup> . Chlorophyll a (Chl a) concentrations as high as 50 µg l−<sup>1</sup> have been recorded off the West Antarctic Peninsula (WAP), although maximum Chl a concentrations off East Antarctica are usually an order of magnitude less (Nelson et al., 1987; Smith and Gordon, 1997; Wright and van den Enden, 2000; Garibotti et al., 2003; Wright et al., 2010; Goldman et al., 2015). The majority of phytoplankton production in the SO is grazed by microheterotrophs or consumed and remineralized by bacteria (Lochte et al., 1997; Christaki et al., 2014). Production that escapes these fates sinks to depth, often in the form of dead cells, aggregates of biogenic material (marine snow), or fecal pellets, sequestering carbon in the deep ocean.

Some phytoplankton, such as prymnesiophytes and dinoflagellates, also synthesize substantial quantities of dimethylsulfoniopropiothetin (DMSP), which when enzymatically cleaved, forms dimethylsulfide (DMS). Oxidation of DMS in the atmosphere forms sulfate aerosols, which nucleate cloud formation and increase the reflectance of solar radiation (Charlson et al., 1987). The microbial food web plays a vital role in metabolizing these sulfur compounds (Kiene et al., 2000; Simó, 2004). The active involvement of phytoplankton in the sequestration and synthesis of climate-active gases (CO2) and biogenic sulfur compounds (DMSP and DMS), plus the mediation of the fate of these compounds by protozoa and bacteria means that microbes are a crucial determinant of future global climate (**Figure 1**).

The SO plays a substantial role in mediating global climate. The world's oceans have taken up between 25 and 30% of the anthropogenic CO<sup>2</sup> released to the atmosphere, with ∼40% of this uptake occurring in the SO (Raven and Falkowski, 1999; Sabine et al., 2004; Khatiwala et al., 2009; Takahashi et al., 2009; Frölicher et al., 2015). Without this, the atmospheric CO<sup>2</sup> concentration would be ∼50% higher than it is today. Drawdown of CO<sup>2</sup> by phytoplankton photosynthesis and vertical transport of this biologically sequestered carbon to the deep ocean (the biological pump) is responsible for around 10% this uptake (Cox et al., 2000; Siegel et al., 2014). Any climate-induced change in the structure or function of phytoplankton communities is likely to alter the efficiency of the biological pump, with feedbacks to the rate of

**Abbreviations:** SO, Southern Ocean; SAZ, sub-Antarctic zone; POOZ, permanently open ocean zone; SSIZ, seasonal sea ice zone; MIZ, marginal ice zone; CZ, Antarctic continental shelf zone; DMSP, dimethylsulfoniopropiothetin; DMS, dimethylsulfide; Chl a, Chlorophyll a; HNLC, high nutrient, low chlorophyll; UCDW, upper circumpolar deep water; SAM, Southern Annular Mode; WAP, west Antarctic peninsula; ASL, Amundsen Sea Low; ENSO, El Niño-Southern Oscillation; SIE, sea ice extent; CCM, carbon concentrating mechanism; PAR, photosynthetically active radiation; UV, ultraviolet.

climate change (Matear and Hirst, 1999; Le Quéré et al., 2007).

The SO is a region of seasonal extremes in productivity that reflect the large fluctuations in the SO environment. In summer, the development of large blooms of phytoplankton support a profusion of Antarctic life. Their metabolic activity also affects biogeochemical cycles in the SO, which in turn can influence the global climate. Whilst their effect on global climate is substantial, their microscopic size means they are intimately exposed to changes in their environment and are also likely to be affected by climate change. Already, climate change is causing the southward migration of ocean fronts, increasing sea surface temperatures, and changes in sea ice cover (Constable et al., 2014). Further changes in temperature, salinity, wind strength, mixed layer depth, sea ice thickness, duration and extent, and glacial ice melt are predicted. These changes are likely to affect the composition, abundance, and productivity of phytoplankton in the SO and feed back to threaten the ecosystem services they provide, namely sustaining biodiversity, fueling the food web and fisheries, and mediating global climate (Moline et al., 2004).

The SO is a vast and diverse environment, and hence the effect of climate change on the phytoplankton community is likely to be complex. For the purposes of this review we define the SO as waters south of the Sub-Tropical Front, thereby comprising ∼20% of the world's ocean surface area. We subdivide these waters into five regions that group waters according to the environmental drivers of the phytoplankton community in a similar manner as Tréguer and Jacques (1992) and Sullivan et al. (1988), namely the Sub-Antarctic Zone (SAZ), Permanently Open Ocean Zone (POOZ), Seasonal Sea Ice Zone (SSIZ), Marginal Ice Zone (MIZ), and the Antarctic Continental Shelf Zone (CZ) (**Figure 2**). Differences in environmental factors (physical, chemical, and biological) and processes (e.g., stratification, mixing, grazing) define the composition, abundance, and productivity of the phytoplankton community, both within and between these regions. Climate change is expected to elicit widespread changes in oceanography in each region, such as the displacement of oceanographic fronts (Sokolov and Rintoul, 2009b), as well as different permutations of climate-induced stressors that may interact synergistically or antagonistically, with either beneficial or detrimental effects on the phytoplankton community (Boyd and Brown, 2015; Boyd et al., 2016).

Here we identify the factors and processes that critically affect phytoplankton communities in each region of the SO, consider the impacts of climate change on each of these regions, examine the likely effect of these changes on the phytoplankton inhabiting these waters, and predict the possible repercussions for the Antarctic ecosystem.

#### 2. SUB-ANTARCTIC ZONE

The Sub-Antarctic Zone (SAZ) comprises more than half the total area of the SO and incorporates three important frontal regions; the Sub-Tropical Front, the Sub-Antarctic Front, and the Polar Front (**Figure 2**) (Orsi et al., 1995). Within this region, the waters between the Sub-Antarctic Front and the Polar Front are also referred to as the Polar Frontal Zone (e.g., Tréguer and Jacques, 1992). This region forms an important transitional boundary within the SO between the dominance of coccolithophores that construct carbonate shells to the north and diatoms with silicate frustules to the south (**Figures 2**, **3**) (Trull et al., 2001a,b; Honjo, 2004). Macro- and micronutrients are more abundant at the Polar Frontal Zone where nutrients are entrained across the bottom of the mixed layer, supporting deep chlorophyll maxima at depths up to 90 m. These deep chlorophyll maxima support blooms of large diatoms, such as Rhizosolenia sp. and Thalassiothrix sp., which can grow to high abundance and contribute significantly to carbon and silica flux (Tréguer and Van Bennekom, 1991; Kopczynska et al., 2001; Kemp et al., ´ 2006; Assmy et al., 2013). For the purpose of this review we are combining all waters between the Sub-Tropical Front to the north and the Polar Front to the south as the SAZ, as the physical and biological characteristics of these regions are similar.

This region of the SO is a major contributor to the uptake of CO<sup>2</sup> by the ocean (Metzl et al., 1999; Sabine et al., 2004; Frölicher et al., 2015). The westerly winds that circulate Antarctica carry water from the Antarctic Slope Front north across the SAZ by Ekman transport (**Figure 3**). This water has a partial pressure of carbon dioxide (pCO2) below that of the atmosphere, allowing CO<sup>2</sup> to dissolve into the ocean (the solubility pump). North of the Sub-Antarctic Front, surface water is convected to hundreds of meters, forming Antarctic Intermediate Water and Sub-Antarctic Mode Water (**Figure 3**) (Wong et al., 1999; Matear et al., 2000; Rintoul and Trull, 2001; Lumpkin and Speer, 2007). In doing so, it carries an estimated ∼1 Gt C yr−<sup>1</sup> to the ocean's interior and connects the upper and lower components of the global overturning circulation (Metzl et al., 1999; Sloyan and Rintoul, 2001a,b).

The SAZ is the largest high nutrient, low chlorophyll (HNLC) province in the world's ocean. Over the year phytoplankton productivity in this region is limited by a variety of bottomup (silicic acid, iron, and light) and top-down (grazing) factors (**Figure 4A**) (e.g., Banse, 1996; Boyd et al., 2001; Hiscock et al., 2003; Doblin et al., 2011). Iron is the main factor limiting phytoplankton growth in the SAZ, despite inputs from dust, shelf sediments, and hydrothermal vents (Boyd et al., 2004; Blain et al., 2007; Cassar et al., 2007; Pollard et al., 2009; Boyd and Ellwood, 2010; Tagliabue et al., 2010). Silica is replete in these waters in spring but it is drawn down by silicifying plankton, such as diatoms, silicoflagellates, and radiolarians, to limiting concentrations by autumn (Trull et al., 2001a; Salter et al., 2007; Pollard et al., 2009). In iron-limited regions of the SAZ, Si:C ratios are high, resulting in low carbon export (Salter et al., 2007, 2012; Assmy et al., 2013). In addition, light levels experienced by phytoplankton can be very low due to cloudiness and mixed layer depths ranging from 70 to 100 m in summer to as deep as 600 m in winter (Bishop and Rossow, 1991; Rintoul and Trull, 2001). In regions of shallow or complex bathymetry, such as sea mounts, or in waters downstream of sub-Antarctic islands, resuspension of iron-rich sediments naturally fertilizes the SAZ waters creating areas of high productivity (Salter et al., 2007; Pollard et al., 2009). Large, heavily-silicified diatoms, such as Eucampia antarctica and Fragilariopsis kerguelensis, are responsible for high levels of export in these naturally fertilized regions (Salter et al., 2007, 2012; Assmy et al., 2013; Rembauville

FIGURE 2 | Summer near-surface Chlorophyll a concentration, frontal locations and sea ice extent in the Southern Ocean. Chlorophyll a is determined from MODerate-resolution Imaging Spectroradiometer, Aqua satellite estimates from austral summer season between 2002/03 and 2015/16 at 9 km resolution. Black lines represent, frontal positions from Orsi et al. (1995). The red line denotes the maximum extent of sea ice averaged over the 1979/80 to 2007/08 winter seasons, derived from Scanning Multichannel Microwave Radiometer and Special Sensor Microwave/Image satellite data. Light blue lines depict the 1000 m depth isobath, derived using the General Bathymetric Chart of the Oceans, version 20150318. STF, Sub-Tropical Front; SAF, Sub-Antarctic Front; PF Polar Front; SACCF, Southern Antarctic Circumpolar Current Front.

et al., 2016b,c). This export is aided by silica limitation, the exhaustion of which ceases diatom growth and accelerates rates of sinking. Nutrient limitation also causes a succession in the phytoplankton community to picoeukaryotes, such as Phaeocystis sp. and coccolithophorids (Salter et al., 2007; Quéguiner, 2013; Balch et al., 2016).

Small taxa, including nanoflagellates, cyanobacteria, dinoflagellates, coccolithophores, and small or lightly silicified diatoms, dominate the protistan community in the SAZ (Odate and Fukuchi, 1995; Kopczynska et al., 2001, 2007; de Salas ´ et al., 2011). Copepods and mesopelagic fish, particularly myctophids, are important primary and secondary consumers of the phytoplankton in these waters and form an alternative food web for squid, predatory mesopelagic fish, and penguins (Kozlov, 1995; Cherel et al., 2010; Murphy et al., 2016). Measured rates of microzooplankton grazing (Jones et al., 1998; Griffiths et al., 1999; Safi et al., 2007; Pearce et al., 2011), together with high grazer biomass (Kopczynska et al., 2001 ´ ) suggest that grazers consume much of the primary productivity in this region. As a result of the physical and biological factors limiting primary

productivity in the SAZ, phytoplankton abundance is moderately low and varies little among seasons (Banse, 1996). The SAZ is more productive in the Atlantic sector and around 170◦W where iron concentrations are higher due to the proximity of land (**Figure 2**) (Comiso et al., 1993; de Baar et al., 1995; Moore and Abbott, 2000). Despite the low levels of primary productivity, export efficiency is high in HNLC waters of the SAZ, suggesting that small taxa contribute to a high proportion of carbon export (Trull et al., 2001b; Lam and Bishop, 2007; Cassar et al., 2015; Laurenceau-Cornec et al., 2015).

Climate predictions suggest that waters of the SAZ will become warmer, fresher and more acidic; the frequency of storms will increase, bringing more wind-blown dust to the region; and phytoplankton will experience increased irradiances of photosynthetically active radiation (PAR) and ultraviolet (UV) radiation (**Figure 4B**) (Matear and Hirst, 1999; Caldeira and Wickett, 2003; Orr et al., 2005; Marinov et al., 2010; Boyd and Law, 2011; Boyd et al., 2016). Together, these changes may have profound consequences for phytoplankton in the SAZ and the role of this region in mediating global climate.

Models suggest that global warming is likely to reduce the efficiency of both the solubility and biological pumps (Sarmiento and Le Quéré, 1996; Matear and Hirst, 1999). For phytoplankton, increased precipitation and warming increases the buoyancy of surface waters, enhancing stratification and reducing mixed layer depths over much of the SAZ. This reduces the delivery of nutrients to surface water, thereby reducing phytoplankton production and the vertical flux of biogenic carbon to the deep ocean via the biological pump (Matear and Hirst, 1999 and references therein; Boyd and Law, 2011; Petrou et al., 2016). The declining efficiency of the biological pump means it would be unable to compensate for any decline in the solubility of CO<sup>2</sup> as the ocean warms (Matear and Hirst, 1999). Recent studies also indicate that rising temperatures cause rates of grazing to increase more rapidly than rates of phytoplankton growth (Sarmento et al., 2010; Evans et al., 2011; Caron and Hutchins, 2013; Behrenfeld, 2014; Biermann et al., 2015; Cael and Follows, 2016). Thus, phytoplankton standing stocks are likely to decline and the proportion of primary production respired in near-surface waters by prokaryotes and grazers will increase. The nutritional quality of phytoplankton may also decline at higher temperatures (Finkel et al., 2010 and references therein; Hixson and Arts, 2016), suggesting grazers will also need to consume more phytoplankton to obtain the nutrition they require. Together, these factors are predicted to reduce phytoplankton productivity and the uptake of CO<sup>2</sup> by the ocean in the SAZ region.

The absence of iron is regarded as the primary cause of HNLC waters of the SO having the world's highest inventory of unused surface macronutrients (Martin et al., 1990; Boyd et al., 2007). As the largest HNLC region in the ocean, low rates of iron supply to the SAZ restrict primary production, alter phytoplankton species composition, increase Si:C export ratios, and constrain the biological pump (Ridgwell, 2002; Salter et al., 2012; Assmy et al., 2013; Salter et al., 2014). Aeolian dust makes a significant contribution to iron supply in the SAZ in areas downwind of landmasses and any increase in storm activity as a result of climate change may enhance delivery of iron-rich dust to these areas, enhancing productivity and carbon drawdown in this region (Cassar et al., 2007; Boyd and Law, 2011; Boyd et al., 2012, 2016). Investigations into sediment cores taken in the sub-Antarctic South Atlantic have correlated increased aeolian iron supply to the SAZ with increased productivity during ice ages, strengthening the biological pump and causing significant declines in atmospheric CO<sup>2</sup> (Anderson et al., 2014; Martínez-García et al., 2014). Increased desertification through climate change-related vegetation loss may result in a 10-fold increase in dust over the Southern Hemisphere (Woodward et al., 2005). However, the increase in dust will depend on both climate change and anthropogenic changes in land-use and re-vegetation, the net effects of which are currently uncertain (Ridgwell, 2002; Hutchins and Boyd, 2016).

While oceanic uptake of CO<sup>2</sup> ameliorates the accumulation of this gas in the atmosphere, it also alters the carbonate chemistry of the ocean. Absorption of CO<sup>2</sup> by the ocean reduces its pH (termed ocean acidification) and increases the solubility of calcium carbonate by reducing its saturation state () (Caldeira and Wickett, 2003; Orr et al., 2005). Coccolithophorids are the only calcifying phytoplankton in the SO and are most abundant in naturally iron-fertilized regions in the SAZ, such as fronts and downstream of sub-Antarctic islands (Salter et al., 2014; Balch et al., 2016). Calcification releases CO<sup>2</sup> (the carbonate counterpump), resulting in the elevation of pCO<sup>2</sup> concentrations in SAZ waters where coccolithophores are highly abundant, particularly around the Sub-Antarctic Front (Patil et al., 2014; Saavedra-Pellitero et al., 2014; Balch et al., 2016). Studies of the dominant coccolithophore, Emiliania huxleyi, found morphological variations in calcification that closely followed the southerly decline in calcite saturation state but were strainspecific rather than caused by acidification (Cubillos et al., 2007; Patil et al., 2014; Saavedra-Pellitero et al., 2014; Malinverno et al., 2015). However, culture studies by Müller et al. (2015) reported that calcification by E. huxleyi decreased at pCO<sup>2</sup> concentrations >1000 µatm. This suggests that calcifying phytoplankton in the SAZ will be vulnerable to predicted increases in pCO2. A decrease in calcification is anticipated to have a greater negative impact on the carbonate counter-pump than cell growth, leading to greater surface water pCO<sup>2</sup> uptake but potentially reducing vertical carbon flux through a decline in the ballasting effect of calcification (Riebesell et al., 2009; Müller et al., 2015; Balch et al., 2016).

Minimal research has been performed on the effect of ocean acidification on non-calcifying phytoplankton in the SAZ. Boyd et al. (2016) included ocean acidification in their multi-stressor study on a sub-Antarctic diatom and whilst their experimental design did not allow for full analysis of each individual stressor, they found that ocean acidification was not likely to be a primary controller in diatom physiology. Studies on other sub-Antarctic diatom species have reported an increase in productivity with increased CO<sup>2</sup> concentration, likely due to reduced energetic costs associated with the down-regulation of carbon concentrating mechanisms (CCMs) (Hopkinson et al., 2011; Trimborn et al., 2013). Most SO phytoplankton use CCMs to increase the intracellular concentration of CO<sup>2</sup> for fixation by RubisCO (Hopkinson et al., 2011). This process requires

substantial energy consumption and the down-regulation of CCMs is thought to decrease the energy cost of carbon acquisition for phytoplankton photosynthesis (e.g., Raven, 1991; Rost et al., 2008; Hopkinson et al., 2011). However, iron and light limitation in these waters is likely to inhibit any positive effects of increased CO<sup>2</sup> supply (Hoppe et al., 2013, 2015).

Stratification of the water column is predicted to increase in the SAZ region, trapping phytoplankton in a shallowing mixed layer where they are exposed to higher irradiances of PAR and UV radiation (280–400 nm) (Davidson, 2006; Gao et al., 2012; Häder et al., 2015). Light wavelengths are differentially attenuated by sea water. Blue wavelengths (∼500 nm) can reach depths exceeding 250 m in clear oceanic water but the penetration rapidly decreases as radiation tends toward infrared (longer) and ultraviolet (shorter) wavelengths (**Figure 5**) (Davidson, 2006). Thus, red and infrared wavelengths only warm the very surface of the ocean, while damaging irradiances of UV-B penetrate to ≤30 m depth (Karentz and Lutze, 1990; Buma et al., 2001; Davidson, 2006). Rates of phytoplankton productivity in the SAZ are commonly limited by light availability due to cloudiness and deep mixing. Increased stratification could mitigate this limitation by keeping cells in sunlit near-surface waters. Overall, productivity would still be constrained by the availability of key nutrients (iron and silicate), which already limit phytoplankton production in the SAZ despite the low light. Thus, increased rates of productivity are unlikely to result in higher biomass or carbon export in this region without a coincident increase in nutrient supply (see above).

Exposing phytoplankton in the SAZ to higher irradiances of PAR, Ultraviolet-A (UV-A, 315–400 nm), and Ultraviolet-B (UV-B, 280–315 nm) is also likely to increase photodamage. The damage to intracellular molecules or structures become progressively less repairable as wavelengths decline below 350 nm, reducing phytoplankton productivity, growth and survival, and changing the species composition, with implications for ecosystem structure and function (e.g., Karentz, 1991; Marchant and Davidson, 1991; Davidson, 2006). The amount of damage sustained by cells is a function of the dose and dose rate of UV exposure; the frequency and duration of exposure to low irradiances to allow repair; and species-specific differences in the UV-tolerance of component species in natural phytoplankton communities (e.g., Cullen and Lesser, 1991; Davidson, 2006; Häder et al., 2015). It is hard to assess the additional risk UV exposure may have to phytoplankton in the SAZ as such details are currently unavailable. Studies by Helbling et al. (1994) and Neale et al. (1998a,b) showed that increasing the rate of change in the light climate altered the balance between damage and repair and greatly increased the biological impact of a specific UV dose. Thus, trapping cells in a shallow mixed zone where they receive repeated exposure to high PAR and UV irradiances over short time scales (see above, **Figure 5**) may have a far greater impact on the growth, production, and survival of phytoplankton than ozone depletion (Davidson, 2006).

The SAZ region is being increasingly penetrated by both sub-tropical and polar waters. The climate-induced increase in the positive phase of the Southern Annular Mode (SAM) has caused the westerly wind belt to intensify and move south

Ozone of 300 Dobson units. Photon energy was calculated after Kirk (1994). SAZ, Sub-Antarctic Zone; POOZ, Permanently Open Ocean Zone; SSIZ, Seasonal Sea Ice Zone; PAR, photosynthetically active radiation.

(see POOZ below). This increase in the velocity of westerly winds to the south of the SAZ has enhanced upwelling at the Antarctic Slope Front and increased its Ekman transport into the SAZ from the south, increasing phytoplankton growth in the cool, nutrient-rich water (Lovenduski and Gruber, 2005; DiFiore et al., 2006). A 37 year dataset of surface Chl a measurements south of Australia from vessels of the Japanese Antarctic Research Expeditions show as similar trend of increasing Chl a spreading northward from these northern limits of the POOZ (55◦ S) into the Polar Frontal Zone (40◦ S) (Hirawake et al., 2005). The southward movement of the westerly wind belt has also increased the penetration of sub-tropical waters into the SAZ; supplementing iron supply, exacerbating warming, and intensifying climate-induced stratification (Lovenduski and Gruber, 2005; Poloczanska et al., 2007; Ridgway, 2007). Warmer waters also allow the incursion of sub-tropical phytoplankton and grazers into SAZ waters, causing additional grazing competition and unknown effects on the SO food web (McLeod et al., 2012).

Not all of the SAZ is expected to experience shallowing mixed layer depth as a result of climate change. At the sub-Antarctic convergence, increased wind will deepen the mixed layer, causing declines in phytoplankton productivity through light limitation (Lovenduski and Gruber, 2005). In addition, there are zonal differences in the effect of the increasingly positive SAM on mixed layer depth in the SAZ region, with deepening over the eastern Indian Ocean and central Pacific Ocean, and shallowing over the western Pacific Ocean (Sallée et al., 2010). Resulting in a mosaic of changing factors that limit phytoplankton productivity, from nutrient limitation in shallower regions to light limitation in deeply mixed waters.

Clearly, phytoplankton occupying the SAZ region are likely to experience a range of environmental stressors as a result of climate change. The net effect of these changes is uncertain. Most studies investigate the physiological effects of change on phytoplankton by imposing single stressors (e.g., Boyd et al., 2013; Trimborn et al., 2013) but research shows interaction among stressors alter their response. A multi-stressor study by Boyd et al. (2016) using a sub-Antarctic diatom showed that its response to environmental change was governed by the range of stresses to which it was exposed. Negative responses to several stressors (CO2, nutrients, and light) were offset by positive responses to others (temperature and iron). Thus, the response of an organism is determined by the interactive effect of all the stresses they experience (Boyd et al., 2016). Equally, responses of single species (e.g., Boelen et al., 2011; Trimborn et al., 2014; Müller et al., 2015) provide valuable insights into the mechanisms of sensitivity and tolerance but avoid interactions among species and trophic levels that can alter the responses or sensitivity of a community to a stressor (Davidson et al., 2016; Thomson et al., 2016). Yet gaining maximum predictive strength by simultaneously performing multi-stressor and multitrophic level studies is often logistically so demanding as to be impractical.

Predicted responses by phytoplankton in the SAZ to climate change differ. Many propose that the stratification-induced decline in nutrient supply to surface waters will reduce their productivity and favor small flagellates (e.g., Matear and Hirst, 1999; Marinov et al., 2010; Petrou et al., 2016), heightening the role of the microbial food web and reducing carbon export. While Boyd et al. (2016) indicates that increases in iron and temperature may double growth rates and favor diatoms; scenarios which have major and opposing influences on regional productivity and biogeochemistry. It is likely that the effect of climate change on phytoplankton in the SAZ is going to be determined by the timing, rate, and magnitude of change in each stressor. Stochastic inputs of iron, wind, and storms disrupt stratification; influencing productivity, species composition, and export production through changes in nutrients and light climate. Changes in community composition from diatoms to flagellates also affect particulate matter stoichiometry in this region, causing a decline in nutritional quality for grazing zooplankton (Martiny et al., 2013; Rembauville et al., 2016a) and subsequent flow on effects throughout the food web (Finkel et al., 2010). Ocean acidification will also cause declines in carbonate saturation, affecting coccolithophore calcification, resulting in greater surface pCO<sup>2</sup> uptake and decreased carbon export. Overall, our synthesis suggests that productivity will decline in the SAZ due to the net response of nutrient limitation and increased grazing, especially in silicate-limited waters.

# 3. PERMANENTLY OPEN OCEAN ZONE

The Permanently Open Ocean Zone (POOZ) lies between the Polar Front and the northern limit of the winter sea ice, covering approximately 14 million km<sup>2</sup> (**Figure 2**). The Polar Front at the northern extent of the POOZ forms a natural barrier between the warm SAZ water (5–10◦C) and the cold Antarctic water (<2◦C) (Pollard et al., 2002; Sokolov and Rintoul, 2009a). These waters are predominantly HNLC with a phytoplankton community dominated by nano- and picoflagellates but characteristically contain even less Chl a than the SAZ (Becquevort et al., 2000; Moore and Abbott, 2000; Kopczynska et al., 2001; Olguín and ´ Alder, 2011). The exception to this is where iron concentrations in surface waters are enhanced by upwelling and/or sediment input/resuspension from sea floor bathymetry and sub-Antarctic islands (**Figure 2**) (e.g., Pollard et al., 2002; Ardelan et al., 2010; Rembauville et al., 2015b). This pattern differs from that of macronutrients, which decline northwards across the POOZ region, nitrate falling from ∼25–20 µmol l−<sup>1</sup> and silicate from ∼60–10 µmol l−<sup>1</sup> . These nutrients are upwelled at the Antarctic Slope Front and are progressively drawn down by phytoplankton as they are transported northward across the POOZ by Ekman drift (Tréguer and Jacques, 1992; Pollard et al., 2002).

The POOZ displays a strong seasonality in biological production (Abbott et al., 2000). Strong winds in winter deepen the mixed layer, bringing nutrient-rich water to the surface. These nutrients fuel phytoplankton growth in spring when sunlight increases, conditions are calmer, and phytoplankton are confined to shallower mixed depths by stratification (**Figure 6A**) (Abbott et al., 2000; Pollard et al., 2002; Constable et al., 2014). Whilst the POOZ is considered to be an iron-limited environment, silicate limitation and grazing by micro- and metazooplankton also limit the duration of the diatom-dominated bloom in this region (Abbott et al., 2000; Becquevort et al., 2000; Timmermans et al., 2001; Strzepek et al., 2011; Christaki et al., 2014). Like the SAZ, large, heavily silicified diatoms contribute significantly to carbon export (Rembauville et al., 2015a,b, 2016b; Rigual-Hernández et al., 2015). In regions of natural iron fertilization (e.g., the Kerguelen Plateau), phytoplankton production appears to be strongly linked to higher trophic levels rather than making a substantial contribution to carbon export (Obernosterer et al., 2008; Christaki et al., 2014; Laurenceau-Cornec et al., 2015; Rembauville et al., 2015b).

Modeling studies predict the POOZ region will experience a poleward shift and strengthening of the westerly winds; deepening of the summertime mixed layer depth; increasing cloud cover; warming and freshening of surface waters; and decreasing pH (**Figure 6B**) (Orr et al., 2005; McNeil and Matear,

2008; Meijers, 2014; Leung et al., 2015; Armour et al., 2016; Haumann et al., 2016). Thus far, sea surface warming in the POOZ of only 0.02◦C per decade has been slower than the global average of 0.08◦C per decade, since 1950 (Armour et al., 2016). This is due to heat taken up by surface water in the POOZ being transported northward by Ekman drift into the SAZ (**Figure 3**). Despite this, it has been proposed that rising temperatures may be contributing to an observed range extension of E. huxleyi below 60◦ S (Cubillos et al., 2007; Winter et al., 2014).

Whilst warming is expected to increase phytoplankton productivity (Sarmiento et al., 2004; Behrenfeld et al., 2006; Steinacher et al., 2010), this effect is offset against the increasingly positive phase of SAM, which is causing an intensification and southerly shift of westerly winds in summer (Lenton and Matear, 2007; Lovenduski et al., 2007). The SAM controls the north-south shift of the circumpolar westerly winds and is the dominant climate-induced environmental change in Antarctic waters, substantially affecting SO circulation and CO<sup>2</sup> uptake (Thompson and Solomon, 2002; Lenton and Matear, 2007; Lovenduski et al., 2007; Swart et al., 2014). In the last 50 years there has been an observed increase in the positive phase of SAM, strongly related to the depletion of ozone in the atmosphere above Antarctica (Son et al., 2008; Polvani et al., 2011). Leung et al. (2015) predict that the positive SAM will continue to deepen the summer mixed layer and increase cloud cover in the POOZ, resulting in decreasing light availability and causing a decline in phytoplankton biomass and productivity. Observed trends in summertime mixed layer depth, cloud cover, and Chl a (since 1950, 1980, and 1997, respectively) correspond well to the modeled projections (Leung et al., 2015).

Conversely, some predict the increase in positive SAM may enhance phytoplankton productivity in the POOZ. Deepening of the mixed layer can increase the upwelling of nutrients, which some models predict will promote phytoplankton productivity and export production south of 60◦ S (Lovenduski and Gruber, 2005; Hauck et al., 2013, 2015; Laufkötter et al., 2015). It is hard to assess the validity of such predictions for the POOZ region as these models combine all waters south of the Polar Front, including the SSIZ. Using satellite and Argo data, Carranza and Gille (2015) reported a correlation of increased Chl a in the SO with increased mixed layer depth. A positive SAM also increases eddy formation and transports SAZ water across the Polar Front (Meredith and Hogg, 2006; Kahru et al., 2007; Hogg et al., 2008). These cyclonic eddies trap warm water at their core, enhance stratification, and upwell nutrients and iron, creating ideal conditions for phytoplankton productivity (Kahru et al., 2007) and may also contribute significantly to ocean warming in the POOZ (Hogg et al., 2008).

Increased nutrient input from melting icebergs may also increase productivity in the POOZ. Climate warming and the breakup of Antarctic ice shelves (Scambos et al., 2000) could increase the number of icebergs in the POOZ (see CZ below). Melting icebergs enrich the surrounding water with iron, enhancing phytoplankton growth and productivity (Cefarelli et al., 2011; Lin et al., 2011; Shaw et al., 2011; Vernet et al., 2011, 2012), and increasing export of carbon from surface waters (Smith et al., 2011). This heightened productivity also attracts large grazing populations that increase food availability to higher trophic levels and facilitates the sequestration of carbon to the deep ocean through fecal pellet production (Vernet et al., 2011).

Climate change is expected to change the location and area of the POOZ. The Polar Front, which denotes the northern limit of the POOZ has already shifted 60 km south since 1992 and this southward migration is expected to continue as the climate warms (Sokolov and Rintoul, 2009b). To the south, the northernmost extent of sea ice coverage is also predicted to retreat with ocean warming. Overall, this would result in a net increase in the area of the POOZ in the future (Bracegirdle et al., 2008; McNeil and Matear, 2008; Boyd et al., 2014). Some studies suggest that an increase in open ocean habitat will increase production in this region (Bopp et al., 2001; Behrenfeld et al., 2006). However, it is not yet understood how the multi-stressor effects of the accompanying environmental changes, such as ocean warming, decreased pH, light availability, and nutrient supply will affect the phytoplankton community.

The effect of climate change on phytoplankton productivity in the POOZ will strongly depend on the changes in light limitation and nutrient supply. Deepening of the summertime mixed layer depth due to increases in the strength of westerly winds are likely to further reduce the light available to phytoplankton, reducing their productivity over much of the POOZ (see above). However, increased nutrient concentrations as a result of increased mixing and melting icebergs, together with the incursions of warmcore eddies from the Polar Front may promote localized phytoplankton blooms when light is not limiting. Furthermore, increased nutrient concentrations might promote the growth of large diatoms (Timmermans et al., 2001), as well as increased abundance of phytoplankton in near surface waters rather than forming deep chlorophyll maxima. This increase in abundance is likely to increase the functioning of the microbial loop and promote grazing, as has been observed in naturally iron-fertilized regions of the POOZ (Christaki et al., 2014). It is also likely that with a future southward shift in SSIZ extent (see SSIZ below) the brief but substantial blooms of Phaeocystis sp. and large diatoms of the MIZ will be replaced by a prolonged but subdued bloom of phytoplankton over summer in waters that are now part of the POOZ (see MIZ below, Behrenfeld et al., 2006).

# 4. SEASONAL SEA ICE ZONE

In the following sections we divide the region of the SO covered by sea ice into two distinct zones. First we consider the effects of climate change on the extent, advance and retreat of ice over the entire Seasonal Sea Ice Zone (SSIZ) and examine the implications for phytoplankton. Then we consider the processes occurring at the northern margin of the sea ice (the marginal ice zone, MIZ), and how these are predicted to respond to a changing climate.

The SSIZ encompasses the region of the SO between the winter maximum and summer minimum of sea ice cover (**Figure 2**). The sea ice is one of the largest and most dynamic ecosystems on earth, extending to over 19 million km<sup>2</sup> in winter and retreating to ∼3 million km<sup>2</sup> over summer (Brierley and Thomas, 2002; Comiso and Nishio, 2008; Convey et al., 2009). Total productivity within the SSIZ has been estimated at ∼140–180 Tg C yr−<sup>1</sup> (Arrigo et al., 1997, 2008b). Sea ice cover plays an important role in the regulation of climate by controlling heat and gas exchange between the atmosphere and the ocean (Massom and Stammerjohn, 2010). Snow covered sea ice creates a high albedo surface that reflects most of the sun's energy back into space, thereby reducing warming of the polar oceans (Perovich, 1990). Conversely, in winter the ice cover insulates the ocean from direct exposure to the cold atmosphere (Stroeve et al., 2016 and references therein). Not only is sea ice itself an important regulator of global climate, it also provides a vital environment for Antarctic life.

Sea ice supports a diverse community of algae that possess some of the most extreme adaptations to environmental stress recorded. They inhabit a range of environments throughout the ice; from surface ponds to brine channels in the sea ice interior and at the bottom ice-water interface (Knox, 2007; Arrigo, 2014). Here they can experience extremely low temperatures (<-20◦C), light irradiances (<1 µmol m−<sup>2</sup> s −1 ), CO<sup>2</sup> concentrations (<100 µatm), and salinities up to ∼200 PSU (Thomas and Dieckmann, 2002 and references therein). Primary production by sea ice algae contributes between 24-70 Tg C yr−<sup>1</sup> (Legendre et al., 1992; Arrigo et al., 1997; Saenz and Arrigo, 2014) and phytoplankton biomass averages between 1 and 100 mg Chl a m−<sup>2</sup> , although it can exceed 1000 mg Chl a m−<sup>2</sup> in some regions (Lizotte, 2001; Arrigo et al., 2010). Ice algal biomass and productivity varies greatly at small spatial and temporal scales, primarily due to changes in snow cover, ice thickness, surface flooding, and ice rafting (McMinn et al., 2007; Meiners et al., 2012; Arrigo, 2014 and references within). Thus, ice algae are able to thrive in this harsh physical environment.

Ice algal productivity is essential to the nutrition of higher trophic levels in Antarctic waters. Productivity and algal biomass within the sea ice is generally low during the winter (Arrigo et al., 1998a). Conditions are most favorable at the ice-water interface, where warmer temperature (-1.8◦C), lower salinity (∼35 PSU), and high nutrients maintain higher productivity rates than the sea ice interior (Lizotte, 2001). These bottom ice algal communities are an essential food source for zooplankton over winter (Brierley and Thomas, 2002 and references therein, Jia et al., 2016), when phytoplankton biomass in the waters beneath the sea ice are very low due to light limitation (Perrin et al., 1987; Legendre et al., 1992; Robins et al., 1995). For example, the phenology of the Antarctic krill, Euphausia superba, a keystone organism in SO food webs, is integrally liked to sea ice and seasonality, largely due to its being a refuge and source of algal nutrition over winter (Kawaguchi and Satake, 1994; Daly, 1998; Atkinson et al., 2004; Smetacek and Nicol, 2005; Quetin and Ross, 2009) and is associated with the ice at all stages of its life cycle (Flores et al., 2012 and references therein). Thus, changes in the timing and/or extent of sea ice cover are likely to have major implications for the Antarctic food web (see below, Quetin and Ross, 2009).

Changes in the extent, duration, thickness, and transparency of sea ice will have major implications for the algae that inhabit the ice and processes that drive phytoplankton productivity during sea ice retreat. In stark contrast to the decline currently observed in the Arctic (Stroeve et al., 2012 and references therein), the overall sea ice extent (SIE) around Antarctica has experienced a modest increase of between 0.9 and 1.5% since 1979 (Comiso and Nishio, 2008; Turner et al., 2009; Parkinson and Cavalieri, 2012; Simmonds, 2015), and modeled increases in sea ice volume of ∼0.4% yr−<sup>1</sup> between 1992 and 2010 due to approximately equal increases in both SIE and thickness (Holland et al., 2014). This culminated in the National Snow and Ice Data Center (NSIDC) reporting a maximum recorded SIE >20 million km<sup>2</sup> in September 2014, 1.54 million km<sup>2</sup> above the 1981 to 2010 average (Fetterer et al., 2016a). However, the SIE around Antarctica in November 2016 was only 14.54 million km<sup>2</sup> , 1.81 million km<sup>2</sup> below the 1981 to 2010 average (Fetterer et al., 2016b), demonstrating substantial interannual variability. Furthermore, the long term trend in increasing SIE is not uniform around Antarctica, with a significant decrease in the Amundsen and Bellingshausen Seas of between −5.1 and −6.6% per decade but a large increase in the Ross Sea of between 4.2 and 5.2% per decade due to the Amundsen Sea Low (ASL) (see below, Comiso and Nishio, 2008; Massom and Stammerjohn, 2010; Parkinson and Cavalieri, 2012).

Dramatic changes in SIE in some regions around Antarctica have altered the timing of sea ice growth and retreat. The large changes in SIE between the Ross Sea and the WAP are driven by the combined influence of the El Niño-Southern Oscillation (ENSO), the SAM, and their interaction with the ASL, the deepest low pressure cell around Antarctica (Arrigo and Thomas, 2004; Liu et al., 2004; Massom et al., 2008; Stammerjohn et al., 2008; Pezza et al., 2012; Raphael et al., 2016). The positive SAM phase and the La Niña phase of the ENSO have deepened the ASL. Increasing greenhouse gasses and stratospheric ozone recovery may also exacerbate the current SIE trends in these regions by further deepening the ASL (Raphael et al., 2016). The resultant strengthening of winds associated with the ASL lead to the compression of the sea ice in the Amundsen and Bellingshausen Seas and expansion in the Ross Sea. As a result, sea ice extent around the West Antarctic Peninsula (WAP) has declined by up to 40% over the past 26 years (Smith and Stammerjohn, 2001; Ducklow et al., 2007; Parkinson and Cavalieri, 2012). Modeling studies predict that continued global warming will eventually override the SAM and ENSO effects, increasing warming to the atmosphere and ocean, and resulting in significant declines in SIE around Antarctica (Bracegirdle et al., 2008; Ferreira et al., 2015).

Changes in sea ice concentration, extent, and seasonality critically affect the timing and productivity of phytoplankton blooms. In the western Ross Sea, sea ice retreats later and advances earlier, reducing the ice-free season by ∼2.6 months (Stammerjohn et al., 2012). The delay in ice retreat has delayed the onset of the summer bloom and decreased its duration, thereby reducing total seasonal productivity (Arrigo and van Dijken, 2004). Conversely, earlier retreat and delayed advance of sea ice has resulted in a 3 month lengthening of the summer ice-free season in the Amundsen and Bellingshausen Seas (Stammerjohn et al., 2012). While this extension of the ice-free period was expected to increase annual phytoplankton production and growth (Sarmiento et al., 2004; Moreau et al., 2015), no such trend has yet been observed (Smith et al., 2008; Montes-Hugo et al., 2008). This may be due to constraints imposed by nutrient and light limitation that are also key drivers of phytoplankton growth in the SSIZ (Pearce et al., 2010; Westwood et al., 2010).

The observed increase in SIE is contrary to modeling studies that predict a decline in SIE with global warming (Maksym et al., 2012 and references therein), reflecting the complex interaction of factors influencing the distribution and concentration of sea ice around Antarctica (Sen Gupta et al., 2009; Parkinson and Cavalieri, 2012 and references therein, Turner et al., 2013). Models indicate that the continued warming of the Earth's climate will result in a 33% decline in Antarctic SIE by 2100 (Bracegirdle et al., 2008). Historical records (whaling records, ice charts, and direct observations) and concentrations of methane sulfonic acid in ice cores suggest SIE has declined at least 20% since the 1950s (Curran et al., 2003; de la Mare, 2009).

The seasonal southward retreat of the sea ice initiates the phytoplankton bloom (see MIZ below) and changes in the timing of sea ice growth and retreat will alter the timing of these blooms. Such changes can impose temporal asynchronies and spatial separations between grazers and their food, reducing grazer abundance, reproductive success, and altering the distributions of higher trophic levels (Moline et al., 2008). SO zooplankton use the sea ice as a refuge and food source in the winter (Daly, 1998; Murphy et al., 2007; Jia et al., 2016). MIZ phytoplankton blooms supply the essential fatty acids required for reproduction and over-wintering strategies (Schnack-Schiel et al., 1998; Hagen, 1999). It is not yet known how changes in sea ice retreat will affect higher trophic levels in SSIZ but a delay in the summer bloom may restrict the availability of an essential food source during vulnerable life-stages, resulting in significant grazer mortality and less food availability to higher trophic organisms.

A decline in SIE is likely to decrease overall ice algal abundance, reducing carbon flux to the deep ocean. Decaying sea ice releases plumes of ice algal aggregates that can sink from surface waters at rates ≤200 m d−<sup>1</sup> (Thomas et al., 1998; Wright and van den Enden, 2000; Wright et al., 2010). Given that sea ice algae contribute to ∼12% of annual productivity in the SSIZ (Saenz and Arrigo, 2014); the large accumulations of algal biomass amongst the sea ice (see above); and the fact that the rate of sedimentation would largely preclude remineralization of these algal aggregates; it is likely that declining ice algal abundance would reduce this region's contribution to vertical carbon flux.

A reduction in SIE extent, and therefore sea ice algal biomass, is also likely to reduce the contribution of Antarctic sea ice algae to the global biogenic sulfur budget via synthesis of DMSP and subsequent release of DMS. Many intracellular roles have been proposed for DMSP and DMS, including cryoprotectant, antioxidant, metabolic overflow product, and even a compound that mediates grazer interactions (Kirst et al., 1991; Malin, 2006 and references therein). DMS is oxidized in the atmosphere to sulfate aerosols which nucleate cloud condensation, altering global albedo (Charlson et al., 1987, 1992). Estimates suggest that the Antarctic region contributes 17% of the global DMS emissions (Curran and Jones, 2000), with the highest concentrations of these DMSP and DMS compounds often found amongst sea ice (e.g., Kirst et al., 1991; Turner et al., 1995; Trevena and Jones, 2006; Jones et al., 2010; Vance et al., 2013). Any climate-induced decline in SIE and/or duration (see above) could reduce the magnitude of DMS production in the SSIZ, feeding back to global climate by reducing cloud-induced albedo.

Thinning of sea ice could substantially contribute to the loss of sea ice volume within the SSIZ, impacting ice algal communities. Observations of ice thickness in the SSIZ are sparse and difficult to obtain, displaying large variability within regions and among seasons (Worby et al., 2008). As a result, current trends in Antarctic sea ice thickness are not well understood (Kwok, 2010; Hobbs et al., 2016 and references therein) and based upon model estimates (Holland et al., 2014). The majority of the sea ice in the SSIZ is first-year ice, with ice thickness seldom exceeding 2 m (Worby et al., 2008; Meiners et al., 2012). Ice algal biomass is often concentrated in the bottom 20 cm of the ice (Palmisano and Sullivan, 1983; McMinn et al., 2007; Meiners et al., 2012), with thicker ice (>1.0 m) supporting higher algal biomass than thin ice (<0.4 m), due to longer time for colonization and growth of the bottom ice algal community, along with development of internal communities from the rafting of ice floes (McMinn et al., 2007; Meiners et al., 2012). Thus, a decline in sea ice thickness may result in a reduction in bottom community biomass, which is an important food source for zooplankton (Brierley and Thomas, 2002; Jia et al., 2016), thereby causing a shift in the diet of Antarctic birds and mammals toward less efficient pathways (Murphy et al., 2007; Moline et al., 2008; Flores et al., 2012; Ballerini et al., 2014).

A warming atmosphere is predicted to result in more precipitation that could cause an increase in snow deposits on the surface of the sea ice (Bracegirdle et al., 2008; Massom et al., 2008). Increased snow load depresses ice floes, flooding the ice surface and fostering phytoplankton blooms in the high light, high nutrient environment at the snow-ice interface (Arrigo et al., 1997; Massom et al., 2006). Surface communities are most often associated with thin ice (<0.4 m) (Meiners et al., 2012) and as such, could become more prominent in the future. Increased albedo caused by greater snow cover on the ice would also limit light transmission through the ice, reducing ice algal productivity in internal and bottom communities (Grossi et al., 1987; Palmisano et al., 1987).

Sea ice is a substantial sink for CO<sup>2</sup> over winter. Air-ice exchange at the ice surface over-saturates the CO<sup>2</sup> in sea ice brine and contributes as much as 58% of the annual atmospheric CO<sup>2</sup> uptake in the SO (Delille et al., 2014). Ice cover provides a barrier between the atmosphere and the surface water, slowing atmospheric CO<sup>2</sup> uptake (Boyd et al., 2008) and limiting predicted pCO<sup>2</sup> levels by 2100 to 500–580 µatm. Furthermore, it prohibits outgassing of upwelled water supersaturated in CO<sup>2</sup> over winter (Gibson and Trull, 1999; Roden et al., 2013). The few studies investigating the effect of ocean acidification on sea ice algal communities suggest they can tolerate CO<sup>2</sup> concentrations up to 10,000 µatm (McMinn et al., 2014; Coad et al., 2016).

The increasingly positive SAM (see SAZ above) exposes the SSIZ is to stronger winds. However, future recovery of the ozone hole will reduce the SAM favoring increasing warming and stratification (see Conclusion), with consequent declines in the SIE extent, thickness and duration of ice cover. This is likely to have a strong negative effect on sea ice algal abundance, through a loss of habitat. Whilst ice algae are not major contributors to overall SO primary productivity, they are essential in the life cycles of many zooplankton species. Thus, declines in ice algal abundance will likely have a significant negative effect on critical links in the SO food web, especially krill, and promote different and less energy efficient trophic pathways such as consumption of phytoplankton by salps or via copepods to myctophids. Such changes would reduce the capacity of the SO to support the current abundance of iconic, krill-dependent Antarctic wildlife (Murphy et al., 2007, 2016).

The development of the phytoplankton bloom and succession of the pelagic phytoplankton community is initiated by the seasonal retreat of the sea ice across the SSIZ. Here we consider the effects of climate-induced changes on processes in the MIZ.

# 5. MARGINAL ICE ZONE

The region where the dense sea ice pack transitions to open ocean is known as the marginal ice zone (MIZ). It is an area of high productivity that accounts for the majority of the springsummer phytoplankton blooms (**Figure 2**) (Arrigo et al., 2008b). The area of the MIZ varies greatly over spring and summer, ranging from 6 million km<sup>2</sup> in December to ∼0.2 million km<sup>2</sup> by March (Fitch and Moore, 2007). Sea ice formation in the winter scavenges phytoplankton cells into the ice and concentrates iron from the surface water (de Baar et al., 1995; Boyd, 2002; Lannuzel et al., 2010, 2016). In the spring, low salinity, high iron melt water is released from the sea ice, creating a buoyant layer of fresher water that traps phytoplankton in an environment where conditions are ideal for growth (high light, and high macroand micronutrients). This fosters large phytoplankton blooms (**Figure 7A**) (Smith and Nelson, 1986; Sullivan et al., 1988), which can reach biomasses of over 200 mg Chl a m−<sup>2</sup> (e.g., Smith and Nelson, 1986; Nelson et al., 1987; Wright et al., 2010). The region was thought to house very high rates of productivity (∼400 Tg C yr−<sup>1</sup> ) (Smith et al., 1988; Arrigo et al., 1998b) and contribute 40-50% of the productivity of the entire SO (Smith and Nelson, 1986; Sakshaug, 1994). Advances in satellite technology and modeling algorithms provide more conservative results (Arrigo et al., 2008b; Taylor et al., 2013), suggesting the MIZ contributes ∼114 Tg C yr−<sup>1</sup> . This equates to a total annual productivity of 54–68 g C m−<sup>2</sup> yr−<sup>1</sup> , which is ∼5 times that in the sea ice (∼24 Tg C yr−<sup>1</sup> ) but is similar to that in the POOZ (∼62 g C m−<sup>2</sup> yr−<sup>1</sup> ) (Moore and Abbott, 2000; Arrigo et al., 2008b; Saenz and Arrigo, 2014).

A diverse array of phytoplankton inhabit the MIZ, undergoing successional change due to ice retreat, warming, nutrient depletion, and grazing (Davidson et al., 2010; Wright et al., 2010). Phytoplankton blooms in East Antarctica and the Weddell Sea, are commonly co-dominated by the colonial life-stage of Phaeocystis sp. and diatoms, with increasing diatom abundance

over time and the appearance of dinoflagellates, silicoflagellates, and heterotrophic protists later in the season (Waters et al., 2000; Kang et al., 2001; Davidson et al., 2010). Once the available iron has been exhausted, the community shifts to one more typical of the POOZ, consisting of small diatoms and flagellates (Pearce et al., 2010; Wright et al., 2010). In the WAP, diatom-dominated blooms in the spring shift to flagellate communities as melting sea ice and glacial run-off reduce the salinity of surface waters (Kang et al., 2001). However, icebergs released by the breakup of ice shelves will increase nutrient input, as in the POOZ (see above, Duprat et al., 2016), promoting additional blooms of large diatoms.

Phytoplankton in the MIZ can contribute directly or indirectly to vertical flux. During large blooms phytoplankton aggregate to form marine snow, which fall rapidly through the water column, contributing to carbon sequestration into the deep ocean (Alldredge and Silver, 1988). High algal biomass within decaying sea ice in summer is also a rich source of nutrition and a site of reproduction for grazers (Schnack-Schiel et al., 1998; Thomas et al., 1998). This grazing transfers carbon to higher trophic levels but can also contribute to vertical carbon flux by reparceling cells into rapidly sinking fecal pellets (Cadée et al., 1992; Burkill et al., 1995; Perissinotto and Pakhomov, 1998; Pearce et al., 2010).

Climate change is predicted to decrease SIE, increase icebergs, and cause SAM-induced increases in wind and wave action (**Figure 7B**). The effect of decreased SIE on total annual productivity in the SO may not be large. Reduced SIE would shift the latitudinal range of the MIZ southward, resulting in an increase in the area of the POOZ (Smetacek and Nicol, 2005). However, the restriction of intense primary productivity in the MIZ to the spring-summer season results in area-normalized annual primary production similar to that of the POOZ (see above, Moore and Abbott, 2000; Arrigo et al., 2008b), suggesting that an increase in the size of the POOZ may not significantly affect total SO productivity (Arrigo et al., 2008b). Admittedly, this does not take into account other potential effects of climate change on the POOZ (see above), nor does it consider the effect of the absence of ice on the timing and magnitude of the phytoplankton bloom. It is likely that blooms would start earlier due to the higher light climate but may develop slower due to greater mixed depths (see below) and the lack of iron fertilization from the ice melt (Behrenfeld et al., 2006).

The most profound change in the MIZ may be caused by the increasingly positive phase of SAM. The poleward shift and intensification of wind strength and storms is predicted to deepen the mixed layer and reduce phytoplankton production in the MIZ (**Figure 7B**) (Lovenduski and Gruber, 2005; Yin, 2005; Hemer et al., 2010; Massom and Stammerjohn, 2010; Young et al., 2011; Dobrynin et al., 2012). Phytoplankton blooms in the MIZ are patchy in space and time (Smith and Nelson, 1986). They generally occur in shallow mixed layers where wind speeds are <5 m s−<sup>1</sup> (Fitch and Moore, 2007). Storms, wind mixing, and waves deepen mixed depths in the MIZ, reducing the light availability and inhibiting bloom development (**Figure 7B**) (Venables and Meredith, 2014). As a result, blooms only cover 17–24% of the MIZ over summer with maximum coverage of only 0.36 million km<sup>2</sup> in December (Savidge et al., 1996; Fitch and Moore, 2007). Evidence from culture studies and blooms in the Ross Sea indicate that Phaeocystis sp. is more tolerant of deeply mixed, low light environments than diatoms (Arrigo et al., 1999; Moisan and Mitchell, 1999). Therefore, a more deeply mixed MIZ could cause a shift toward Phaeocystis sp. dominated blooms.

Large, early season blooms of Phaeocystis sp. can be responsible for substantial carbon export, rapidly sinking from surface waters and avoiding grazing pressure. Phaeocystis sp. colonies are encased in a tough outer coating, providing an effective defense against grazing protozoa and small zooplankton (Smetacek et al., 2004). In combination with their ability to draw down larger amounts of CO<sup>2</sup> than diatoms (Arrigo et al., 2000), it is likely that an increase in blooms dominated by Phaeocystis sp. may enhance carbon export in the MIZ (DiTullio et al., 2000). Phaeocystis sp. are also responsible for generating large amounts of DMSP (DiTullio and Smith, 1995; Turner et al., 1995; Vance et al., 2013). If increased mixing favors Phaeocystis sp. growth, it may counteract some of the loss of DMSP from decreased SIE in the SSIZ (see SSIZ above).

An increase in wind and wave action could also potentially increase the area of the MIZ by increasing the breakup and dispersal of sea ice by waves (Yin, 2005; Hemer et al., 2010; Young et al., 2011; Dobrynin et al., 2012; Stroeve et al., 2016). In spring and summer, large waves propagate through the sea ice up to 200 km, breaking up ice floes and accelerating ice retreat (Kohout et al., 2014; Horvat et al., 2016). Some satellite derived estimates of the MIZ region suggest a positive trend in MIZ area over time during spring (Stroeve et al., 2016), although not all models agree due to difficulties in accurately mapping the MIZ from satellite images (Ackley et al., 2003). However, changes in MIZ area are not likely to be uniform within the SSIZ, with Massom et al. (2006) reporting a contraction of the MIZ in the WAP due to strong northerly winds from the ASL (see SSIZ above). Interestingly, intense phytoplankton blooms still occurred amongst in the slurry of frazil ice between floes in this region (Massom et al., 2006), suggesting MIZ size is not necessarily a good indicator of its productivity.

Bloom formation within the MIZ is reliant on the coincidence of optimal conditions for phytoplankton growth. Increases in turbulent mixing by wind and waves would decrease light availability through a deepened mixed layer, with likely reductions in productivity and changes in the phytoplankton community structure within MIZ blooms. Additional nutrient inputs from melting icebergs are likely to cause localized increases in productivity but the extent of this effect would be felt most in the SSIZ, where growth of phytoplankton has drawn down nutrient concentrations. The net effect of future increases in MIZ area and decline in overall SIE remain uncertain.

#### 6. ANTARCTIC CONTINENTAL SHELF ZONE

Antarctic Continental Shelf Zone (CZ) waters make up the smallest area of the SO (1.28 million km<sup>2</sup> ) but they are also highly productive, contributing 66.1 Tg C yr−<sup>1</sup> or an average of 460 mg C m−<sup>2</sup> d −1 (Arrigo et al., 2008b). The high productivity in this region is due to high surface nutrient concentrations; iron enrichment from coastal sediments and basal shelf melt; and upwelled upper circumpolar deep water (UCDW, **Figure 3**) onto the continental shelf from the easterly-flowing Antarctic Slope Current, which approximately follows the 1,000 m isobath (**Figure 2**) (Jacobs, 1991; Smetacek and Nicol, 2005; Westwood et al., 2010; Williams et al., 2010). Blooms in CZ waters make a vital contribution to supporting the abundance and diversity of life in Antarctica. They attract large numbers of grazers that consume phytoplankton, that in turn feed higher tropic levels, while also producing fecal pellets, that are either remineralized into nutrients by heterotrophic microbes or sink rapidly into deep water, supporting the biological pump (Cadée et al., 1992; Turner, 2002; Honjo, 2004; Schnack-Schiel and Isla, 2005). Open water regions over the CZ are important foraging areas for many Antarctic species, especially during the summer breeding season (Arrigo and van Dijken, 2003; Smith et al., 2007; Stroeve et al., 2016). For example, DMS released from grazed phytoplankton acts as an olfactory foraging cue for white-chinned petrels (Nevitt et al., 1995) and Adélie penguin breeding success has been related to the proximity of colonies to open water (Ainley et al., 1998). The CZ is also a significant CO<sup>2</sup> sink over the summer as high rates of primary productivity cause surface CO<sup>2</sup> undersaturation (Hoppema et al., 1995; Gibson and Trull, 1999; Ducklow et al., 2007; Arrigo et al., 2008a; Roden et al., 2013).

Polynyas contribute to high productivity over the CZ with average annual primary production rates up to 105.4 g C m−<sup>2</sup> yr−<sup>1</sup> (Arrigo and van Dijken, 2003; Arrigo et al., 2015). Strong, cold katabatic winds freeze the surface water of the polynya, creating ice that is pushed north, adding to the seasonal sea ice extent and contributing to the generation of Antarctic Bottom Water through exclusion of high salinity brine by sea ice as it forms (Orsi et al., 1999). The Ross Sea polynya is the largest and the most productive polynya in Antarctica, contributing on average, 22.2 Tg C yr−<sup>1</sup> (Arrigo et al., 2015), with daily production as high as 6 g C m−<sup>2</sup> d −1 (Smith and Gordon, 1997). These high productivity rates are likely due to substantial iron input from upwelling of underlying sediments and basal melt of nearby ice shelves (Arrigo et al., 2015). Future increases in sea surface temperature are likely to accelerate the melting of ice shelves, increasing the input of fresh, stratified, iron-rich water to polynyas. increasing productivity in these regions (Feng et al., 2010).

Spatial differences in the factors controlling phytoplankton production have been observed within CZ waters. Consequently, the cause and rate of climate-induced change in these waters differs with location. Substantial differences have already been observed between East and West Antarctica (Turner et al., 2014 and references therein) and as such, we separately address the effects climate change on the phytoplankton communities in each of these two regions.

#### 6.1. West Antarctica

The West Antarctic CZ spans from the Amundsen and Bellingshausen Seas in the west to the Weddell Sea in the east and is dominated by the Antarctic Peninsula. Productivity is highest along the WAP and the Weddell Sea with rates of over 600 mg C m−<sup>2</sup> d <sup>−</sup><sup>1</sup> during the peak of summer (El-Sayed and Taguchi, 1981; Arrigo et al., 2008b; Vernet et al., 2008). The flow of warm, nutrient-rich UCDW onto the continental shelf (**Figure 3**) in the WAP accelerates sea ice retreat and enhances phytoplankton productivity (Kavanaugh et al., 2015), fostering diatom blooms as in the MIZ (see above). These are replaced by small flagellate and cryptophyte communities in the fresher, more stratified surface water later in the season (Moline et al., 2004; Ducklow et al., 2007). High production in the WAP and Scotia Sea support abundant krill populations, which are in turn food for a wealth of higher predators (Ducklow et al., 2007 and references therein).

Climate change threats to the West Antarctic CZ include warming, freshening, increased stratification, the melting and break up of glaciers and ice shelves, and ocean acidification (**Figure 8B**). The WAP is one of the fastest warming regions on Earth with an increase in the mean atmospheric temperature of 2 ◦C (6◦C in the winter) since 1950 (Meredith and King, 2005; Ducklow et al., 2007). No similar warming event has occurred on Earth in the last 1,800 years (Vaughan et al., 2003). Along with atmospheric warming in the WAP, increased heat delivery of UCDW from the Antarctic Circumpolar Current onto the shelf has caused a 0.6◦C increase in temperature of the upper 300 m of the water column (Meredith and King, 2005; Turner et al., 2014). This warming trend has resulted in increased glacial melt, with 87% of glaciers in the Antarctic Peninsula showing signs of retreat since 1950 (Cook et al., 2005; Peck et al., 2010). Glacial melt has resulted in an influx of fresh water to coastal regions of the WAP, freshening and increasing the stratification of surface waters over the summer. While phytoplankton productivity is expected to increase with increasing sea surface temperature (Rose et al., 2009), the phytoplankton community is likely to be more affected by resultant changes in SIE and freshwater inputs to the CZ (Arrigo et al., 2015; Moreau et al., 2015).

Freshening of surface waters from glacial melt has led to a documented change in the phytoplankton community in the WAP from diatom-dominated assemblages to cryptophytes and small flagellates (Moline et al., 2004; Montes-Hugo et al., 2008). The resultant shift in size distribution from large to small phytoplankton cells has had a significant flow on effect to zooplankton grazers, particularly krill and salps (Moline et al., 2004). This region is historically an area of high krill abundance, which is the preferred food source for the many Antarctic birds and mammals that live in the WAP (Atkinson et al., 2004). Changes to the phytoplankton community structure, favoring small cells, negatively affects krill grazing as they feed most efficiently on cells >10 µm and are unable to capture particles <6 µm in size (Kawaguchi et al., 1999). This has caused a shift in dominance to salps, mucoid feeders that are unaffected by the particle size of their prey (Moline et al., 2004), and a shift toward a less efficient fish-based food web (Murphy et al., 2007). Reductions in the krill population in the WAP are expected have a significant negative effect on the food chain in this region (Ballerini et al., 2014).

Surface water freshening causes a concurrent stratification of the water column, elevating phytoplankton into shallow mixed layers with higher light intensity. Phytoplankton productivity is enhanced through increased light, however, excessive light and elevated UV-A and UV-B exposure can lead to photoinhibition and cell damage (see SAZ above, Moreau et al., 2015). In order to limit the damage of these conditions, phytoplankton can channel metabolic reserves into photoprotection and tolerance mechanisms (Davidson, 2006). A lengthening of the open water season in the WAP, caused by earlier sea ice retreat (see SSIZ above), has increased productivity in the CZ, whilst also increasing photoinhibition rates (Moreau et al., 2015). Thus far, the increase in production is much greater than the loss due to photoinhibition so it is expected that increased stratification will lead to a net increase in primary productivity in the future (Moreau et al., 2015).

Stronger westerly winds, as a result of a positive SAM, are bringing warmer air across the Antarctic Peninsula, increasing

snowfall and causing a break up of large ice shelves (e.g., Scambos et al., 2000; Rack and Rott, 2004; Turner et al., 2014). The break up of the Larsen A ice shelf created new areas of high nutrient open water, stimulating phytoplankton blooms and increasing productivity in a previously ice-covered pelagic habitat (Bertolin and Schloss, 2009). The continued retreat of glaciers and breaking up of ice shelves has led to the creation of new carbon sinks around the Antarctic Peninsula that have increased productivity up to 3.5 Tg C m−<sup>2</sup> yr−<sup>1</sup> (Peck et al., 2010). The continued break up of ice shelves will also lead to an increase in icebergs over the CZ. Melting icebergs have been found to provide a significant amount of iron and nutrients to surface waters, leading to increased phytoplankton productivity (Lin et al., 2011; Vernet et al., 2011; Duprat et al., 2016). Increased iceberg numbers will also contribute to increased productivity throughout the SSIZ and POOZ as they are propelled by ocean currents around Antarctica (see above).

Little work has investigated the effect of ocean acidification on phytoplankton in West Antarctic waters. The CO<sup>2</sup> concentration in waters over the West Antarctic CZ vary seasonally from ∼176 to 503 µatm through to the uptake of CO<sup>2</sup> by phytoplankton in the summer and return to super-saturated levels in winter under the sea ice (Moreau et al., 2012). Coastal phytoplankton communities from the WAP (both diatom-dominated and mixed diatom-flagellate communities) displayed no significant change in community composition, cell size, or growth rate when exposed to 800 µatm CO<sup>2</sup> (Young et al., 2015). Yet, results of this study did demonstrate the differences in physiological carbon uptake among phytoplankton species as production by diatoms may be enhanced by down-regulation of CCMs at high pCO2, while a slight decline in production by Phaeocystis sp. was attributed to the alternative bicarbonate transport pathway used by this species.

#### 6.2. East Antarctica

The East Antarctic CZ ranges from the Ross Sea in the east to the eastern edge of the Weddell Sea in the west. The Ross Sea is the most productive region in the CZ, contributing ∼24 Tg C m−<sup>2</sup> yr−<sup>1</sup> and accounting for ∼30% of the total annual production in shelf waters (Sweeney et al., 2000; Arrigo et al., 2008b). Iron and light availability are the dominant factors controlling growth of phytoplankton in the Ross Sea (**Figure 8A**) (Smith et al., 2000b; Feng et al., 2010; Sedwick et al., 2011). In addition, the relative abundances of the dominant phytoplankton (diatoms and Phaeocystis sp.) are linked to mixed layer depth, with diatoms dominant in highly stratified water and Phaeocystis sp. where it is deeply mixed (Arrigo et al., 1999). These phytoplankton blooms support a unique food web in the Ross Sea, structured around the crystal krill, Euphausia crystallorophias, and the Antarctic silverfish, Pleuragramma antarcticum (Smith et al., 2007). Elsewhere around East Antarctica the CZ is relatively narrow and contributes ∼12 Tg C m−<sup>2</sup> yr−<sup>1</sup> (Arrigo et al., 2008b) and like the West Antarctic CZ, the Antarctic krill, E. superba, is a keystone species (Nicol et al., 2000, 2010). Here the phytoplankton community is dominated by blooms of diatoms and Phaeocystis sp. during the summer and shifts to small flagellates once nutrients have been exhausted (Waters et al., 2000; Wright and van den Enden, 2000; Davidson et al., 2010).

The East Antarctic CZ is expected to experience increased freshening, stratification, the melting and break up of glaciers and ice shelves, ocean acidification, and modest warming (**Figure 8B**). In contrast to the warming trend around West Antarctica, there has been a measured cooling over East Antarctica for the same period (1969–2000) (Thompson and Solomon, 2002). Despite this, most recent model projections for the Ross Sea by the end of the century predict a 0.15– 0.4◦C increase in SST, with decreases in the mixed layer depth (∼50–70 m), sea ice concentration (2–11%), and macronutrient concentrations (Rickard and Behrens, 2016). Freshening has already been reported in the Ross Sea and has been attributed to changes in precipitation, sea ice production, and melting of the West Antarctic ice sheet (Jacobs et al., 2002). Projected changes to the remaining area of East Antarctica are not well understood but similar trends are anticipated (Watanabe et al., 2003; Convey et al., 2009; Gutt et al., 2015).

Ocean acidification is anticipated to affect polar waters sooner than the rest of the world, due to the increased solubility of CO<sup>2</sup> in cold water (Orr et al., 2005; McNeil and Matear, 2008). Phytoplankton communities in Antarctic shelf waters are already exposed to strong annual variations in pCO<sup>2</sup> (Gibson and Trull, 1999; Sweeney et al., 2000; Roden et al., 2013; Shadwick et al., 2013; Kapsenberg et al., 2015). Sea ice cover during the winter restricts air-sea gas transfer, allowing for CO<sup>2</sup> oversaturation of the water column (up to 450 µatm) through upwelling of high CO<sup>2</sup> UCDW water from the Antarctic Slope Current (**Figure 3**). Photosynthetic drawdown over the summer can result in CO<sup>2</sup> levels falling below 100 µatm. This large seasonal variation seems to favor species that tolerate large fluctuations in pH. Phytoplankton communities have been observed to show little change in composition when grown at CO<sup>2</sup> concentrations similar to those already experienced in coastal environments (84–643 µatm) (Davidson et al., 2016). However, superimposing anthropogenic pCO<sup>2</sup> increase upon the large natural fluctuation already occurring in the natural environment may push some species past their limit sooner than anticipated (McNeil and Matear, 2008), causing changes in phytoplankton productivity, growth and community composition. Concentrations of CO<sup>2</sup> exceeding 1,000 µatm induced a change in phytoplankton community composition in Prydz Bay, increasing the abundance of small phytoplankton species (Davidson et al., 2016; Thomson et al., 2016). Studies on Ross Sea phytoplankton communities also suggest that high CO<sup>2</sup> concentrations (760–800 µatm) may cause a shift in dominance in this region from Phaeocystis sp. to large chain-forming diatom communities (Tortell et al., 2008; Feng et al., 2010). Investigation into the physiological reasons for changes in growth rates link increased growth and carbon fixation to the energy saved through the down-regulation of CCMs (Rost et al., 2008; Tortell et al., 2008), while inhibition of growth and productivity may be related to the metabolic costs of proton pumps to exclude hydrogen ions (Gao et al., 2012; McMinn et al., 2014).

A change in phytoplankton community composition will likely have significant effects on carbon export in the CZ. A shift toward smaller cell communities will allow for increased remineralization of cells through the microbial consumption, decreasing the downward flux of carbon into the deep ocean (Finkel et al., 2010 and references therein). These cells are also likely to be less efficiently grazed by zooplankton, resulting in less carbon transfer to higher trophic organisms. Any CO2-induced increase in the dominance of diatoms in the Ross Sea may cause a decline in net carbon export as blooms of Phaeocystis sp. are capable of exporting more carbon than diatoms (Arrigo et al., 2000). However, diatom-dominated communities are likely to be grazed more than Phaeocystis sp., providing better nutrition for the Antarctic food web and also producing negatively buoyant feces that can assist in the sinking of diatoms (Schnack-Schiel and Isla, 2005).

Whilst most studies have focused on individual factors predicted to alter as a result of climate change, phytoplankton in the SO will be simultaneously exposed to multiple climate change stressors (Gutt et al., 2015). Recent work has focused on the interaction of multiple stressors on phytoplankton growth in the Ross Sea, highlighting the complex interaction between environmental changes and the phytoplankton community (Rose et al., 2009; Feng et al., 2010; Xu et al., 2014; Zhu et al., 2016). Iron promotes phytoplankton growth, whereas interactive effects between iron, warming, increased CO2, and light favor the dominance of diatoms over Phaeocystis sp. (Rose et al., 2009; Xu et al., 2014; Zhu et al., 2016). In contrast, high pCO<sup>2</sup> only affected diatoms, favoring the growth of large centric species (Feng et al., 2010). As well as causing shifts in phytoplankton taxa, changes in temperature and iron supply caused modifications to microzooplankton abundance, suggesting possible changes in predator/prey interactions (Rose et al., 2009). No multi-stressor experiments have yet been performed on other East Antarctic phytoplankton communities, though it appears likely that climate-induced change will alter the competitive interactions among dominant phytoplankton taxa and change trophodynamics throughout continental waters.

Freshening, increased stratification, ocean acidification, and the melting and break up of glaciers and ice shelves are all occurring across the Antarctic CZ due to climate change. Phytoplankton growth is promoted by freshening, increased stratification, and the break-up of ice shelves by establishing conditions that are optimal for growth, most notably an increase in iron supply and light availability. However, freshening and ocean acidification also appear to be responsible for shifts in community composition that could result in a decrease in food quality and availability for grazers. This could have a significant negative effect on the structure and function of the Antarctic food web as well as reducing carbon export. In contrast, the proposed CO2-induced increase in abundance of large diatoms in the Ross Sea may benefit the food web in this region but may still result in a decline in carbon export.

Temperature trends currently differ between East and West Antarctica, with significant warming in West Antarctica and a slight cooling trend over East Antarctica. Increases in temperature appear to promote phytoplankton growth and may accelerate sea ice retreat, changing the timing and magnitude of bloom onset in this highly productive region. However, the interactive effects that this combination of climate stressors will have on phytoplankton communities in this region is not well understood. Further work will be required before we can fully understand how phytoplankton over the CZ will be affected by a changing climate.

# 7. CONCLUSION

The SO comprises a vast expanse of ocean containing a diverse array of environments, each of which exposing phytoplankton to environmental factors that limit their production, growth, survival, and composition. Despite these stressors, phytoplankton thrive in some of the most extreme conditions on earth. Climate-induced changes in the physical characteristics of the SO and the responses by phytoplankton differ substantially among environments. No long-term trends in satellite-derived Chl a or primary productivity are yet detectable due to the large background of interannual/decadal variability (Henson et al., 2010; Gregg and Rousseaux, 2014). It is unlikely that unambiguous trends due to climate change will be seen until approximately 2055 (Henson et al., 2010). However, some longer time series of underway Chl a measurements exist that could indicate climate-induced trends (see below).

Given the competing influences on phytoplankton within each region of the SO, predictions are bound to be tentative and contentious. Our assessment of the available information suggest the responses of phytoplankton in various regions of the SO are:


and destabilizing the seasonal progression of phytoplankton blooms (**Figure 7**). Such changes would reduce the frequency of ice edge blooms and cause taxonomic shifts in the phytoplankton community toward small diatoms and flagellates.

• In the CZ, the few available studies suggest that warming, freshening, and ocean acidification are likely to elicit changes to community composition (**Figure 8**), with reports of a shift toward communities composed of smaller cells and flagellates (Moline et al., 2004; Davidson et al., 2016). Increased nutrients and stratification from melting glaciers and icebergs are likely to increase productivity (**Figure 8**). Localized shifts in community composition in the Ross Sea toward diatomdominated communities will potentially decrease carbon export but may provide better nutrition for higher trophic levels.

These changes are likely to have a significant effect on the biogeochemical processes in the SO, affecting the biological pump, microbial loop, and nutrition for higher trophic levels. It is likely that the effect of climate change on phytoplankton in each of these regions is going to be determined by the timing, rate, and magnitude of change in each stressor; as well as the sequence in which these stressors are imposed. Climate change models of the SO still contain large uncertainties, in part due to knowledge gaps in biogeochemical processes and carbon uptake (Frölicher et al., 2016). The vast majority of phytoplankton research in the SO have been observational studies, providing essential data on phytoplankton communities, seasonal community succession, nutrient utilization, primary and export production, and food web interactions (e.g., El-Sayed, 1994; Nicol et al., 2000, 2010; Smith et al., 2000a; Olguín and Alder, 2011; Quéguiner, 2013). These studies are essential for our understanding of the current and potential future state of SO phytoplankton. Relatively few studies have focused on the manipulation of climate stressors on SO phytoplankton species/communities (e.g., Tortell et al., 2008; Rose et al., 2009; Hoppe et al., 2013; Müller et al., 2015; Boyd et al., 2016; Coad et al., 2016; Davidson et al., 2016). More of these studies are necessary in all of the regions of the SO to determine the thresholds for climate-induced stressors on phytoplankton communities. It is also important to perform multi-stressor experiments, incorporating a range of environmental factors affected by climate change, if we are to understand the interactive effects (from synergistic to antagonistic) of future stressors on phytoplankton species and communities (e.g., Feng et al., 2010; Xu et al., 2014; Boyd et al., 2016; Zhu et al., 2016).

The vastness and environmental diversity of the SO; the inherent spatial and temporal variability in phytoplankton communities; and the logistical costs and difficulty in obtaining data from the SO, especially year-round observations, means the effect of climate change on phytoplankton in this region is poorly understood. In some instances, advances in remote sensing technology and computer modeling have allowed access to data sets that can assist in understanding trends. However, they are still limited in their ability to detect some physical changes, such as sea ice thickness and Chl a concentration in waters covered by ice (Massom et al., 2006; Hobbs et al., 2016). There are very few places that have long-term monitoring programs to detect changes in the physical and biological environment (such as the Palmer-Long Term Ecological Research program, Smith et al., 1995) and few of these have collected data for a sufficient duration to detect trends in phytoplankton against the background of natural variation. Decades long monitoring programs should be established as a matter of urgency to detect changes in SO phytoplankton abundance, production, and composition.

Stratospheric ozone concentrations exert a pervasive effect on atmospheric circulation in the Southern Hemisphere and recovery of the ozone hole will change the trajectory of climate. Concern over ozone depletion and the consequent rise in short wave UV radiation reaching the Earth's surface, galvanized the international community, culminating in the Montreal protocol, which banned the use of ozone depleting substances, such as chlorofluorocarbons (CFCs) and halons. Unrecognized at the time, ozone depletion was also the primary cause of increases in the positive phase of the SAM, resulting in the acceleration and poleward shift of westerly winds over the SO (see POOZ above, Polvani et al., 2011; Thompson et al., 2011). This proved to be the most obvious and persistent characteristic of Southern Hemisphere climate change in the last half century (Thompson and Wallace, 2000; Polvani et al., 2011). Modeling studies indicate that recovery of the ozone hole will decelerate the westerly winds (Son et al., 2008) and result in a more rapid rise in Antarctic temperatures than elsewhere in the Southern Hemisphere (Shindell and Schmidt, 2004). Nearly 30 years after the Montreal protocol came into effect, the first signs are emerging that the ozone hole is beginning to heal (Solomon et al., 2016). Projections suggest that ozone concentrations in the stratosphere are likely to return to pre-ozone hole values around 2065 (Son et al., 2008; Schiermeier, 2009). Thus, the main factor presently driving climate change and phytoplankton responses over much of the SO will decline over the next half century. Ozone depletion and positive SAM cause increases in wind and wave action, deeper mixing, and increased nutrient entrainment into surface waters (see POOZ, SSIZ, MIZ above). Replenishment of ozone is likely to reverse these climate-induced drivers of phytoplankton dynamics in Antarctic waters, moving to a scenario reminiscent of the SAZ region and dominated by increased warming, stratification, and declining nutrient availability in surface waters. The effect of this reversal in climate fortunes is unknown but the rate of change (∼50 years) may prove too fast for some species to adapt and/or evolve to the changing environment.

The response of phytoplankton to anticipated future environmental conditions in the SO will eventually depend upon their capacity to adapt and evolve (Boyd et al., 2016 and references therein). Phytoplankton communities have short generation times and high genetic diversity, which allow for adaptation to changing environmental conditions through natural selection (Collins et al., 2014). Some SO phytoplankton communities are already exposed to large variations in their environment, such as sea ice and coastal communities. Phytoplankton that are already exposed to large variations in their environment are considered inherently more tolerant and capable of adapting to future changes (Sackett et al., 2013; Schaum and Collins, 2014). Davidson et al. (2016) showed that exposing natural microbial communities to the large range in CO<sup>2</sup> concentrations they encounter in nature over a year had little effect. Concentrations above this reduced productivity and changed the composition of the phytoplankton community, suggesting that their tolerance to variability outside of those normally encountered was low. Furthermore, current experiments, which determine the tolerance limits of phytoplankton over short time scales, may not be a good indicator of long-term resilience as the metabolic costs of climate-induced stress may not be sustainable over numerous generations (Schaum and Collins, 2014; Torstensson et al., 2015). It is currently unknown whether the rate of environmental change will outpace the ability of SO phytoplankton to adapt and/or evolve. It is, however, inevitable that changes at the base of the SO will influence trophodynamics, biogeochemistry, and climate change.

#### AUTHOR CONTRIBUTIONS

SD and AD both wrote, edited and produced the figures for the manuscript.

#### FUNDING

This review was funded by the Australian Government, Department of Environment and Energy as part of Australian Antarctic Science Project 4026 at the Australian Antarctic Division and an Elite Research Scholarship awarded by the Institute for Marine and Antarctic Studies, University of Tasmania.

# ACKNOWLEDGMENTS

We would like to thank Indiah Hodgson-Johnston and the staff at the Australian Antarctic Division Data Centre for assistance with the production of our figures.

#### REFERENCES


on the basis of historical and modern sea ice records. Polar Res. 22, 19–25. doi: 10.3402/polar.v22i1.6439


and its biogeochemical significance. Global Biogeochem. Cycles 30, 1124–1144. doi: 10.1002/2016GB005414


antarctica blooms in the Ross Sea, Antarctica. Nature 404, 595–598. doi: 10.1038/35007061


Ocean across the CMIP5 model suite. Biogeosciences 12, 5715–5734. doi: 10.5194/bg-12-5715-2015


in the Weddell-Scotia Confluence during the austral spring. Limnol. Oceanogr. 43, 433–448. doi: 10.4319/lo.1998.43.3.0433


over the central Kerguelen Plateau. Deep Sea Res. I Oceanogr. Res. Papers 111, 91–101. doi: 10.1016/j.dsr.2016.02.017


naturally iron-fertilized bloom in the Southern Ocean. Glob. Biogeochem. Cycles 26:GB1014. doi: 10.1029/2010GB003977


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Deppeler and Davidson. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Microbial Eukaryotes in an Arctic Under-Ice Spring Bloom North of Svalbard

Archana R. Meshram1, 2, Anna Vader <sup>1</sup> , Svein Kristiansen<sup>3</sup> and Tove M. Gabrielsen<sup>1</sup> \*

<sup>1</sup> Department of Arctic Biology, University Centre in Svalbard, Longyearbyen, Norway, <sup>2</sup> Department of Biosciences, Centre for Ecological and Evolutionary Synthesis, University of Oslo, Oslo, Norway, <sup>3</sup> Faculty of Biosciences, Fisheries and Economics, UiT The Arctic University of Norway, Tromso, Norway

Microbial eukaryotes can play prominent roles in the Arctic marine ecosystem, but their diversity and variability is not well known in the ice-covered ecosystems. We determined the community composition of microbial eukaryotes in an Arctic under-ice spring bloom north of Svalbard using metabarcoding of DNA and RNA from the hypervariable V4 region of 18S nrDNA. At the two stations studied, the photosynthetic biomass was dominated by protists >3 µm and was concentrated in the upper 70–80 m, above the thermocline and halocline. Hierarchical cluster analyses as well as ordination analyses showed a distinct clustering of the microbial eukaryote communities according to a combination of water mass and local environmental characteristics. While samples collected in the surface mixed layer differed distinctly between the two sites, the deeper communities collected in Atlantic Water were fairly similar despite being geographically distant. The differentiation of the microbial eukaryote communities of the upper mixed water was probably driven by local development and advection, while the lack of such differentiation in the communities of Atlantic Water reflects the homogenizing effect of water currents on microbial communities.

Keywords: Arctic, under-ice spring bloom, microbial eukaryotes, 18S V4 region, pyrosequencing

# INTRODUCTION

The Arctic spring bloom fuels production in the seasonally ice-covered shelves surrounding the Arctic Ocean. The sea ice and pelagic communities are dominated by microalgae such as diatoms (von Quillfeldt et al., 2009; Hodal et al., 2012) and to varying degrees colonies of the haptophyte Phaeocystis pouchetii (Eilertsen et al., 1981; Wassmann et al., 1999; Degerlund and Eilertsen, 2010). During the last decade, sequencing of 18S rDNA has been used to metabarcode communities of also the pico- (0.2–3 µm) and nano- (3–20 µm) sized eukaryotic fraction of Arctic marine systems (e.g., Lovejoy et al., 2006; Comeau et al., 2011; Kilias et al., 2014; Marquardt et al., 2016) as reviewed by Lovejoy (2014). The insights of arctic microbial eukaryotes gained by the use of molecular genetic tools include the identification of ecotypes endemic to the Arctic (Lovejoy et al., 2007; Terrado et al., 2013; Percopo et al., 2016), the existence of taxa-specific biogeographic patterns (Thaler et al., 2015), and a linking of distinct protist assemblages to water masses (Hamilton et al., 2008; Bachy et al., 2011; Monier et al., 2014; Metfies et al., 2016). It has been shown that the pronounced Arctic seasonality extends to the succession of microbial eukaryotes (Marquardt et al., 2016; Joli et al., 2017). Molecular data can also provide insight into responses of marine microbes to environmental change (Comeau et al., 2011) with potential changes higher up in the marine food web.

#### Edited by:

Eva Ortega-Retuerta, UMR7621 Laboratoire d'Océanographie Microbienne (LOMIC), France

#### Reviewed by:

Ludwig Jardillier, Université Paris-Sud, France Charles Bachy, Monterey Bay Aquarium Research Institute, United States

> \*Correspondence: Tove M. Gabrielsen tove.gabrielsen@unis.no

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Microbiology

Received: 17 October 2016 Accepted: 30 May 2017 Published: 28 June 2017

#### Citation:

Meshram AR, Vader A, Kristiansen S and Gabrielsen TM (2017) Microbial Eukaryotes in an Arctic Under-Ice Spring Bloom North of Svalbard. Front. Microbiol. 8:1099. doi: 10.3389/fmicb.2017.01099

The area north of Svalbard in the European Arctic is seasonally covered by sea ice, although with a diminishing ice concentration the last decade (Onarheim et al., 2014). The region is strongly influenced by inflow of Atlantic Water (AW) with relatively high salinity and temperature (≥34.8 and 2◦C, respectively) from the West Spitsbergen Current (Cokelet et al., 2008). Increased temperature in the inflowing AW (Walczowski et al., 2012) is probably the main driver of the reduced sea ice concentration in the area which is seen predominantly in winter (Onarheim et al., 2014). At the same time there is a trend of increased boreal plankton in the Atlantic inflow water (Weydmann et al., 2014; Paulsen et al., 2016) pointing to the importance of AW for the transport of taxa along the Spitsbergen coastline.

The upper mixed layer north of Svalbard is characterized by reduced salinity (≤34.4; Manley, 1995) due to riverine input and melting of sea ice throughout the Arctic Ocean (Lind and Ingvaldsen, 2012). Upwelling of the deeper AW to the surface was demonstrated in the winter of 2012 along the northern Svalbard shelf (Falk-Petersen et al., 2015), thus mixing this more saline and warmer water and its northward transported plankton to the surface. The extreme environmental variability in this high Arctic region makes it an excellent area to study the contrasts between homogenization of the microbial communities due to potential large-scale dispersal with water currents and environmental filtering due to local conditions.

In this study, we investigated microbial eukaryote diversity and community compositional differences in an under-ice spring bloom north of Svalbard. Our main objectives were: (1) To indentify the community of microbial eukaryotes in two icecovered stations, (2) to unravel the influence of water masses of different characteristics and history vs. local processes on the composition of the microbial protist communities, and (3) to study the metabolically active fraction of the community by comparing inventories based on rRNA and the rRNA gene.

#### MATERIALS AND METHODS

#### Sample Collection

Seawater samples were collected in May 2010 from two icecovered stations, Stations 1 and 2, both located north of Spitsbergen (**Figure 1**). Station 1 in this paper was sampled between St. 2 and St. 3 of Johnsen et al. (submitted). After having drifted out of the sea ice, as described by Johnsen et al. (submitted), the ship re-located further NE into the sea ice, where our Station 2 was sampled. Sampling depths were selected based on hydrography and fluorescence profiles aiming to sample potentially different habitats for microbial protists. Water samples from five depths (0, 10, 45, 65, and 200 m) were collected at Station 1 and from four depths (0, 15, 120, and 200 m) at Station 2. From each depth, seawater was sequentially filtered through 3 µm filters (Isopore membrane filter, Millipore, USA) onto 0.22 µm filters (Durapore membrane hydrophilic PVDF filter, Millipore, USA) using a peristaltic pump at 40 rpm. One L was filtered for DNA extraction and 2 L were filtered for RNA extraction. Immediately after filtration, the RNA filters were submerged in 1 mL of Trizol (Invitrogen), the tube was shaken vigorously and kept at room temperature for 5 min. Filters were snap-frozen in liquid nitrogen and kept at −80◦C until extraction. To measure total and >3 µm Chl a biomass, 250 mL seawater was filtered in triplicate onto GFF filters (Whatman, England) and 3 µm Isopore membrane polycarbonate filters (Millipore, USA), respectively. The filters were extracted in methanol for 12–24 h at 4◦C in the dark and analyzed using a 10-AU-005-CE fluorometer (Turner, USA). Seawater for nutrient analyses was collected in acid-washed and sample-rinsed bottles and kept frozen until analysis. Nitrate + nitrite (NO3+NO2), phosphate (PO4), and silicate (Si(OH)4) were analyzed in triplicate by standard seawater methods using a Flow Solution IV analyzer from O.I. Analytical, USA. Photosynthetically active radiation (PAR) from the surface down to 40 m was measured using a cabled, flat Li-Cor sensor (LI-1400). The temperature, salinity and density measurements taken from the onboard SeaBird CTD was plotted in a TS diagram, and the water masses were characterized according to Aagaard and Carmark (1989); Manley (1995), and Cokelet et al. (2008). The thickness of the sea ice and the snow depths on top of the sea ice were measured in 6–10 drilled holes per station.

#### Nucleic Acid Extraction and Pyrosequencing

DNA was extracted using a modified CTAB protocol according to Vader et al. (2014) and each filter was cut into two halves and extracted individually. After thawing the filters sampled for RNA, the Trizol was transferred to a new tube and the filter was washed with 1 mL of fresh Trizol. In both tubes, the cells were lysed by beating two times 1 min with 200 µm zirconium beads (Molecular grade, pre-filled tubes, OPS Diagnostics) in a bead beater at 1/22s. Subsequently total RNA was extracted according to the manufacturers' protocol, except that 10 µg glycogen (RNA grade, Fermentas) was added to each tube as a carrier during precipitation of the RNA. The RNA was dissolved in DEPC-treated water, and DNAse-treated with Turbo DNAse (Ambion) according to the manufacturers protocol. RNA was reverse transcribed using the Retro Script kit (Ambion) with random decamers at 44◦C for 1 h, after an initial denaturation step of 3 min at 75◦C.

The hypervariable V4 18S rDNA region was amplified from the rRNA gene and from cDNA from reverse transcribed rRNA (referred to DNA and RNA, respectively) using eukaryotespecific 454 primers (Comeau et al., 2011). Samples were amplified in triplicate 25 µl reactions containing 1xHF-buffer, 0.2 µM of each dNTP, 0.2 µM of each primer, 0.5 U Phusion Hot start HF II polymerase and 1 µl undiluted DNA or RNA. All products were cleaned using AMPure XP beads (Agencourt, Beckman Coulter) according to the manufacturers instruction using a 4:5 ratio of beads to PCR product. Subsequently all PCR products were pooled at equimolar amounts and sequenced unidirectionally on a Roche 454 GS-FLX Titanium platform at the IBIS/Plateforme d'Analyses Génomiques de l'Université Laval.

#### Sequence Processing and Statistical Analysis

Sequence analyses were done in QIIME v1.5.0 and v1.8.0 (Caporaso et al., 2010). Low quality reads were removed using default parameters except for min quality score 30, min and max

sequence lengths 330 and 550 bp, max homopolymer length 7, and no ambiguous bases. Reads that passed the quality control were denoised using DeNoiser v0.851 (Reeder and Knight, 2010) as implemented in QIIME v1.5.0. Chimeras were removed in MOTHUR v1.32.1 (Schloss et al., 2009) with uchime (Edgar et al., 2011) using self as reference and default settings. Demultiplexed reads were clustered into operational taxonomic units (OTUs) at 98% sequence similarity (as often used in protist metabarcoding; cf. Comeau et al., 2011) using UCLUST (Edgar, 2010) with default parameters and the taxonomy of the most abundant representatives from each OTU were assigned using BlastN (evalue 0.00001) against the curated protist databases of Lovejoy et al. (2016) and PR2 (gb203; Guillou et al., 2013). Global singletons and reads assigned to Metazoa were removed. The number of sequence reads per sample was rarefied by random sampling to the sample having the lowest read number (6066). Diversity analyses were performed using the Vegan (version 2.0– 10) package in R (Version 0.98.994–RStudio, Inc.). The Shannon diversity index and Pielou's index of evenness were calculated to compare alpha-diversity between the samples. Rarefaction analyses were conducted and computed every 100 sequences using 5,000 subsampling iterations.

Beta diversities were analyzed by clustering samples using UPGMA (unweighted pairwise grouping method with averaging) based on Bray-Curtis dissimilarities. To identify a possible differentiation of the communities under the constraint of environmental factors, canonical correspondence analysis (CCA) was done using the PAST software (v3.04, Hammer et al., 2009). CCA was performed on a Bray-Curtis dissimilarity matrix of the samples. A permutation test was performed to calculate the significance of constraints on the data for the environmental parameters (fluorescence, density, depth, temperature, salinity, PAR and concentrations of nitrate/nitrite, phosphate, and silicate). The environmental parameters were tested for statistical dependence using the Spearman rank correlation coefficient (data not shown). For two highly co-varying parameters (salinity co-varied with density and silicate co-varied with phosphate) only one parameter was shown.

To visualize the differences among the libraries, a heat map of relative abundances was generated (in R) for the abundant OTUs (>1% of total reads) and the five most abundant OTUs in each library, and overlaid the UPGMA dendrogram.

# RESULTS

# Study Location and Hydrography

At the time of sampling both ice stations were snow-covered (30 ± 9.7 cm and 23 ± 6.3 cm at Station 1 and Station 2, respectively) and had thick overlying sea ice (121 ± 11.7 cm and 160 ± 32.3 cm at Station 1 and Station 2, respectively). The water masses were identified based on their potential temperature salinity characteristics; Atlantic Water (AW), Polar Water (PW), and Arctic Surface Water (ASW; Aagaard and Carmark, 1989; Cokelet et al., 2008; **Figures 2**, **3**). The hydrographic profiles of the two stations were similar, although the PW was deeper at Station 2 (90 m) compared to Station 1 (45 m; **Figure 3**). The upper samples from both stations were all collected within the PW (S < 34.4; θ < 0 ◦C), and designated PW1 (0, 10, and 45 m from Station 1) and PW2 (0 m and 15 m from Station 2; **Table 1**). The samples from 200 m at both locations displayed Atlantic Water properties. The samples from depths 65 m at Station 1 and 120 m at Station 2 were collected beneath the halocline and

thermocline at the two stations, and represented Arctic Surface Water (ASW1 and ASW2 from stations 1 and 2, respectively). The two stations both consisted of ice floes with leads in between, and the euphotic zone (1% of surface PAR) reached 27 m at Station 1 and 34 m at Station 2 (**Figure 3**).

squares represent Atlantic Water (AW) samples.

A distinct algal bloom had developed at Station 1, with Chl a concentrations of 10.7–16.0 µg L−<sup>1</sup> in the PW (**Table 1**). Station 2 saw an order of magnitude lower Chl a biomass at 1.0–1.3 µg L−<sup>1</sup> in the PW. The size fractionated Chl a-values showed that cells >3 µm dominated the photosynthetic biomass at both stations; at Station 1 the autotrophic biomass of the picoplanktonic fraction was low or below detection, and at Station 2 it was very low (0.06–0.08 µg L−<sup>1</sup> ).

The depth distribution of the measured nutrients (NO<sup>3</sup> + NO2, PO4, Si(OH)4) was similar at the two stations, with comparatively low nutrient concentrations in the PW, minimum concentrations at intermediate depths (120 m at Station 2) and considerably higher concentrations in the AW (**Table 1**).

# Sequence Analyses and Diversity Estimates

A total of 286,979 reads were obtained, which were reduced to 224,272 after quality filtering and chimera checking. Clustering the reads at 98% similarity produced 1,547 OTUs, which after removing OTUs classified as Metazoa, global singletons and subsampling to the smallest sample size (6,066 reads) were further reduced to 1,288 OTUs. The rarefaction curves (Figure S1) showed that all libraries had been sequenced to saturation. There was a general trend of increasing diversity with depth both in OTU richness and evenness, also reflected in the Shannon-Wiener diversity index (**Table 1**). A higher eukaryotic diversity was seen in the DNA samples compared to the RNA samples.

#### Microbial Eukaryote Community Structure

Hierarchical cluster analyses as well as the CCA showed a distinct clustering according to a combination of water mass and local environmental characteristics (**Figures 4**, **5**) where the two first axes in the CCA described 54.2% of the variation in the data. The two samples collected from ASW were clearly associated with the communities of either AW (the ASW2 sample from 120 m depth; **Figures 4**, **5**) or PW (the ASW1 sample clustered with the PW1 samples, in particular with the 45 m PW1 sample; **Figure 5**). The permutation test performed on the CCA confirmed the importance of water mass as a structuring factor, as significant interactions were identified between the community composition of individual samples and density (R 2 = 0.13, p = 0.01). In addition, the ANOVA identified fluorescence (R <sup>2</sup> = 0.19, p = 0.001) as the other significant factor structuring the communities. While density mostly separated along the first CCA axis (which explained 34.8% of the variability in the data), fluorescence separated the communities along the second CCA axis (which explained 19.4% of the variability in the data) thus separating the surface samples of the two stations. Other parameters explaining the variability seen in the community composition (although not statistically significant) were temperature, depth, nutrient concentrations and PAR (**Figure 4**).

# Microbial Eukaryote Communities

Alveolates were abundant in all samples, recruiting 30–80% of reads in individual libraries (**Figure 6**). While Ciliophora dominated in PW (7–73% of library reads), Dinophyceae and MALV were more common in AW (including the 120 m ASW sample from Station 2). The Dinophyceae reads were dominated by one abundant OTU (assigned to Dino\_clone\_NIF\_3A1; **Table 2**, **Figure 5**) that accounted for 3–44% of the total reads in individual libraries and was present in all libraries. While the maximum number of reads was detected in the 0 m DNA sample from Station 1 (1D0 library), RNA abundances were highest in the deep samples (all AW and 65 m from Station 1, **Figure 5**). Reads assigned to Woloszynskia halophila was also highly abundant in the 1D0 library, but did not have a similar read abundance in the RNA library (1R0; **Figure 5**). Reads assigned to MALV were more abundant in the DNA libraries than in the RNA libraries, while the dominant Ciliophora and Dinophyceae OTUs displayed the opposite pattern, recruiting a larger proportion of reads within the RNA libraries (**Figure 5**).

Three Ciliophora OTUs were among the abundant taxa; Choreotrichia-1 was abundant in the 45 and 65 m libraries from Station 1, whereas Strombidiidae and Oligotrichia were prevalent in PW, with especially high read numbers of the latter in RNA from Station 1 (**Figure 5**). Four additional Ciliophora OTUs were among the five most abundant in single libraries; Strombidium sp. did not show a very distinct distribution pattern, but was most abundant in PW, Monodinium sp. was

particularly abundant in the 0 m RNA sample of Station 1, Mesodiniidae was abundant in the RNA libraries from AW, and Protostomatea-1 was found in the PW (including at 65 m) of Station 1.

The haptophytes were dominated by an OTU assigned to Phaeocystis sp. (**Table 2**), which was the most abundant OTU in the total dataset (16% of all reads) and was especially prevalent in the PW (and ASW) at Station 1 (6–49% of library reads). The same OTU also recruited 5–21% of all reads in the deeper AW sample from the same station. Another dominant group was the pico-sized green algal phylum Mamiellophyceae which recruited 35–45% of reads from the PW samples of Station 2 (**Figure 5**). Nearly all of these sequences were assigned to the Arctic ecotype of Micromonas (Arctic Micromonas; **Table 2**).

A Bacillariophyceae OTU assigned to Thalassiosira antarctica was found at both stations, but showed highest relative abundance in PW at 45 m of Station 1. OTUs assigned to marine stramenopiles were found throughout the water column at both stations, with the most abundant OTU (MAST 7; ANT12\_19) along with an OTU assigned to Chrysophyceae-Synurophyceae Clade-H predominating at 45 m and 65 m in Station 1 (**Figures 5**, **6**). Another stramenopile OTU, assigned to Labyrinthulaceae, TABLE 1 | Environmental characteristics of the water samples collected from North-West Spitsbergen.


Samples from Station 1 (80.40◦ N, 5.60◦ E) and Station 2 (81.23◦ N, 9.30◦ E) were sampled on 16th and 19th of May, 2010, respectively. PW, Polar Waters; ASW, Arctic Surface Waters; AW, Atlantic Waters. No. of OTUs mentioned in the table are after subsampling. The OTU numbers are written for both DNA/RNA samples, where upper numbers are from DNA samples and lower numbers are from RNA samples. The estimated diversity indices values were also written in a similar way (DNA/RNA).

<sup>a</sup>Temperature ◦C.

<sup>b</sup>Salinity.

<sup>c</sup>Density.

<sup>d</sup>Chl a measured in µg L−1.

<sup>e</sup>µM; na: not analyzed.

bd, below detection level.

was only abundant in AW. An OTU assigned to the cryptophyte Teleaulax gracilis was abundant in PW at both stations, but had higher RNA read numbers at Station 1 (**Figure 5**). Picozoa were found at all depths at both stations, but were most abundant in the libraries from intermediate and deeper depths (**Figure 6**). An OTU assigned to the rhizarian Cryomonadida (OTU 20) was highly abundant in the surface DNA library from Station 1. Other abundant taxa in the deep AW included two OTUs assigned to the radiolarian Taxopodida B (**Table 2**, **Figures 5**, **6**).

represents neighbor joining clustering with bray-curtis distances. Bootstrap values was based on 999 permutations, and values below 60 are not shown. Colored branches in the dendrogram shows different watermasses (Red: AW; Orange: ASW; Blue: PW). Naming of samples: 1 or 2 states the sampling stations, D or R refer to the DNA or RNA samples and the number refers to sampling depth (i.e., 1R200- RNA sample from 200 m from Station 1). This is a modified image after combining the heatmap and clustering figures together to enhance the illustration.

#### DNA vs. RNA

The results of the UPGMA clustering indicated that among the deeper samples (120 and 200 m), libraries made from the same molecule type (DNA vs. RNA) were more similar than libraries from the same location or depth (**Figure 5**). In particular, reads assigned to the two Taxopodida B and the five abundant MALV OTUs were more abundant in the DNA libraries compared to the RNA libraries. This effect was, however, not apparent in the CCA analysis where libraries prepared from the same depth and station were most similar (**Figure 4**). A higher number of rare OTUs were identified from the DNA libraries compared to the RNA libraries, whereas the numbers of abundant OTUs were similar in the two molecule types (data not shown).

FIGURE 6 | Barchart of eukaryotic OTU (98% sequence similarity) composition in DNA and RNA samples from different depths at Stations 1 and 2. Taxonomy according to Adl et al. (2012). OTUs that did not give clear taxonomic information in BLAST and that had <1% reads in every sample were lumped together in "Others." The sample names used in this figure is similar to Figure 5.



Assignment represents best hit to the lowest taxonomic level using blastn with e-value 0.00001 against databases PR2 (Guillou et al., 2013) and CLJ (Lovejoy et al., 2016). Assignment to CLJ was used in cases having identical % identity for both databases.

# DISCUSSION

#### Surface Water Community Differentiation

The two stations showed limited sympagic growth underneath the melting sea ice (personal observation), and the depth of the salinity-reduced surface water (PW; 45 m at Station 1 and 90 m at Station 2; **Figure 3**) suggest that sea ice melting had been ongoing for a while, although typical sea ice algae as well as the dinoflagellate Polarella glacialis with resting cysts were identified in the bottom 3 cm of ice cores taken in the vicinity (Johnsen et al., submitted). The high Chl a concentrations along with reduced nutrient concentrations and the algal community composition suggest that Station 1 was in a peak bloom phase (cf. Kristiansen et al., 2000; Johnsen et al., submitted). Station 2, on the other hand, had an order of magnitude lower Chl a concentrations while nutrient levels were similar to those of Station 1 (**Table 1**). Although a permanent stratification with surface mixed layer nitrate values around 5 µM may exist in the region in the pre bloom phase (Randelhoff et al., 2016), the low Chl a and nutrient concentrations at Station 2 suggest that this area had already developed into a post-bloom stage. The more shallow euphotic zone at Station 1 was probably due to shading from the ongoing bloom (**Table 1**, **Figure 3**).

The pelagic bloom in Station 1 was dominated by large and typical spring-bloom species of diatoms and had probably advected into the area (Johnsen et al., submitted). Common spring bloom species of Arctic waters such as Phaeocystis sp. and Thalassiosira antarctica (von Quillfeldt, 1997; Degerlund and Eilertsen, 2010) were among the most abundant OTUs in our dataset (**Figures 5**, **6**), and were also abundant in phytoplankton samples from the area studied by microscopy (Johnsen et al., submitted). An advection of Phaeocystis sp. into ice-covered polar water in the Fram Strait was suggested also by Metfies et al. (2016), although a recent study report an under-ice bloom dominated by Phaeocystis sp. developing in situ (Assmy et al., 2017). Under-ice blooms may become more common in a scenario of thinner sea ice and a changing under-ice light climate (see also Arrigo et al., 2012). The clear dominance of photoautotrophs >3 µm in the advected bloom (Station1, **Table 1**) is similar to previous reports from Svalbard waters (Hodal et al., 2011; Marquardt et al., 2016). The OTU assigned to the cold water dinoflagellate Woloszynskia halophila, a bloomforming species of the Baltic Sea (Kremp et al., 2005), was highly abundant in the DNA library from the surface (0 m), while very few reads were found in the corresponding RNA library. As 18S sequence divergence is known to not always distinguish species of Dinophyceae (Logares et al., 2007), it is possible that the identified OTU represents another Suessiales species. Interestingly, Suessiales cysts (identified to Polarella glacialis; Johnsen et al., submitted) were found in sea ice cores as well as in the pelagic, and this OTU may represent cysts of a Suessiales taxon which is not metabolically active and thus not well represented in the RNA library.

Other potentially photosynthesizing plankton among the abundant OTUs include oligotrich ciliates (**Table 2**; **Figure 5**) where several representatives have sequestered chloroplasts that are retained and function as kleptoplastids, enabling photosynthesis by the host (reviewed by Esteban et al., 2010). In support, the very abundant Oligotrichia OTU had particularly high RNA read abundances at 0 and 10 m depth at Station 1, suggesting high metabolic activity of this taxon within the photic zone at this station. However, other Strombidiidae are heterotrophic, and we cannot exclude the possibility that the abundant oligotrichs are bacterivorous as suggested for the abundant Strombidinium in the Amundsen Gulf (Terrado et al., 2011).

The community composition of Station 2 was distinctly different from the advected bloom community identified in Station 1 (**Figures 4**, **6**), and although most of the abundant OTUs were present at both stations their relative abundances differed. The arctic clade of Micromonas sp. (Arctic Micromonas) recruited over 30% of the reads in both RNA and DNA libraries from PW in Station 2, in line with this cold-and shade adapted ecotype of Micromonas (Lovejoy et al., 2007) being important in blooms developing in colder water (see also Metfies et al., 2016). In a recent temporal study from Isfjorden, West Spitsbergen, Arctic Micromonas had especially high relative read abundances during pre- and post-bloom stages (Marquardt et al., 2016), in agreement with Station 2 being in a post-bloom (alternatively pre-bloom) stage. The dominance of Arctic Micromonas over Phaeocystis sp. in the colder station 2 suggests that development in this area, located further to the northeast, was less influenced by the large pelagic bloom advected under the ice at Station 1.

An OTU assigned to the cryptomonad Teleaulax gracilis was abundant in surface waters of both stations, supporting the developed stages of the blooms as cryptophytes often appear late in the bloom (Leu et al., 2006). RNA read counts suggested that the cryptomonad was especially active at Station 1, in agreement with the ongoing bloom. Fluorescence and PAR were the two environmental parameters identified to separate the two surface communities (**Figure 4**), suggesting that succession and light conditions were the main drivers of the community differentiation in the surface.

The samples collected around the halocline (45 and 65 m in Station 1) had increased abundances of OTUs assigned to the autotrophs Phaeocystis sp. and T. antarctica probably reflecting sinking of the huge bloom (**Figure 5**). The same pattern was seen for an OTU assigned to another bloom species, Chaetoceros socialis (data not shown), which also was abundant in microscopy counts from the vicinity (Johnsen et al., submitted). Abundant heterotrophic flagellates most likely feeding on the sinking bloom included the ciliates Choreotrichia-1 and Protostomatea, Picozoa, and MAST 7 (ANT12\_10; **Figure 5**). MAST 7 is more often found in and below the subsurface chlorophyll maximum layer compared to the upper mixed layer in the Arctic Ocean (Monier et al., 2013).

As commonly found in studies of picoplankton diversity based on filtration approaches, sequences from larger protists and metazoans were recovered, probably due to cell breakage and deformation of flexible walled cells allowing their DNA and RNA to pass through the 3 µm filters (Terrado et al., 2011; Sørensen et al., 2013). Another technical issue using 18S sequence data to compare protist communities is the potential for overestimation of certain OTUs due to high rDNA copy numbers or artifacts of the sequencing procedure (e.g., Zhu et al., 2005; Potvin and Lovejoy, 2009).

### Atlantic Water Communities Are Highly Diverse and Similar Despite Geographical Distance

The increased diversity and evenness found in the AW microbial communities compared to the PW surface samples (**Table 1**), suggest that these deep communities are both diverse and metabolically active. Also, the increased representation of rare taxa in AW libraries may represent a seeding stock that could function as a biological buffer to environmental change (Sogin et al., 2006; Pedros-Alio, 2012). The deep, dark and nutrientrich Atlantic current extends from the Norwegian Coastal Current, flowing in a 200–500 m deep layer. This deep AW layer may access the sunlit part by upwelling processes, as was shown to occur in January 2012 onto the shelf of northern Spitsbergen (Falk-Petersen et al., 2015). The AW harbors a distinct community of microbial protists brought northwards with the North Atlantic Current from more southern waters as well as Arctic species introduced through mixing with local water. The increased temperature identified in the waters of outer Isfjorden over the last decade (Pavlov et al., 2013) suggests that the organisms transported northwards with the West Spitsbergen Current may have increased survival, both in the deep and if they are transported to the surface. As argued by Pedros-Alio (2012), the rare biosphere may be composed of species whose individual requirements do not fit the current environment, but that may be able to bloom under different conditions.

There were distinct differences in the abundances of several OTUs occurring in the deeper AW between the DNA and RNA libraries (**Figure 5**). Most of the OTUs predominating in the AW (MALV, RAD, Labyrinthulaceae, Mesodiniidae) have heterotrophic or parasitic lifestyles, except the Mesodiinidae which have representatives at different trophic levels. Radiolarians are known to be associated with a large diversity of marine alveolates of which they are often hosts (Guillou et al., 2008; Bråte et al., 2012). The radiolarian OTU RAD B (Taxopodida; Suzuki and Not, 2015) and the five most abundant marine alveolate OTUs (**Figure 5**) were all abundant in the AW DNA libraries, but were not active (i.e., not abundant in the RNA libraries), as has previously been shown in other studies (Not et al., 2009; Hu et al., 2016; Marquardt et al., 2016). The limited activities shown by the parasitic protists (RAB B and MALV) in AW suggests that these taxa occur in less active life cycle forms.

Radiolaria are known to host a wide diversity of eukaryote symbionts, and were shown to harbor both the MALV I and MALV II lineages (Bråte et al., 2012). Whether the association between RAD B and MALV lineages in the deep AW is due to a parasite-host relationship is unknown, but a similar correlation of MALV sequence abundances with radiolarians was found also in samples from the Western Antarctic Peninsula with increased abundances in deep waters (Cleary and Durbin, 2016).

Other OTUs displayed depth or station related patterns in their RNA to DNA ratios, suggesting that their activity levels were modulated by changes in the environment. One such case was the very abundant dinoflagellate Dino\_Clone\_NIF\_3A1 that only had high activity levels in the deep, suggesting it to be a heterotrophic taxon (**Figure 5**).

All of the abundant phototrophic OTUs were detected in the AW, albeit at reduced read numbers. Interestingly Phaeocystis sp. was also abundant in the AW at Station 1, both in the DNA and RNA libraries, showing that the detected cells were alive. This finding is in agreement with previous reports on viable Phaeocystis sp. in the deep (Vader et al., 2014), suggesting that these haptophytes could have a mixotrophic lifestyle as was shown for Arctic Micromonas (McKie Krisberg and Sanders, 2014).

In conclusion, the microbial eukaryote communities of the two under-ice stations differed due to a combination of their water mass history and local processes, the homogenizing effect of water currents being less important in surface waters. Furthermore, the combined use of RNA and DNA libraries allow interpretation on the activity levels of the identified taxa. Thus, an extended use of RNA libraries in similar studies may improve our understanding of the active part of these communities.

# ACCESSION NUMBERS

The 454 sequencing data obtained in this study was submitted to the Sequence Read Archive (SRA) at NCBI GenBank (BioProject ID PRJNA384116).

# AUTHOR CONTRIBUTIONS

AM, AV, and TG participated during the field sampling. AM and AV did the molecular lab work, and SK analyzed the samples for nutrient concentrations. AM did the bioinformatic analyses with input from AV and TG. AM drafted the manuscript text, and all co-authors contributed to discussing the results and editing the manuscript.

# FUNDING

This project is funded by The University Centre in Svalbard, the Norwegian Research Council and the Northern Area Program of ConocoPhillips and Lundin Petroleum.

# ACKNOWLEDGMENTS

We would like to thank the crew of R/V Helmer Hanssen and the UNIS AB330 and AB323 students from 2010 for great support during field sampling. The help of Eva Falck to interpret the oceanographic data and of Geir Johnsen for sharing unpublished data from the sampling sites in the vicinity of our stations is greatly appreciated. This work was performed on the Abel Cluster (UiO) and the Norwegian metacenter for High Performance Computing (NOTUR), and operated by USIT, UiO (http://www.hpc.uio.no/). We acknowledge the help received from the University Centre for Information Technology at UiO. We would also like to acknowledge the use of the Lifeportal (lifeportal-help@usit.uio.no) for analyses.

#### REFERENCES


Esteban, G. F., Fenchel, T., Finlay, B. J. (2010). Mixotrophy in ciliates. Protist 161, 621–641. doi: 10.1016/j.protis.2010.08.002


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.01099/full#supplementary-material

marine parasitoids belonging to Syndiniales (Alveolata). Environ. Microbiol. 10, 3349–3365. doi: 10.1111/j.1462-2920.2008.01731.x


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Meshram, Vader, Kristiansen and Gabrielsen. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Atlantic Advection Driven Changes in Glacial Meltwater: Effects on Phytoplankton Chlorophyll-a and Taxonomic Composition in Kongsfjorden, Spitsbergen

Willem H. van De Poll <sup>1</sup> \*, Douwe S. Maat <sup>2</sup> , Philipp Fischer <sup>3</sup> , Patrick D. Rozema<sup>1</sup> , Oonagh B. Daly <sup>1</sup> , Sebastiaan Koppelle<sup>2</sup> , Ronald J. W. Visser <sup>1</sup> and Anita G. J. Buma1, 4

*<sup>1</sup> Department of Ocean Ecosystems, Energy and Sustainability Research Institute Groningen, University of Groningen, Groningen, Netherlands, <sup>2</sup> Department of Marine Microbiology and Biogeochemistry, NIOZ Royal Netherlands Institute for Sea Research, and Utrecht University, Den Burg, Netherlands, <sup>3</sup> Biosciences, Shelf Sea System Ecology, Alfred Wegener Institute, Helgoland, Germany, <sup>4</sup> Faculty of Arts, Arctic Centre, University of Groningen, Groningen, Netherlands*

#### Edited by:

*Connie Lovejoy, Laval University, Canada*

#### Reviewed by:

*Heather Bouman, University of Oxford, UK Kevin Arrigo, Stanford University, USA*

#### \*Correspondence:

*Willem H. van De Poll w.h.van.de.poll@rug.nl*

#### Specialty section:

*This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Marine Science*

Received: *26 July 2016* Accepted: *29 September 2016* Published: *18 October 2016*

#### Citation:

*van De Poll WH, Maat DS, Fischer P, Rozema PD, Daly OB, Koppelle S, Visser RJW and Buma AGJ (2016) Atlantic Advection Driven Changes in Glacial Meltwater: Effects on Phytoplankton Chlorophyll-a and Taxonomic Composition in Kongsfjorden, Spitsbergen. Front. Mar. Sci. 3:200. doi: 10.3389/fmars.2016.00200* Phytoplankton biomass and composition was investigated in a high Arctic fjord (Kongsfjorden, 79◦N, 11◦40′E) using year round weekly pigment samples collected from October 2013 to December 2014. In addition, phytoplankton dynamics supplemented with physical and chemical characteristics of the 2014 spring bloom (April–June 2014) were assessed in two locations in Kongsfjorden. The goal was to elucidate effects of Atlantic advection on spatial phytoplankton chlorophyll-a (chl-a) and taxonomic composition. Chl-a declined during the polar night to a minimum of 0.01 mg m−<sup>3</sup> , followed by a 1000-fold increase until May 28. Atlantic advection prevented sea ice formation and increased springtime melting of marine terminating glaciers. This coincided with spatial and temporal differences in abundances of flagellates (prasinophytes, haptophytes, cryptophytes, and chrysophytes) and diatoms in early spring. More flagellated phytoplankton were observed in the non-stratified central Kongsfjorden, whereas diatoms were more abundant in the stratified inner fjord. Contrasting conditions between locations were reduced when glacial melt water stratification expanded toward the mouth of the fjord, mediating a diatom dominated surface bloom at both locations. We suggest that glacial melt water governs spring bloom spatial timing and composition in the absence of sea ice driven stratification. The spring bloom exhausted surface nutrient concentrations by the end of May. The nutrient limited post bloom period (June–October) was characterized by reduced biomass and pigments of flagellated phytoplankton, consisting of prasinophytes, haptophytes, chrysophytes, and to a lesser extent cryptophytes and peridinin-containing dinoflagellates.

Keywords: Arctic phytoplankton, pigments, taxonomic composition, Atlantic advection, Kongsfjorden, stratification, glacial melt water, seasonal cycle

# INTRODUCTION

Phytoplankton dynamics in the coastal Arctic are shaped by extreme seasonality in irradiance. During the polar night phytoplankton experience a prolonged period of darkness. The return of light marks the start of the spring bloom (Berge et al., 2015). Arctic sea ice cover in spring and summer months has declined for several decades, also around the Svalbard archipelago (Stroeve et al., 2007). Reduced sea ice cover enhances phytoplankton irradiance exposure and pelagic phytoplankton productivity in vast parts of the Arctic Ocean (Arrigo and van Dijken, 2015). Density differences due to melting sea ice or glacial meltwater stabilize the water column, allowing phytoplankton to maintain their position and form highly productive surface blooms. Stratification is associated with diatom blooms in the marginal ice zones (Syvertsen, 1991; Perrette et al., 2011). However, strong stratification can also lead to nutrient limitation and changes phytoplankton taxonomic composition and biomass (Li et al., 2009). In the absence of stratification, convection and strong winds can mediate deep turbulent mixing of the water column that can reduce phytoplankton irradiance exposure and reduce phytoplankton growth (Townsend et al., 1994).

The Greenland and Barents Sea experience inflow of Atlantic water with variable heat content (Lien et al., 2013). Kongsfjorden (79◦N, 11◦ 40′E) and the West coast of Spitsbergen are influenced by warm saline Atlantic water of the West Spitsbergen Current (WSC) and colder less saline Arctic water of the East Spitsbergen Current (ESC) that mix on the continental shelf (Cottier et al., 2005). As a result, fjords on the West coast of Spitsbergen can experience highly variable sea water temperature conditions. The temperature of the WSC showed an increasing trend from 1997 to 2010 (Beszczynska-Möller et al., 2012). Recent increases in advection of Atlantic water in Kongsfjorden have been attributed to changes in density by warming of the WSC, reduced drift ice in the ESC, and changes in wind direction (Cottier et al., 2007, 2010).

Advection of warm Atlantic water in winter and early spring can prevent sea ice formation in Kongsfjorden. The changing conditions in Kongsfjorden have been suggested to affect phytoplankton productivity and taxonomic composition (Hegseth and Tverberg, 2013; Kubiszyn et al., 2014). These changes may arise from multiple causes. Warmer Atlantic water may harbor different phytoplankton species and a more boreal grazer community (Willis et al., 2008). Furthermore, transition from an ice covered to an open fjord in winter can change the starting population by affecting growth conditions of pelagic phytoplankton. An open fjord in spring can promote convective mixing, but advection of Atlantic water at the surface can also reduce the depth of convective mixing (Hegseth and Tverberg, 2013). Moreover, increasing sea water temperature promotes glacial melting, which affects the strength and timing of stratification as well as turbidity (Hop et al., 2002; Luckman et al., 2015; Bartsch et al., 2016). In the absence of melting sea ice, melt water from marine terminating glaciers is the dominant source of fresh water in Kongsfjorden (MacLachlan et al., 2007), producing a steep fresh water gradient during spring and summer (Piquet et al., 2014).

Kongsfjorden spring blooms show considerable variability in timing and composition, typically consisting of diatoms and Phaeocystis pouchetii, whereas small flagellates dominate during the nutrient limited post bloom period (Iversen and Seuthe, 2011; Hegseth and Tverberg, 2013; Piquet et al., 2014). Phytoplankton abundance and composition affect carbon fluxes to pelagic and benthic higher trophic levels (Bhatt et al., 2014). In addition, this eventually affects carbon storage in the ocean. The present study aims to improve our understanding of the processes that affect phytoplankton biomass and composition in Kongsfjorden. We investigated an annual time series of phytoplankton pigment samples from the 2013 to 2014 polar night, spring bloom, and following post bloom. Furthermore, we investigated effects of Atlantic water advection on stratification, phytoplankton biomass, and composition during the spring bloom from April to June (2014) by sampling two stations in the glacial meltwater gradient. The goal was to understand the effects of Atlantic water advection on stratification and phytoplankton biomass and composition dynamics. We specifically address the hypothesis that in the absence of sea ice, glacial melt water is the dominant factor that defines the timing and composition of the phytoplankton spring bloom.

# METHODS

The research was conducted at 3 locations in Kongsfjorden, Spitsbergen (79◦N, 11◦ 40′E), covering the period from October 2013 to December 2014. Samples were collected at the ferry box of the AWIPEV Underwater Fjord Observatory in Ny Ålesund (Fischer et al., 2016) that collects water from 11 m depth (100 m off shore) in Kongsfjorden (**Figure 1**). Samples were taken over weekly intervals, whereas sampling frequency was increased from April to June ranging from daily to twice a week, with 24 samples being collected during this period. Seawater samples (4–8 L) were filtered on 47 mm GF/F (Whatman) using 0.2 bar overpressure, snap frozen in liquid nitrogen, and stored at −80◦C. The samples were analyzed for phytoplankton pigments by High Performance Liquid Chromatography (HPLC) as described below. During the 2013–2014 polar night and following spring bloom (October 2013–June 2014), temperature from 11 m depth was obtained from a SBE38 temperature sensor (Sea Bird Electronics) from which daily averages were calculated. Average daily air temperature (2 m) was calculated from the BSRN measurement platform in Ny Ålesund.

In addition, regular sampling (19–28 times) was performed by boat from April to June 2014 at two stations in Kongsfjorden: "Glacier station" (G), located in front of the Kronebreen glacier (inner Kongsfjorden) and "Midfjord station" (M), located in central Kongsfjorden (**Figure 1**). Vertical profiles of the water column were collected at station G (50 m) and M (100 m) using a CTD (SBE 19 plus, Sea-Bird Electronics) equipped with sensors for PAR (Licor, Sea-Bird Electronics), fluorescence (Wetstar), and turbidity (ECO NTU, Wetlab). From these profiles, salinity, temperature, potential density, and turbidity at 5, 25, and 50 m depth were extracted. Potential density differences in excess of 0.005 kg m−<sup>3</sup> between surface (5 m) and 50 m depth

were considered indicative for stratification (Kara et al., 2000). Irradiance attenuation (Kd) was calculated by linear regression of log transformed PAR data. The euphotic zone (defined as the depth interval down to 0.1% irradiance) was calculated as: Zeu (m) = − ln (0.001)/Kd.

Samples were collected at 5, 25, and 50 m depth at both stations using a 24 L Niskin bottle. The samples were kept cold and stored in darkness until processing in the lab. Pigment samples were obtained by mild vacuum filtration (0.2 bar) of 4–10 L seawater on 47 mm GF/F (Whatman) filters. Afterwards, filters were snap frozen in liquid nitrogen and stored at −80◦C until HPLC pigment analysis. Five mL subsamples were filtered (0.2µm Acrodisc, Pall) for nutrient analysis and frozen at −20◦C (nitrate and nitrite, phosphate, ammonium) or stored at 4◦C (silicate) until analysis using a Bran and Luebbe QuAAtro auto analyzer to determine dissolved inorganic phosphate, ammonium, nitrate, and nitrite, and silicate at the NIOZ, The Netherlands.

Subsamples for microscopy (100 mL) were fixed with 1.5 mL Lugol's iodine solution and stored dark at 4◦C until analysis using an Olympus IMT-2 inverted microscope (Utermöhl technique, Edler and Elbrächter, 2010). This procedure was conducted at 20 samples collected between April 26 and June 10, 2014 at station M (n = 10) and G (n = 10) to observe the general phytoplankton composition (presence of diatoms, flagellates, Phaeocystis, dinoflagellates, cryptophytes). The phytoplankton was not enumerated in these samples and the results are not shown. We report phytoplankton abundance and the abundance of large heterotrophic dinoflagellates and ciliates of 4 samples collected at 5 m depth (station M: May 5, 10, and 28, station G: May 28).

#### PHYTOPLANKTON PIGMENT AND CHEMTAX ANALYSIS

Filters were freeze dried for 48 h and pigments were extracted using 90% acetone (v/v) for 48 h (4◦C, darkness). Pigments were separated by HPLC (Waters 2695) with a Zorbax Eclipse XDB-C8 column (3.5µm particle size), using the method of Van Heukelem and Thomas (2001), modified after Perl (2009). Detection was based on retention time and diode array spectroscopy (Waters 996) at 436 nm. Pigments were manually quantified using standards for all used pigments (DHI lab products). The absolute and relative abundances of phytoplankton groups were assessed by CHEMTAX analysis of pigments, using the steepest descent algorithm (version 1.95) (Mackey et al., 1996). Pigments were partitioned among 7 groups (based on microscopic observations): diatoms (fucoxanthin), prasinophytes (chlorophyll b, neoxanthin, prasinoxanthin), haptophytes (19′butanoloxyfucoxanthin, fucoxanthin, 19′hexanoloxyfucoxanthin, chlorophyll c3), pelagophytes (19′butanoloxyfucoxanthin, fucoxanthin, chlorophyll c3), chrysophytes (19′butanoloxyfucoxanthin, fucoxanthin, 19′hexanoloxyfucoxanthin, chlorophyll c3), dinoflagellates (peridinin), and cryptophytes (alloxanthin). CHEMTAX results of the haptophyte and pelagophyte subgroups were pooled (these groups showed similar dynamics) and will be denoted as "haptophytes." Ferry box samples with chl-a below 0.01 mg m−<sup>3</sup> were excluded for CHEMTAX analysis because many accessory pigments became undetectable in these dilute samples. Ferry box samples collected during the nutrient replete spring bloom (n = 40) and those collected during the nutrient depleted post bloom (n = 24) were grouped in separate bins. Non-stratified samples of station M (5, 25, and 50 m; n = 24) were grouped in a single bin. Stratified samples of M and G were analyzed in two depth bins: 5 and 25 m, (n = 52), and 50 m (n = 26). We used identical low light acclimated initial pigment ratios for all bins (Supplement Table 1A). All pigments were allowed to vary during CHEMTAX analysis (chl-a: 100%, other pigments: 500%). Final pigment ratios are shown in Supplement Table 1B.

#### DATA ANALYSIS

Analysis of the ferry box data series (October 2013–December 2014) and the data series of stations M and G (April–June 2014) were performed separately. Annual distribution of 2014 phytoplankton chl-a and taxonomic composition was assessed for weekly averaged ferry box samples (chl-a n = 81, CHEMTAX n = 64). Exponential growth rates of chl-a during the nutrient replete spring bloom were calculated by fitting log transformed chl-a vs. time with a linear function.

During the spring bloom we used linear correlation (Pearson Product Moment) to investigate relationships between environmental data and non-linear correlation (Spearman Rank Order) when investigating relationships between environmental and biological data for dates with a complete biological and physical data set of stations M and G. Correlations were considered significant at p < 0.05.

#### RESULTS

#### Annual Cycle 2014 (Ferry Box)

The underwater observatory recorded declining sea water temperatures from October 2013 (maximum 5.60◦C) to the beginning of January 2014 (minimum 0.67◦C, **Figure 2A**). After January 16 temperature increased to 3.36◦C (first inflow event, **Figure 2A**). This was preceded by elevated air temperature. Sea water and air temperature declined from the end of February to the beginning of April (minimum 0.32 and −16.55◦C, respectively), and increased from April onwards (second inflow event, **Figure 2A**). Chl-a concentrations declined by two orders of magnitude during the polar night from October 20 to February 23, 2014 (**Figure 2B**). The 2014 ferry box time series showed average concentrations of 0.023 ± 0.018 mg chl-a m−<sup>3</sup> during the polar night (samples from October 20 to December 27, 2014, and from October 20, 2013 to February 20, 2014, n = 21). Increasing chl-a coincided with the return of light, light dose correlated linearly with chl-a from February 20 to April 12 (r<sup>s</sup> = 0.97, n = 10). Peak concentrations of up to 4.9 mg m−<sup>3</sup> were observed on June 5. This coincided with depletion of nitrate, dissolved inorganic phosphate, and silicate (**Figure 3A**, Supplement Figure 1). The non-stratified period (February 20 to May 2, see below) showed lower chl-a based exponential growth rates (**Table 1**) and overall lower chl-a concentrations (average 0.15 ± 0.90 mg m−<sup>3</sup> )

FIGURE 2 | (A) Daily averaged sea water temperature at the ferry box inlet (11 m) from October 2013 to June 2014 (left y axis) and air temperature in Ny Ålesund at 2 m. The numbers 1 and 2 indicate Atlantic advection events. (B) Ferry box chlorophyll-a concentration and daily PAR dose (photosynthetically active radiation, 400–700 nm) in Ny Ålesund from October 2013 to June 2014.

as compared to the stratified spring bloom period (May 5–June 5, average 2.4 ± 2.0 mg m−<sup>3</sup> ).The nutrient depleted post bloom period (June 9–October 20) showed lower average chl-a (0.60 ± 0.36 mg m−<sup>3</sup> ). Chl-a increased from July 14 to August 25 (average 0.79 ± 0.17), before declining during the polar night.

CHEMTAX pigment analysis revealed shifts in phytoplankton taxonomic groups during the spring bloom and the following nutrient depleted post bloom (**Figure 3**). Diatom and chrysophyte chl-a peaked on June 5, representing 68 and 26% of the spring bloom chl-a concentration peak at the ferry box, respectively. Prasinophyte chl-a peaked prior to the spring bloom peak (May 6) and during the post bloom (July 31) period at a maximum relative abundance of 41% of chl-a. The post bloom period (June 9–October 20) showed a decline in relative abundance of diatoms and chrysophytes, averaging of 25 and 26% of chl-a in June and July. Peridinin-containing dinoflagellates and cryptophyte chl-a peaked in July, at a maximum of 14 and 18% of chl-a, respectively. Chrysophytes increased to 40% of chl-a in August, whereas haptophytes peaked in early September (maximal 40% of chl-a). Flagellated photosynthetic phytoplankton declined in relative abundance during the polar night, with diatoms increasing in relative abundance (>90% of chl-a).

#### SPRING 2014 (STATIONS M AND G)

#### Physicochemical Data

Marked differences in salinity and temperature were observed over time and at depth between stations M and G during the

analysis. The vertical dashed line marks the transition from non-stratified to stratified conditions. The shaded area indicates the polar night.



*R* 2 *is shown in brackets, n shows the number of data points fitted by the exponential function. Data were collected from April to June 2014.*

spring of 2014 (**Figure 4**). Salinity at 25 and 50 m of station M (35.08 ± 0.01) was higher than at station G (34.94 ± 0.05). Potential density at station M of these depths was higher than at station G up to May 5. Temperature at station M (5 m average: 2.10 ± 0.70◦C) was higher than at station G (5 m average 0.54 ± 0.99◦C), and increased from April 11 to 14, and from May 5 onwards (**Figure 4**). Water column temperature at station G increased from April to June. Surface temperature (5 m) at M (May 5–May 28) and G (April 14–May 28) to was typically lower compared with 50 m.

The temperature and salinity increase at station M from April 11 to 14 was followed by a period with minimal potential density differences (no stratification) between 5 and 50 m up to May 2. During this period temperature and salinity showed similar and significant positive correlations at 5, 25 and 50 m (ρ = 0.97, 0.96, 0.95). After stratification of station M due to decreasing surface salinity (May 5 and onwards), this relationship was not observed at 5 m, and became weaker at 25 and 50 m. Station G was always stratified from April 14 to June 10 (**Figure 5**). Stratification strength correlated inversely with surface salinity (ρ = −0.96, −0.95 for G and M, respectively). Stratification strength increased with surface temperature at station G and M (ρ = 0.93, 0.64, respectively). In May, stratification strength increased at M, and differences between M and G became minimal by the end of May. At station G an inverse correlation was found between temperature and salinity at 5 m (ρ = −0.78), whereas positive correlations were observed at 25 and 50 m (ρ = 0.96, 0.94) up to May 10.

Irradiance attenuation was strong at station G, resulting in a euphotic zone of 23 ± 5 m that showed little change over time (**Figure 5**). The euphotic zone of station M was on average 70 ± 19 m up to May 20, and declined to 24 ± 7 m afterwards. Surface turbidity correlated positively with the attenuation coefficient K<sup>d</sup> of M and G (ρ = 0.65). Between April and June, turbidity in the upper 25 m was on average 0.53 ± 0.37 at station M whereas this was 2.43 ± 0.91 for station G (not shown).

Surface (5 m) nitrate (10.78 ± 0.37µM), phosphate (0.65 ± 0.06µM, Supplement Figure 2) and silicate (4.75 ± 0.05µM) concentrations were high from April to early May (**Figure 5**), and showed little variability with depth (Supplement Figure 2). Surface (5 m) nutrient concentrations declined steeply during the second half of May, and were close to the detection limits by the end of May at stations G and M (average phosphate: 0.019, nitrate:0.06, and silicate: 0.08 µM). In June, average surface N:P

(nitrate: phosphate) ratios were 3.6 and 2.4 for stations M and G, respectively. Prior to May 7, N:P ratios were on average 16.72 ± 0.89 and 18.02 ± 2.25 at M and G, respectively. Ammonium increased at station M and G from 0.15 ± 0.07 to 0.34 ± 0.06µM at 5 m from April 14 to June 10. Ammonium increased to 1.35 ± 0.28µM at 25 and 50 m depth (Supplement Figure 2).

# Phytoplankton Chlorophyll-a

Chlorophyll-a concentrations at stations M and G were low (0.026 mg m−<sup>3</sup> ) in early April (**Figure 5**). Chl-a at station M was uniformly distributed over the water column from April 9 to May 7, and increased in the upper 25 m in May. Chl-a (5 m) was on average 28% higher at M compared to G during the non-stratified period (April 9–May 2), whereas it was on average (26%) lower than that at G after stratification (May 5–May 28). Surface chl-a (5 m) showed an exponential increase over time, peaking on May 28 (>10 mg m−<sup>3</sup> ) at both stations (**Figure 5**). Chl-a based growth rates were higher at station G as compared to station M (**Table 1**). Prior to stratification of station M (April 11–May 2) the growth rate was ∼50% lower than at station G. Chl-a based growth at stations G and M was similar after stratification. Surface chl-a declined to 0.6 mg m−<sup>3</sup> in June.

#### Phytoplankton Composition

Taxonomic composition at station M and G showed differences in early spring (April–May, **Figure 6**). At station M diatoms were on average 31 ± 18% of chl-a from April 14 to May 16, and increased to 72 ± 14% on May 28. Diatoms comprised a high fraction of chl-a at station G between April and June (average 5 m: 71 ± 13%, **Figure 6**). Absolute diatom chl-a was on average 2.2-fold higher at G than at M during April– May 15 (**Figure 6**). During the same period flagellates (chla of haptophytes, prasinophytes, cryptophytes, dinoflagellates, and chrysophytes, combined) were 2.3-fold higher at station M than at G (**Figure 6**). During the stratified period, differences in absolute flagellates chl-a were 10% between M and G. Haptophyte relative abundance at station M declined from 34 ± 2.1% (April 14) to 4.0 ± 4.5% of chl-a (May 28). At station G haptophytes were 17 ± 4.1% of chl-a up to May 10 and declined to 4.0 ± 4.1% afterwards. Relative abundance of prasinophytes at M increased from April 9 (4%) to May 2 (34 ± 1%), and declined toward the end of May (**Figure 6**). At station G prasinophytes increased to 17 ± 1.6% of chl-a on May 10. Cryptophytes increased from April to May 16 from to 5.9% at station M, but were < 1% of chl-a at station G. The contribution of chrysophytes to chl-a was variable, ranging from 0 to 30% and

0 to 16% at station M and G, respectively. Peridinin-containing dinoflagellates were on average 1.6 and 0.4% of chl-a at M and G respectively (not shown). The spring bloom chl-a peak (May 28) was dominated by diatoms (72 and 85% of chl-a at M and G, respectively).

Relative abundances of haptophytes, diatoms, prasinophytes, and cryptophytes showed the strongest correlations with physical variables (**Table 2**). Relative abundance of haptophytes, prasinophytes, cryptophytes at station M and G (5, 25, and 50 m) showed positive correlations with surface salinity (r<sup>s</sup> = 0.76, 0.66, 0.59) and euphotic zone depth (r<sup>s</sup> = 0.76, 0.55, 0.54). Surface density showed positive correlations with haptophytes and prasinophytes (r<sup>s</sup> = 0.72, 0.65). Relative abundance of haptophytes, prasinophytes, and cryptophytes showed an inverse correlation with stratification strength (r<sup>s</sup> = −0.75, −0.58, −0.45). Relative abundance of diatoms at station M and G (5, 25, and 50 m) showed inverse correlations with surface salinity (r<sup>s</sup> = −0.77), surface density (r<sup>s</sup> = −0.67), and euphotic zone depth (r<sup>s</sup> = −0.73). Relative abundance of diatoms showed a positive correlation with stratification strength (r<sup>s</sup> = 0.66). Relative abundance of dinoflagellates correlated significantly with temperature (r<sup>s</sup> = 0.70), and surface temperature (r<sup>s</sup> = 0.67). All reported rs-value were significant at p < 0.05.

#### Microscopy (Stations M and G)

Light microscopy revealed small (∼2µm) and larger (5–10 µm) flagellates dominating station M during the first half of May (**Table 3**). The larger flagellates (10 µm) were most likely Dictyocha speculum (Chrysophyceae). Small flagellates (∼2µm) were most likely Micromonassp. (Prasinophyceae). Diatoms were rare in these samples. However, the peak of the bloom (May 28) was dominated by diatoms (mostly Chaetoceros sp., and Thalassiosira sp.) at stations M and G. At station M significant numbers of small flagellates were still observed, whereas these were less abundant at station G. Tintinnids (Ciliates) increased at station M, whereby concentrations were 8-fold higher compared to G during the peak of the bloom. Large heterotrophic dinoflagellates were also higher at M as compared to G.

#### DISCUSSION

Atlantic advection was observed during 2013–2014 polar night (underwater observatory) and spring (underwater observatory, CTD) in central Kongsfjorden. Atlantic water interrupted winter cooling of the fjord, with a temperature of 3◦C at 11 m depth detected as early as January. Together with relatively high air temperature this prevented sea ice formation in 2014. Atlantic advection was episodic in early spring, followed by a period of cooling. During the second Atlantic advection event in April–May a salinity and temperature gradient was observed in the fjord, expanding from central to inner Kongsfjorden over time. Similar patterns were reported for Adventfjorden in 2014 (Wiedmann et al., 2016). The density of the intruding Atlantic water was higher than that of the inner fjord, resulting in advection at depth. This increased the temperature of the inner fjord, thereby promoting melting of marine terminating glaciers from below. The Atlantic water was the only heat source available during the 2014 spring bloom. The seasonally increasing contribution of air temperature and downward radiation became apparent by warming of the stratified surface layer in June. The

stations M and G from April to June 2014.

fresh water influx of glacial melting induced stratification in inner Kongsfjorden. Stratification strength near the glacier correlated with water temperature, and was comparable in magnitude to that reported during sea ice mediated stratification in central Kongsfjorden (Hodal et al., 2012). Melt water driven stratification expanded from the inner fjord toward central Kongsfjorden in early May. Prior to stratification central Kongsfjorden showed deep convective mixing (April–May 2) due to surface cooling and strong wind, with surface nutrient concentrations close to the maximal concentrations reported for Kongsfjorden after winter (Hop et al., 2002; Piquet et al., 2014). The 2014 spring bloom peak in central Kongsfjorden occurred after glacial melt water driven stratification. The spring bloom peak was late compared to recorded Kongsfjorden spring bloom phenology from the last decade, which can be as early as April (Hegseth and Tverberg, 2013). Atlantic advection created a spatial and temporal gradient of stratification and light availability in the fjord in April and May. The glacial melting in the stratified inner fjord coincided with sediment discharge, resulting in a high turbidity and a shallow euphotic zone during spring bloom formation.

Chl-a based growth rates showed exponential growth from April to the end of May in the inner fjord. These growth rates were two times higher compared to those from central Kongsfjorden (April–May 16). Despite high water transparency in central Kongsfjorden, vertical mixing may have reduced phytoplankton light exposure due to mixing below the euphotic zone. This may have delayed the development of a surface bloom in central Kongsfjorden, whereas light limitation from glacial sediment influx may have reduced growth in the inner fjord. Furthermore, phytoplankton composition was different between the two locations during early spring. Early spring phytoplankton in central Kongsfjorden consisted mostly of flagellated phytoplankton (prasinophytes, haptophytes, cryptophytes, chrysophytes), whereas the inner fjord phytoplankton was dominated by diatoms. Therefore, the growth rates may reflect differences in community composition between the two locations. Previous research indicated that years with Atlantic advection during the spring bloom were characterized by increased abundance of the haptophyte P. pouchetii (Hegseth and Tverberg, 2013). Haptophyte pigments and Phaeocystis cells or colonies were not abundant during the 2014 spring bloom. Nevertheless, inverse correlations with stratification strength, and positive correlations with

TABLE 2 | Correlation coefficients of Spearman rank order correlation between relative abundance of prasinophytes, dinoflagellates, cryptophytes, chrysophytes, diatoms, and haptophytes at station M and station G (5, 25, 50 m) between April 14 and June 10, and stratification strength, euphotic zone depth, surface salinity, temperature, and potential density (5 m), and salinity, temperature, and potential density at actual depth (a); (n = 102).


*All shown correlations were significant.*

surface salinity, and euphotic zone depth suggest a link between haptophytes and the Atlantic water influenced central Kongsfjorden. Prasinophytes and cryptophytes showed similar links to Atlantic water characteristics. Vader et al. (2015) provided evidence for the continuous presence of Micromonas pusilla and P. pouchetii in the Arctic and Atlantic water surrounding Spitsbergen during the polar night. In contrast, relative abundance of diatoms showed an inverse relationship with surface salinity, euphotic zone depth, and a positive correlation with stratification strength. These relationships reflect the spatial distribution of the phytoplankton taxonomic groups and water masses (possibly due to introduction of haptophytes by Atlantic advection and diatoms of local origin) rather than the conditions that cause these differences. Overall, these relationships also point to the importance of stratification in shaping phytoplankton composition during the spring bloom. How stratification changed phytoplankton species composition remains unknown. Stratification at central Kongsfjorden did not cause immediate changes in the phytoplankton composition and chl-a based growth rates. Pronounced changes were observed 10 days after stratification of station M, and appeared to be caused by expansion of the low salinity surface layer with a diatom bloom that was initiated at the inner fjord. Apart from influencing phytoplankton light exposure and nutrient concentrations, stratification also affects concentration driven processes such as top down control (Behrenfeld, 2010). Chla based growth rates during the spring bloom were roughly 2–4 times lower as compared to those from cultured Arctic phytoplankton species, suggesting significant losses (due to grazing or viral lysis). Ciliates (30µm) and large dinoflagellates were the most abundant grazers in our samples and were higher at station M compared to G during the peak of the bloom. These groups were previously shown to have a high potential to control small phytoplankton in Kongsfjorden (Seuthe et al., 2011). In addition, diatom resting stages can be suspended in the water column by deep mixing (Hegseth and Tverberg, 2013). Therefore, multiple factors influenced by stratification can potentially affect phytoplankton composition during the spring bloom.


TABLE 3 | Cell counts (cells l−<sup>1</sup> ) of dominant phytoplankton and ciliates of four samples of station M (3) and G (1) in Kongsfjorden obtained by light microscopy.

*Diatoms were typically organized in long chains.*

We used pigments and CHEMTAX to assess changes in phytoplankton taxonomic composition. Light microscopy was used to select the taxonomic groups included in CHEMTAX and to verify the observed patterns such as the difference in abundance of diatoms at station M and G. Dicyocha speculum (Chrysophyceae) flagellates were occasionally observed in considerable numbers in microscopy samples from M and G during the 2014 spring bloom. Diatoms and chrysophytes are difficult to distinguish using CHEMTAX because both groups share high fucoxanthin to chl-a ratios, a feature that can also be observed in some haptophyte groups (Higgins et al., 2011). In addition chrysophytes share pigments with haptophytes. Furthermore, the presence of additional taxonomic groups in the 2014 ferry box data cannot be excluded as there were no microscopy samples of the ferry box time series. Therefore, the presented CHEMTAX based composition should be viewed as a crude estimates. Apart from errors related to CHEMTAX, microscopic identification is difficult for small flagellated phytoplankton groups. Moreover, it is difficult to distinguish between pigmented and non-pigmented dinoflagellates in Lugol's iodine fixed samples. Nevertheless, CHEMTAX calculations matched microscopy observation from station M and G reasonably well. This suggests that pigment samples provided useful taxonomic information of Kongsfjorden phytoplankton.

The ferry box time series identified spring as the most dynamic part of the season with chl-a concentration increasing by 3 orders of magnitude after the polar night. As expected, chla declined during the polar night to very low concentrations (< 0.01 mg m−<sup>3</sup> ). This has consequences for the spring bloom and organisms that rely on phytoplankton for survival. Chla increased when light returned in February. We observed no lag phase in the chl-a response to irradiance as was suggested for Atlantic phytoplankton development in deeply mixed waters (Mignot et al., 2015). Chl-a increased exponentially at all locations from April to the peak concentrations on May 28 and June 5. This coincided with depletion of surface nutrients. The low N:P ratios (dissolved nutrients) suggested nitrate limitation during the post bloom period in June. In June, surface ammonium concentration was higher than the nitrate concentration, signaling the transition from new (nitrate based) production to regenerated phytoplankton production. The average chl-a concentration was higher during the nutrient replete spring bloom as compared to the nutrient limited post bloom period. A considerable part (up to 40% of chl-a) of the 2014 phytoplankton consisted of prasinophytes (Micromonas sp.) in the pico (∼ 2µm) phytoplankton range, both during the early spring bloom and post bloom period. Our limited observations suggested that spatial variability was lower after stratification. Phytoplankton during the post bloom consisted mostly of diverse

#### REFERENCES

Arrigo, K. R., and van Dijken, G. (2015). Continued increases in Arctic Ocean primary production. Prog. Oceanogr. 136, 60–70. doi: 10.1016/j.pocean.2015.05.002

flagellated phytoplankton. Early spring showed strong variability on a small spatial scale, coinciding with spatial differences in phytoplankton composition (diatoms vs. flagellates) and growth rates, and water masses.

Advection of Atlantic water modified the hydrographical conditions in the fjord that shaped the 2014 spring bloom by influencing stratification. This process was crucial for the development of the spring bloom biomass peak. The 1 month period following stratification showed Chl-a concentrations that were on average 5 times higher than the annual average over 2014. Advection of Atlantic water, and colder ESC water as well as the temperature of these water masses control important aspects that influence the spring bloom in Kongsfjorden. The properties of these water masses are likely to change with ongoing climate change, thereby influencing the interaction with the marine terminating glaciers in Kongsfjorden. Continued monitoring of pigments combined with monitoring of hydrographical conditions can increase our understanding of the inter-annual variability of Kongsfjorden phytoplankton biomass and composition and the controlling factors.

#### AUTHOR CONTRIBUTIONS

WV wrote the main manuscript, did the pigment analysis, and conducted field work in Ny Ålesund. DM contributed to field work in Ny Ålesund, and provided feedback on the manuscript. PF provided data from the AWIPEV Underwater Fjord Observatory (Ny Ålesund), and feedback on the manuscript. PR provided feedback and helped with the statistics. OD and SK contributed to field work in Ny Ålesund and provided feedback. RV contributed to field work in Ny Ålesund and developed the pigment sampling equipment. AB contributed to the writing of this manuscript.

#### ACKNOWLEDGMENTS

We express our strong thanks to the AWIPEV technicians for collecting ferry box samples for this manuscript and for facilitating our research at AWIPEV base. Furthermore, we thank the Kingsbay Marine lab staff for their assistance. Special thanks to Loes A. H. Venekamp for the light microscopy analysis and Sean de Graaf for assistance with the HPLC. This is a contribution to NWO project 866.12.408.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00200

Bartsch, I., Paar, M., Fredriksen, S., Schwanitz, M., Daniel, C., Hop, H., et al. (2016). Changes in kelp forest biomass and depth distribution in Kongsfjorden, Svalbard, between 1996–1998 and 2012–2014 reflect Arctic warming. Polar Biol. doi: 10.1007/s00300-015-1870-1. [Epub ahead of print].


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 van De Poll, Maat, Fischer, Rozema, Daly, Koppelle, Visser and Buma. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# *Synechococcus* in the Atlantic Gateway to the Arctic Ocean

Maria L. Paulsen<sup>1</sup> \*, Hugo Doré<sup>2</sup> , Laurence Garczarek <sup>2</sup> , Lena Seuthe<sup>3</sup> , Oliver Müller <sup>1</sup> , Ruth-Anne Sandaa<sup>1</sup> , Gunnar Bratbak <sup>1</sup> and Aud Larsen<sup>4</sup>

<sup>1</sup> Department of Biology, University of Bergen, Bergen, Norway, <sup>2</sup> Centre National de la Recherche Scientifique, UMR 7144, Station Biologique, Sorbonne Universités, UPMC Université Paris 06, Roscoff, France, <sup>3</sup> Department of Arctic and Marine Biology, Faculty of Biosciences, Fisheries and Economics, UiT-The Arctic University of Norway, Tromsø, Norway, <sup>4</sup> Hjort Centre for Marine Ecosystem Dynamics, Uni Research Environment, Bergen, Norway

Increasing temperatures, with pronounced effects at high latitudes, have raised questions about potential changes in species composition, as well as possible increased importance of small-celled phytoplankton in marine systems. In this study, we mapped out one of the smallest and globally most widespread primary producers, the picocyanobacterium Synechococcus, within the Atlantic inflow to the Arctic Ocean. In contrast to the general understanding that Synechococcus is almost absent in polar oceans due to low temperatures, we encountered high abundances (up to 21,000 cells mL−<sup>1</sup> ) at 79◦N, and documented their presence as far north as 82.5◦N. Covering an annual cycle in 2014, we found that during autumn and winter, Synechococcus was often more abundant than picoeukaryotes, which usually dominate the picophytoplankton communities in the Arctic. Synechococcus community composition shifted from a quite high genetic diversity during the spring bloom to a clear dominance of two specific operational taxonomic units (OTUs) in autumn and winter. We observed abundances higher than 1000 cells mL−<sup>1</sup> in water colder than 2◦C at seven distinct stations and size-fractionation experiments demonstrated a net growth of Synechococcus at 2◦C in the absence of nano-sized grazers at certain periods of the year. Phylogenetic analysis of petB sequences demonstrated that these high latitude Synechococcus group within the previously described cold-adapted clades I and IV, but also contributed to unveil novel genetic diversity, especially within clade I.

Keywords: picocyanobacteria, picoeukaryotes, temperature adaptation, *petB* sequences, flow cytometry, high latitude ecosystems, Svalbard, West Spitsbergen Current

#### INTRODUCTION

The widely abundant picocyanobacterium Synechococcus is estimated to be responsible for about 17% of ocean net primary productivity and thus to have a high impact on ocean ecosystems and biogeochemical cycles (Flombaum et al., 2013). Synechococcus is normally not considered to be bloom-forming even though they can appear in abundances as high as 1.2–3.7 × 10<sup>6</sup> cells mL−<sup>1</sup> in the Costa Rica dome (Saito et al., 2005). Using 37,699 discrete global Synechococcus observations between 69◦ S and 81◦N and quantitative niche models, Flombaum et al. (2013) demonstrated temperature to be the main environmental parameter explaining the global distribution of Synechococcus. Accordingly, the regional range of temperature was found to be a relatively good predictor for the seasonal change in Synechococcus abundance (Tsai et al., 2013). Although the

#### *Edited by:*

Connie Lovejoy, Laval University, Canada

#### *Reviewed by:*

David J. Scanlan, University of Warwick, UK William Li, Fisheries and Oceans Canada, Canada

> *\*Correspondence:* Maria L. Paulsen maria.l.paulsen@uib.no

#### *Specialty section:*

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Marine Science

*Received:* 28 June 2016 *Accepted:* 21 September 2016 *Published:* 05 October 2016

#### *Citation:*

Paulsen ML, Doré H, Garczarek L, Seuthe L, Müller O, Sandaa R-A, Bratbak G and Larsen A (2016) Synechococcus in the Atlantic Gateway to the Arctic Ocean. Front. Mar. Sci. 3:191. doi: 10.3389/fmars.2016.00191

**61**

marine Synechococcus that have been studied in culture so far have a temperature optimum ranging from 20 to 33◦C depending on the clade (Pittera et al., 2014), the highest annual average in situ cell abundances were found at temperatures around 10◦C in the Indian and western Pacific Oceans, with averages of 34,000 and 40,000 cells mL−<sup>1</sup> , respectively (Flombaum et al., 2013).

There has been observations of Synechococcus at low temperatures e.g., <4 ◦C in low numbers (<100 cells mL−<sup>1</sup> ; Waterbury et al., 1986) and <2 ◦C (Shapiro and Haugen, 1988; Gradinger and Lenz, 1995), but they are still often considered to be nearly absent from the polar ocean (Pedrós-Alió et al., 2015) in contrast to cold adapted eukaryotic picophytoplankton that occur in high abundances both in Arctic (Sherr et al., 2003; Lovejoy et al., 2007; Tremblay et al., 2009; Zhang et al., 2015) and in Antarctic waters (Doolittle et al., 2008). Only a few studies have actually documented Synechococcus north of 70◦ and none have so far described the genetic diversity of these northern populations or tested their temperature optimum.

During four expeditions Gradinger and Lenz (1995) observed maximal abundances of 5500 Synechococcus cells mL−<sup>1</sup> in the Atlantic inflow to the Arctic Ocean west of Svalbard at 78◦N, while they did not find any Synechococcus cells in surface samples of polar water (defined as water having Temp < 0 ◦C; Salinity < 34). Further south, following a transect from 70.5 to 74◦N, Not et al. (2005) recorded a maximum abundance of 25,000 cells mL−<sup>1</sup> in the Norwegian and Barents Seas in August 2002. In the western Canadian Arctic, Cottrell and Kirchman (2012) found abundances of 40–80 cells mL−<sup>1</sup> in coastal waters of the Chukchi Sea and the Beaufort Sea at 71.5◦N, both during summer and winter cruises. Nelson et al. (2014) concluded in their overview that the Synechococcus distribution in this region is controlled mainly by inflow of the relatively warm Pacific water, but argue that water temperature alone cannot be used to define environments in which Synechococcus may reside as they do persist at water temperatures near the freezing point (−1.8◦C) (Nelson et al., 2014).

Synechococcus is often found in Arctic lakes and rivers, and freshwater runoff may thus also represent a source of Synechococcus cells to the Arctic Ocean (Vincent et al., 2000). Using 16S rRNA analysis, Waleron et al. (2007) revealed that picocyanobacteria present in the Canadian Beaufort Sea originate from the Mackenzie River and other nearby inflows. High abundances of Synechococcus (30,000 cells mL−<sup>1</sup> ) were also found in the Laptev Sea, but were restricted to brackish waters near the Lena River delta, while further away from the delta, abundances decreased with increasing salinity to a total absence at salinities >20 (Moreira-Turcq and Martin, 1998). All these studies support Waterbury et al. (1986) claiming that only few brackish species tolerate wide salinity ranges and that many strains are obligate marine. Assuming that Atlantic Synechococcus have a low tolerance to salinity changes, the question remains whether the low salinity in the Arctic surface waters constrains their distribution in the polar ocean.

The Atlantic inflow is the main conveyor, not only of water and heat, but also of more southern species into the Arctic Ocean. Synechococcus has accordingly been suggested as a bio-indicator for the advection of Atlantic waters into the Arctic Ocean

main flow of the West Spitsbergen Current drawn after Cokelet et al. (2008),

Randelhoff et al. (submitted).

(Murphy and Haugen, 1985; Gradinger and Lenz, 1995). The main transport follows the West Spitsbergen Current (WSC) which is an extension of the Norwegian Atlantic Current splitting up into two branches around 79–80◦N (**Figure 1**). The WSC is about 100 km wide and is confined over the continental slope along the Norwegian coast. It has an average speed of 10 cm s −1 (Cokelet et al., 2008) but can reach a speed of up to 24– 35 cm s−<sup>1</sup> (Boyd and D'Asaro, 1994; Fahrbach et al., 2001). The inflow follows a strong annual cycle with maximum volume transport during winter (20 Sv in February) and minimum during summer (5 Sv in August, Fahrbach et al., 2001) (N.B. the unit sverdrup (Sv) is equal to 1 million m<sup>3</sup> s −1 ). Strong variations in the strength of the Atlantic inflow combined with varying sea ice extension make it challenging to assess the spread of Atlantic organisms in this area. Little is known about how Synechococcus populations, originating from the Norwegian coast or further south, are affected as they are transported into the Arctic Ocean or whether some Synechococcus lineages are favored under the transition to more Arctic conditions.

Temperature is one of the main drivers of Synechococcus biogeography. Among the five globally dominating Synechococcus lineages (clades I, II, III, IV, and CRD1), clades I and IV dominate at high latitudes in cold and coastal waters, while clades II and III are mostly found in warm, (sub)tropical areas (Zwirglmaier et al., 2008; Farrant et al., 2016; Sohm et al., 2016). Populations adapted to distinct thermal niches were also identified within the CRD1 clade, including one co-occurring with clade I and IV in cold, mixed waters of the Pacific Ocean (Farrant et al., 2016).

Increasing ocean temperature in high latitude systems has drawn attention towards the growth of invasive organisms with higher temperature optima and subsequent ecosystem changes. In marine systems, small phytoplankton are expected to become relatively more abundant with warming (Morán et al., 2010; Tremblay et al., 2012) and it has been speculated that the warming of the Arctic Ocean could lead to a shift from picoeukaryotes to picocyanobacteria, with implications for food quality (Vincent, 2010). Flombaum et al. (2013) projected up to a 50% increase in Synechococcus at 60◦N by the end of the twenty first century. Their models were however not able to make projections for higher latitude systems because observations in these areas are scarce. The aim of the present study is therefore to examine the distribution of Synechococcus in relation to environmental parameters and other microbial plankton groups within the Atlantic gateway to the Arctic. The genetic diversity of Synechococcus populations was also unveiled using a high resolution genetic marker, the petB gene (encoding the cytochrome b<sup>6</sup> subunit), in order to trace the geographical origin and seasonal changes of these populations.

#### MATERIALS AND METHODS

#### Locality and Sampling

This study covers the eastern part of the Fram Strait, where Atlantic water (AW) is transported northward by the West Spitsbergen Current (WSC). Data were collected during five cruises in 2014: January (06.01–15.01), March (05.03–10.03), May (15.05–02.06), August (07.08–18.08), and November (03.11– 10.11). Transects were made across the core of AW inflow at 79 and 79.4◦N during May, August and November. Further north (80.5 to 82.6◦N) we investigated the WSC southern branch into the Arctic Ocean in January, March and August (**Figure 1**). The choice of sampling area and stations was largely determined by the extension of the sea ice (**Figure 3**). Vertical profiles of temperature, salinity and fluorescence were recorded on each sampling occasion using a SBE 911plus system. Water masses were defined based on the criteria presented in **Table 1**. Discrete water samples for analyses of nutrients (NO<sup>−</sup> <sup>2</sup> <sup>+</sup> NO<sup>−</sup> 3 , NH<sup>+</sup> 4 , PO3<sup>−</sup> 4 , H4SiO4) and enumeration of phytoplankton, viruses, bacteria, and heterotrophic nanoflagellates (HNF) were collected from 11 depths (1, 5, 10, 20, 30, 50, 100, 200, 500, 750, and 1000 m) using 10 L Niskin bottles. During the summer cruises we collected additional samples from the Deep Chlorophyll Maximum (DCM) (when different from any of the standard depths). The shallow shelf stations were sampled to near bottom and with higher sampling resolution in the surface.

#### Flow Cytometry

Abundances of pico- and nano-sized phytoplankton, viruses, bacteria and HNF were determined on an Attune <sup>R</sup> Acoustic Focusing Flow Cytometer (Applied Biosystems by Life technologies) with a syringe-based fluidic system and a 20 mW 488 nm (blue) laser. Samples were fixed with glutaraldehyde

#### TABLE 1 | Criteria determining the water masses.


For further explanation see Cokelet et al. (2008). PW overlaps with SW and ArW.

(0.5% final conc.) at 4 ◦C for a minimum of 2 h, flash frozen in liquid nitrogen and stored at −80◦C until analysis, except in November, when phytoplankton was enumerated using fresh samples. For analysis of HNF the samples were stained with SYBR Green I (Molecular Probes, Eugene, Oregon, USA) for 2 h in the dark and a minimum of 1 mL was measured at a flow rate of 500 µl min−<sup>1</sup> following the protocol of Zubkov et al. (2007). The HNF population was discriminated from nano-sized phytoplankton based on green vs. red fluorescence and from large bacteria on a plot of side scatter vs. green fluorescence following the recommendations of Christaki et al. (2011). Pico- and nano-sized phytoplankton were counted directly after thawing and the various groups discriminated based on their side scatter signals (SSC) vs. orange fluorescence (**Figure 2A**; Marie et al., 1997; Larsen et al., 2001) as well as their red vs. orange fluorescence (**Figure 2B**). Synechococcus was identified in plots of orange fluorescence vs. side scatter signals (**Figure 2A**). For samples with low abundance of phytoplankton (March and January) a volume of 1.5 mL was counted, while 0.5 mL was sufficient for May, August and November-samples. Regular blank measurements using Milli-Q <sup>R</sup> water were made to ensure that there was no carry over of cells between samples and that electronic noise did not disturb the counts. Due to the inherent uncertainty connected to enumeration of cells when concentrations are low, we only included samples with > 20 cells mL−<sup>1</sup> when relating counts to other environmental parameters (**Figure 5**). Samples for which 0–20 cells were detected (i.e., mainly those deeper than 500 m) are included in our total data set (**Table S1**, **Figure S2**).

#### Microscopy

The presence of Synechococcus was also confirmed by fluorescence microscopy (**Figure 2C**). Samples were fixed and stored as for flow cytometry. The samples were thawed, filtered onto Anodisc filters (Whatman, pore size 0.2 µm) and stained with SYBR Green I (Molecular Probes Inc., Eugene, Oregon) according to Patel et al. (2007). The samples were viewed and photographed at 400X using a Zeiss Axio Imager Z1 microscope with AxioCam MRm BW-camera, extended focus, epifluorescence illumination (HXP Illuminator) and Zeiss filter sets 09 and 43 for SYBR Green and chlorophyll fluorescence, respectively.

# Size-Fractionated Growth Experiments

Water fractionation experiments were used to examine interaction between different size groups of microorganisms and to estimate growth rates of the different microbial components (Simek and Chrzanowski, 1992; Jürgens et al., 2000; Christaki et al., 2001; Sato et al., 2007). Experiments were performed once every cruise using water collected from 20 m (in August and May this depth was near DCM) at stations on the shelf (marked on **Figure 1**). The water was gently screened through 3, 5, 10, and 90 µm mesh size filters by reverse filtration in order to successively exclude grazers of different sizes and thus create communities with increasing "top-predators" sizes. Water from each filtration treatment was gently transferred into triplicate 3.9 L transparent polycarbonate bottles (Nalgene <sup>R</sup> ) by staggered filling using silicone tubing. The incubation experiments ran for 5 to 10 days but we show data only from the initial 5 days of incubation in order to better represent dynamics of the initial communities. Incubation water was sampled daily for enumeration of microorganisms and every second day for nutrients. Prior to setup, all bottles, carboys and silicon tubs were acid washed and then rinsed with Milli-Q <sup>R</sup> water. During the summer cruises (May and August) the experimental bottles were incubated on deck in plexiglass tanks with seawater flow-through (continuously pumped from 7 m depth), keeping the temperature close to in situ (May: 1.7 ± 1.6◦C and August: 1 ± 0.8◦C). A nylon net was wrapped around each bottle to reduce the PAR to about 30% of the surface irradiation. In the winter months (January, March, November), incubations were kept in a cooling room at a constant temperature of 2◦C and in darkness, except in March were an in situ light cycle was set (16 h darkness and 8 h at 5 µmol photons m−<sup>2</sup> s −1 ). The fractionation experiments

provided net growth rates of Synechococcus and HNF by fitting exponential functions to the change in the abundance of cells every 24 h during the first 5 days of the experiments (**Figure S1**).

#### DNA Extraction, PCR Cloning and Phylogenetic Analysis

Environmental samples for molecular analysis were collected by filtering water onto 0.22 µm pore size Millipore <sup>R</sup> Sterivex filters. The filters were immediately flash frozen in liquid nitrogen and stored at −80◦C until extraction. DNA and RNA were extracted simultaneously using the AllPrep DNA/RNA Mini Kit (Qiagen, Hilden, Germany) according to manufacturer's instructions with some optimisation for extraction from Sterivex filters as follows. The filters were thawed on ice and 1 mL extraction buffer (990 µl RLT buffer; containing guanidine isothiocyanate + 10µL ßmercaptoethanol) was added before incubating for 4 min on a Vortex adapter at medium speed. The resulting lysates were recovered using a 10 mL syringe and used for nucleic acids extraction. DNA samples harvested from Arctic surface water collected in May (80◦N, 10.7◦E at 1 m depth, 10 L water filtered), August (80◦N, 10.8◦E at 1 m, 7.5 L filtered) and November (79◦N, 6◦E, 20 m, 20 L filtered) were selected to amplify the Synechococcus petB marker gene (stations marked on **Figure 1** and profiles of picophytoplankton are included in **Figure S2**). Polymerase chain reaction (PCR) were performed using the petB primers and set-up recommended by Mazard et al. (2012) using 30–40 amplification cycles (iCycler, Bio-Rad, CA, USA). Positive PCR products were purified using the Zymo DNA Clean and ConcentratorTM-5 kit (Zymo research, CA, USA) and subsequently cloned with the StrataCloneTM PCR Cloning Kit (Agilent Technologies, CA, USA) following the manufacturer's instructions. A total of 96 clones from each of the three samples were picked (total 288 clones) and sequenced by LCG Genomic GmbH (Berlin, Germany) using Sanger sequencing. A total of 229 petB sequences were obtained and deposited in the GenBank database (accession no. KX345947–KX346174). These sequences are in the following referred to as "MicroPolar sequences."

The 229 MicroPolar sequences include 174 unique fulllength sequences. Together with 721 petB sequences from a non-redundant reference database (representing most of the genetic diversity so far identified within Prochlorococcus and Synechococcus genera; Farrant et al., 2016), the MicroPolar sequences were used to define operational taxonomical units (OTUs) at 97% identity using Mothur v1.34.4 (**Table S2**). Since all MicroPolar sequences clustered with clades I and IV reference sequences, a subset of the petB database, comprising only the 117 reference sequences of these clades, as well as the 174 unique MicroPolar petB sequences was used for a subsequent analysis. Phylogenetic reconstructions were based on multiple alignments of petB nucleotide sequences generated using MAFFT v7.164b with default parameters (Katoh and Standley, 2014). A maximum likelihood tree was inferred using PHYML v3.0, (Guindon and Gascuel, 2003) with the HKY + G substitution model, as determined using jModeltest v2.1.4 (Darriba et al., 2012) and estimation of the gamma distribution parameter of the substitution rates among sites and of the proportion of invariable sites. The tree was drawn using iTOL (Letunic and Bork, 2007). The 229 sequences retrieved from MicroPolar were recruited using BLASTN (v2.2.28+) against the full petB database: reads with more than 90% of their sequence aligned and with more than 80% sequence identity to their BLASTN best-hit were taxonomically assigned to their best-hit and subsequently used to build per-sequence read counts tables. Counts were then aggregated by OTUs and relative abundance was computed for each MicroPolar station.

#### Nutrients

Unfiltered seawater was filled directly from the Niskin bottles into 30 mL acid washed HDPE bottles and stored at −20◦C. Nitrite and nitrate (NO<sup>−</sup> <sup>2</sup> <sup>+</sup> NO<sup>−</sup> 3 ), phosphate (PO3<sup>−</sup> 4 ) and silicic acid (H4SiO4) were measured on a Smartchem200 (by AMS Alliance) autoanalyser following procedures as outlined in Wood et al. (1967) for NO<sup>−</sup> <sup>3</sup> <sup>+</sup> NO<sup>−</sup> 2 , Murphy and Riley (1962) for PO3<sup>−</sup> 4 and Koroleff (1983) for the determination of H4SiO4. The determination of NO<sup>−</sup> <sup>3</sup> was done by reduction to NO<sup>−</sup> 2 on a builtin cadmium column, which was loaded prior to every sample run. Seven-point standard curves were made prior to every run. Two internal standards and one blank were inserted for every 8 samples and these were used to correct for any drift in the measurements. Concentration of NH<sup>+</sup> <sup>4</sup> was determined directly in fresh samples using ortho-phthaladehyde according to Holmes et al. (1999).

### RESULTS

Synechococcus cells were detected by flow cytometry in all samples within the upper 100 m of the water column during all seasons (**Figure 3** and **Figure S2**). The identity of Synechococcus was confirmed by epifluorescence microscopy (**Figure 2C**) and by sequencing of the petB gene (**Figure 7**). The closely related genus Prochlorococcus was never detected.

#### *Synechococcus* Distribution

The highest sampling frequency was obtained in May and August, when the sampling sites were restricted to latitudes below 80◦N (May) and 81◦N (August) by sea ice (**Figure 3**), while the most northern samples were acquired in January and March at around 82.5◦N. Synechococcus was present in abundances higher than 50 cells mL−<sup>1</sup> in 337 samples both within the Atlantic water, Arctic water, Surface water and Polar water (water mass definitions are shown in **Table 1**). Within the cold surface water (<2 ◦C, upper 50 m), 60% of the samples contained Synechococcus with abundances ranging from 50 to 4300 cells mL−<sup>1</sup> . Synechococcus was not detected in the cold Atlantic water or intermediate water masses, which comprise water collected deeper than 500 m (see temperature-salinity plots in **Figure 3**).

In January, the average abundance in the upper 100 m was 51 cells mL−<sup>1</sup> , with highest abundance found at 100 m depth (maximum 106 cells mL−<sup>1</sup> ) and generally low numbers in the surface (**Figures 3F**, **4A**). The lowest average abundance of

FIGURE 3 | (Left panel) Ice-maps provided from the Norwegian Meteorological Institute (istjenesten@met.no) from following dates; January 10, March 7, May 23, August 12, and November 7, 2014. Transects shown in the middle panel as contour plots are marked with black boxes. (Middle panel) Contour plots showing the abundance (cells mL−<sup>1</sup> ) of Synechococcus and salinity of the upper 200 m from 4 cruises. January transect stretches from North-South between 15 and 20◦E, while the remaining transects expands West-East (2–11◦E) following the 79◦N latitude isoline (see transect marked in boxes left of plots). The horizontal light blue lines above the plots roughly mark the cover of open drift ice. Note different scales for Synechococcus abundance. (Right panel) Potential temperature and salinity (TS) diagram for each month. Data included for all depths 1–1000 m. Synechococcus abundance is given on the z-axis by color gradient (N.B. different scales). Potential density (σ0, kg/m<sup>3</sup> ) isolines overlaid with gray and the surface freezing line is show in dashed blue. Following water masses (Table 1) are marked: Atlantic water (AW), cold Atlantic water or Intermediate water (cAW or IW), which consist mainly of deep water samples (>500 m), Arctic water (ArW), surface water (SW) and Polar water (PW).

Synechococcus was found in March with 13 cells mL−<sup>1</sup> and a maximum abundance that did not exceed 40 cells mL−<sup>1</sup> (**Figures 3H**, **4B**). In May the maximum abundance was around 1300 cells mL−<sup>1</sup> and the average ± SD was 181 ± 147 cells mL−<sup>1</sup> (n = 150; **Figures 3J**, **4C**). The highest Synechococcus abundances were detected in August with a maximum of 21,300 cells mL−<sup>1</sup> . When averaged for the upper 50 m at most southern stations (79–79.4◦N), abundances were 5700 ± 4200 cells mL−<sup>1</sup> (n = 61) (**Figure 3L**) over the whole transect, while abundances at the stations north of 80◦N averaged to about 3000 ± 2000 cells mL−<sup>1</sup> (n = 27). In November Synechococcus cells were evenly distributed down to 200 m (**Figure 4E**), with a maximum of 1000 cells mL−<sup>1</sup> and an average abundance of 600 ± 250 cells mL−<sup>1</sup> (n = 18) within the upper 200 m. The vertical distribution of Synechococcus varied from mainly surface peaks in May (upper 20 m) to maximum abundance at depths greater than 50 m in August and November, to a more vertically uniform distribution in January and March with maxima in abundance at around 100 m depth (**Figures 4A–E** and **Figure S2**).

# Biotic and Abiotic Environment

The association between phytoplankton abundances and environmental parameters showed that the abundance of both Synechococcus and picoeukaryotes decreased with increasing latitude, but that picoeukaryotes were relatively more abundant at the northernmost stations (**Figure 5A**). No clear relationship was found for salinity (ranging from 31 to 35 in this study), although the highest Synechococcus abundances were found at salinities >34.5, while picoeukaryotes had their peak abundance at lower salinities of 33.5–34 (**Figure 5B**). Further, we found picoeukaryotes to be strongly dominant over Synechococcus in 14 out of 17 samples with lowest salinity (31–33), all sampled in August. The abundance of Synechococcus ranged from 250 to 4000 cells mL−<sup>1</sup> in these low salinity samples (**Figure 3S**). The presence of sea ice had no clear effect on the vertical distribution of picophytoplankton but at the ice-covered stations, a subsurface maximum of Synechococcus was most prominent. On the other hand, picoeukaryotes tended to peak near surface in ice-covered stations in March and May, while in August the highest surface

(A,B,E–G) contain abundance of picoeukaryotes (green) and (D) nanophytoplankton (blue). Notice log-scale for (A,B,E–G). (Lower panel) (H–K) Synechococcus abundance (mL−<sup>1</sup> ) plotted against HNF abundance during the different months (indicated by color) within the upper 200 m. The broken line illustrates the one-to-one line. Notice different y-axis between months. Only samples with Synechococcus abundance <sup>&</sup>gt;20 cells mL−<sup>1</sup> were included.

maximum of picoeukaryotes was found within the freshwater lens at stations without ice-cover (**Figure S2**).

The highest water temperatures were measured in August followed by those measured in January. The lowest surface temperatures were recorded in the ice-influenced surface waters in March and May. Temperature was the only parameter that displayed a strong relationship with Synechococcus abundance resulting in an exponential fit (r <sup>2</sup> = 0.66, p < 0.005, n = 346; **Figure 5C**), while picoeukaryotes did not show a similar strong relationship (r <sup>2</sup> = 0.31, p < 0.005, n = 372; **Figure 5C**). Synechococcus was more dominant at stations with low chlorophyll a (i.e., chl a fluorescence) compared to larger nanophytoplankton, which correlated positively to chl a (**Figure 5D**).

Nutrients were evenly distributed over the upper 200 m in the winter months (January to November), although a slightly lower concentration was observed in March within the upper 100 m (**Figures 4F,G,J**). In May and August all nutrients were depleted in the upper 10–20 m, with NO<sup>−</sup> 3 reaching the lowest values (**Figures 4H,I**). NH<sup>+</sup> 4 reached the highest values around 2 µM in August at depths below 20 m. At high N concentrations (>2 µM; **Figure 5E**) Synechococcus were generally less abundant than picoeukaryotes, while under low N conditions they were equally numerous. When looking at the N sources separately it appears that at NH<sup>+</sup> <sup>4</sup> <sup>&</sup>gt; 0.5 <sup>µ</sup>M, Synechococcus increased at higher NH<sup>+</sup> 4 levels, whereas they decreased with increasing NO<sup>−</sup> <sup>3</sup> <sup>+</sup> NO<sup>−</sup> 2 (**Figures 5F,G**).

The abundance of HNF increased during the summer months, from less than 200 cells mL−<sup>1</sup> in the winter months up to 1000 and 1500 HNF mL−<sup>1</sup> in May and August, respectively (**Figures 5H–K**). Synechococcus and HNF abundances generally showed a positive relationship within the upper 100 m. In January and March, Synechococcus and HNF cell numbers were within the same order of magnitude, but with slightly more HNF than Synechococcus (i.e., below the dotted line; **Figure 5H**). In May highest Synechococcus abundances were found at the lowest HNF abundance and vice versa (**Figure 5I**). In August, Synechococcus was generally 10 times more abundant than HNF, a trend also observed in November, although less pronounced (**Figures 5J,K**).

#### Growth and Microbial Interactions

Net-growth rates of Synechococcus and HNF were estimated from four different size fractions (<3, <5, <10, and <90 µm) from each of the five cruises and summarized in **Figure 6** (for abundances during incubation see **Figure S1**). Synechococcus showed positive net growth in 9 out of 20 experiments mainly in January and March. Positive growth rates ranged from 0.01 to 0.13 d−<sup>1</sup> . HNF showed positive growth in 14 out of 20 experiments, displaying a maximum growth rate of 0.45 d−<sup>1</sup> when water was filtered through a 5 µm mesh (Treat < 5 µm) in January, otherwise the highest HNF growth rates were measured in May ranging from 0.13 to 0.3 d−<sup>1</sup> . In January, November, and August HNF growth was reduced to close to zero after filtering in the Treat < 90 µm and in March HNF showed negative growth in all treatments. Synechococcus showed positive growth in January, March in all size-fractions and in the Treat < 90 µm in May, which became strongly dominated by Phaeocystis sp. and where both HNF, picoeukaryotes and heterotrophic bacteria increased in abundance simultaneously. Synechococcus had the strongest negative growth in August and in the Treat < 3 µm in May (**Figure 6**). In summary, we measured a positive growth of Synechococcus and negative growth of HNF in March, but in general negative Synechococcus and positive HNF growth in May, August, and November. Only in January and in the May <90 µm treatment, both Synechococcus and HNF displayed positive net growth. Corresponding to the seasonal changes in abundance (**Figures 5H–K**) the prey:HNF ratio (prey being the sum of all picoplankton; Synechococcus + picoeukaryotes + heterotrophic bacteria) of the initial community was highest in May i.e., most prey per HNF grazer and lowest in August, when HNF were more abundant. Generally, the maximum growth rates of Synechococcus were found when prey:HNF was at its highest (**Figure S1**).

#### *Synechococcus* Diversity

The gene petB, which has proved to display a high taxonomic resolution for picocyanobacteria (Mazard et al., 2012), was used as phylogenetic marker for Synechococcus genetic diversity. Only petB sequences related to clade I and IV were retrieved from our dataset (MicroPolar). Based on a petB reference database (including 117 sequences from clade I and IV, described in Farrant et al., 2016), enriched with the 174 unique petB sequences retrieved from MicroPolar samples, 41 OTUs were defined at 97% ID within clade I and IV (**Figure 7**). The petB sequences obtained in the present study correspond more specifically to sub-clades Ib and IVb, with a clear dominance of subclade Ib. Although none of these 41 OTUs form a new subclade, 17 OTUs were composed of only MicroPolar sequences and were not represented in the previous reference database (colored branches, **Figure 7**). In May sub-clade Ib was the only one present, whereas subclade IVb appeared in August and increased in relative abundance in November, indicating seasonal changes in the community composition. Seasonality was also found within subclade Ib. The majority of sequences obtained in August and November belonged to two specific OTUs (Arctic732-2b\_Ib\_IA and Arctic732-35b\_Ib\_IA), which mostly gather reference sequences from the Barents Sea ("Arctic,"

FIGURE 6 | Net growth rates (d−<sup>1</sup> ) of *Synechococcus* plotted against HNF net-growth. Net growth rates are obtained from the fractionation experiments from each cruise where an exponential growth curve was fitted to the change in abundance of the respective cells during a 5-day period (Figure S1). The color indicates the month and the legend at each point indicates the size-fraction treatment from which the values were obtained.

72.5◦N, 19.57◦W) and the North Atlantic Ocean (The Extended Ellett Line; "EEL," 57–63◦N). In contrast, sequences retrieved from samples harvested in May were more evenly distributed over all other OTUs defined within subclade Ib that mainly gather sequences from The Atlantic Meridional Transect ("AMT," http://www.amt-uk.org/) and the North Sea ("MICROVIR" cruise 50–60◦N).

In order to assess whether the genetic populations sampled in MicroPolar cruises could be related to other cold-water populations, we also recruited Illumina reads from 62 surface water metagenomes collected during the Tara Oceans cross-ocean ecosystem study using the same petB database (Karsenti et al., 2011; Farrant et al., 2016) (https://doi.pangaea. de/10.1594/PANGAEA.840718, note that the Tara Oceans samples analyzed here do not include recent Arctic samples from the latest Tara Ocean cruise as they are not yet published). These data showed that the two most abundant MicroPolar OTUs in subclade Ib (Arctic732- 2b\_1b\_1A and Arctic732-35b\_1b\_1A) had a low relative abundance in Tara Oceans stations. Other OTUs identified in MicroPolar samples were also poorly represented in the Tara Oceans dataset, with the notable exception of two subclade Ib OTUs, "MP\_may\_P1\_1m\_E08\_Ib\_I" and "MP\_may\_P1\_1m\_D10\_Ib\_IA" (formed only of MicroPolar sequences), that were dominant in Tara Oceans coldest stations (<14◦C), and of the subclade IVb OTU "Ellet21\_IVb\_IVC" present in the Tara Oceans dataset at all temperatures and especially at cold (<14◦C) and intermediate (18–22◦C) temperatures. Other MicroPolar OTUs were detected at a similarly low level in all temperature ranges of Tara Oceans stations.

# DISCUSSION

#### Arctic Adaptation; *Synechococcus vs. Micromonas*

For the first time we here documented a high abundance of Synechococcus in the Atlantic gateway to the Arctic Ocean north of 79◦N. Synechococcus is generally not thought to be part of the picophytoplankton community in Arctic water masses (e.g., Pedrós-Alió et al., 2015), which has repeatedly been found to be dominated by picoeukaryotes, such as Micromonas spp. (Not et al., 2005). Li et al. (2013) do document their existence in the Canadian Basin of the Arctic proper although as a very small fraction (2%) of the total picophytoplankton community at the only one station higher north than 70◦ . Arctic Micromonas spp. differ from Micromonas genotypes identified elsewhere in the World Ocean (Lovejoy et al., 2007), with these Arctic types being adapted to low temperatures. Similarly, by combining our observations with data from the Tara Ocean we confirmed the latitudinal shift previously described within the Synechococcus genus between the warm-adapted clades II and III and the coldadapted clades I and IV (Zwirglmaier et al., 2008; Mazard et al., 2012; Farrant et al., 2016). Interestingly, clade IV was clearly dominating in Atlantic waters from the Tara Oceans dataset and its relative contribution seemed to increase with temperature in August and November in MicroPolar samples, while clade I appeared to dominate in colder Arctic waters. Thus, although this would need to be confirmed by physiological characterization of representative strains, it suggests that clade I could be adapted to

colder waters than clade IV. Overall, it seems that temperature is the main driver of Synechococcus abundance and diversity in this area.

In laboratory experiments using isolates from tropical sites, Synechococcus has been found not to grow at temperatures below 10◦C (Mackey et al., 2013), even though they have been observed in nature at temperatures as cold as 2◦C (Shapiro and Haugen, 1988), and 0◦C (Gradinger and Lenz, 1995). Our deck incubation experiments showed that northern Synechococcus populations can actually grow at 2◦C, although with a quite low growth rate (maximum of 0.13 d−<sup>1</sup> ), suggesting a physiological adaptation of Arctic populations to low temperatures that further supports the existence of Synechococcus thermotypes (Pittera et al., 2014). This hypothesis is strengthened by our findings that many MicroPolar sequences formed new OTUs, unveiling an important novel genetic diversity (especially within clade I), which seems to be specific to this geographic area (17 OTUs out of the 41 OTU identified within clades I and IV). Furthermore, sequences obtained from August and November are mainly found in two OTUs within subclade Ib, gathering reference sequences retrieved only at high latitude from the Barents Sea (72◦N) and the North Atlantic Ocean (57◦N), but hardly detected in the Tara Oceans dataset. Altogether, these results point toward the existence of Synechococcus populations endemic to these Arctic or subarctic areas.

The peak-values of Synechococcus were clearly associated with the Atlantic inflow (salinity > 34.9) and abundances decreased exponentially with decreasing temperature and were most often low in ice-associated water. This, along with the tendency of decreasing concentrations with decreasing salinity, is in accordance with the suggestion of Synechococcus being an indicator of saline Atlantic water transported into the Arctic (Murphy and Haugen, 1985; Gradinger and Lenz, 1995) as well as the low tolerance to wide salinity ranges of obligate marine Synechococcus (Waterbury et al., 1986). It should also be noted that although Synechococcus peak abundances were found in the relatively warm, saline Atlantic water, equally high abundances were observed in discrete samples from non-Atlantic water masses throughout the year (**Figure 3**), indicating the potential of Synechococcus to adapt to cold, low saline water, as also suggested by Nelson et al. (2014) for Canadian Arctic Synechococcus. The observed maximum abundance of picoeukaryotes, on the other hand, was found at a salinity of 33.5 and they were in general less affected by low salinities than Synechococcus. The dominance of picoeukaryotes over Synechococcus in the Arctic region may thus be connected to their capacity to stand a wide range of salinities in addition to an adaptation to low temperature. As only a few of our samples had a low salinity (17 surface samples in August have salinity <33), more efforts are needed to confirm this trend. In the Canada Basin of the Arctic Ocean proper Synechococcus abundance of 60 cell mL−<sup>1</sup> was found at salinities substantially lower than 33 (Li et al., 2013).

The extreme changes in light conditions in polar environments may also have been a driver for the diversification of the Synechococcus populations. However, in contrast to Prochlorococcus, obvious light partitioning is usually not observed for Synechococcus (Scanlan et al., 2009) since only one study reported a vertical partitioning of some Synechococcus genotypes so far (Gutiérrez-Rodríguez et al., 2014). In our incubations Synechococcus surprisingly showed a net growth in January and March when light was absent or low, respectively, while picoeukaryotes did not grow (data not shown) (**Figure 6**). The ability of Synechococcus to grow under very low light conditions is presumably related to their capacity to consume dissolved organic matter (Palenik et al., 2003; Cottrell and Kirchman, 2009). Yelton et al. (2016) indeed found that the genetic potential for mixotrophy in picocyanobacteria (through osmotrophy) is globally distributed. Although this still needs to be confirmed by laboratory experiments, it is possible that Synechococcus OTUs detected in November, when there is no light, belong to mixotrophic populations that are adapted to slow growth in the dark. Picoeukaryotes may use another mixotrophic strategy, i.e., bacterial grazing, to sustain growth during dark months (Sanders and Gast, 2012). Our observations that Synechococcus can be more abundant than picoeukaryotes in the Arctic in autumn and winter (**Figure 4**) are consistent with previous results [Gradinger and Lenz, 1995; unpublished results from Adventfjorden, Svalbard (I. Kessel Nordgård, personal communication)] and may suggest that cyanobacterial osmotrophy is a more efficient strategy than picoeukaryotic phagotrophy to survive in the dark.

#### Grazing on *Synechococcus*

The highest Synechococcus abundances were observed when NO<sup>−</sup> 3 concentrations were low. Hence, there is no reason to believe that they were resource controlled. The tendency of increased growth when potential grazers were removed, rather points at a top-down control. The all-year-round presence of heterotrophic flagellates (HNF), considered to be their main predators (Sanders et al., 1992; Christaki et al., 2001; Kuipers et al., 2003; Zwirglmaier et al., 2009) indeed allows for grazer control of the Synechococcus populations. Still, grazing losses of Synechococcus are challenging to estimate as potential grazers can include various nano—but also microzooplankton and the specific loss also depends on the presence of other prey types (i.e., bacteria and picoeukaryotes; Pernthaler, 2005). This is illustrated by the different outcomes of successively removing various grazer fractions, which in March, August and November did not result in different growth patterns, but in January and May led to higher growth rates of Synechococcus when organisms larger than 90 µm were removed (**Figure S1**). Thus, this may reflect a trophic cascade where the microzooplankton graze on HNF and thereby release picoplankton from grazing pressure in the <90 µm fraction. In March, August and November, however, there was little effect of size fractionation, which indicates that small HNF (<3 µm) were the main grazers of picoplankton and that these were not grazer-controlled themselves. Exactly "who" were the most important Synechococcus grazers is not possible to deduce from the presented data, and probably varies over the season. In addition, infection by viruses probably also functions as a top down regulator of these Synechococcus populations (Sandaa and Larsen, 2006), however virus counts remained relatively constant in all five experiments (data not shown). Still, we did find the highest net growth rates for Synechococcus when the HNF abundance was lowest (January and March) as well as the highest Synechococcus in situ abundance in water with low HNF concentration (and vice versa), which is in accordance with the view that HNF control their abundance and distribution at large. The picoeukaryote abundance did not follow the same patterns (data not shown), suggesting that they may have different predators. The fact that autotrophs, such as Synechococcus and picoeukaryotes, persist during winter in very low abundances further suggests that low encounter rates between predator and prey in the highly diluted wintry environment release the picophytoplankton from grazing pressure and allows survival despite adverse growth conditions (Kiørboe, 2008). The experiments also illustrate that Synechococcus in both January, March and August have the highest growth rates in the fractions where the total prey:HNF ratio is highest, indicating that Synechococcus might escape the grazers when other potential prey organisms are relatively abundant.

#### *Synechococcus* As an Active Player in the Arctic and Future Implications

It may be questioned whether the observed occurrence of Synechococcus was simply a result of advection and passive transport via the Atlantic water inflow. Since the highest measured abundances were found within the core of the Atlantic water, this probably represents the major source. The seasonal maximum Synechococcus abundance, which was observed in August, does however coincide in time with the seasonal Synechococcus bloom further south along the Norwegian coast. Given the average transportation time is at its minimum in summer (Fahrbach et al., 2001), it seems unlikely that the encountered seasonal change in Synechococcus community we observed was a mere product of advection of Atlantic water. Moreover, the spatial and temporal distribution of clades and OTUs as well as the observed growth at low temperatures when released from grazing pressure, rather suggests that at least some of the observed Synechococcus populations are adapted to Arctic conditions and are indigenous to these waters.

Due to their small size (1.1 ± 0.4 µm diameter in the subarctic Atlantic; Paulsen et al., 2015), Synechococcus cells are largely grazed by HNF and microzooplankton (Christaki et al., 1999, 2005). This implies that their biomass production will be largely recycled in the microbial food web and thus be of minor contribution to higher trophic levels in the grazing food web. Even at the highest abundances observed in this study, Synechococcus only constitutes a minor part of the Arctic epipelagic carbon and energy pool (e.g., 21,000 cells mL−<sup>1</sup> is equal to 2.3 µg C L−<sup>1</sup> , assuming a diameter of 1.1 µm and 250 fg C µm−<sup>3</sup> ; Kana and Glibert, 1987) relative to the total phytoplankton biomass of 42 µg C L−<sup>1</sup> (assuming a carbon to chl a conversion of 30). A warmer Arctic ocean that may favor Synechococcus at the expense of larger phytoplankton species (Flombaum et al., 2013) implies that more energy and carbon could be retained within the microbial food web, further reducing the contribution of Arctic primary production to the top of the food chain.

#### AUTHOR CONTRIBUTIONS

MP led the collection and analysis of data, and the writing of the paper. All other authors contributed to writing the paper and in addition AL, OM, RS, and LS helped collecting data and performing experiments. LG, HD, OM, and GB helped analyse the data and prepare figures.

#### FUNDING

The work was conducted by the projects MicroPolar (RCN 225956) and CarbonBridge (RCN 226415), both funded by the Norwegian Research Council. HD and LG were supported by the French "Agence Nationale de la Recherche" Program SAMOSA (ANR-13-ADAP-0010).

#### REFERENCES


### ACKNOWLEDGMENTS

We thank the helpful crew of RV Helmer Hanssen and RV Lance and to the MicroPolar and CarbonBridge teams for always being helpful during sample collection. A special thank you to Jean-Éric Tremblay for providing ammonium measurements and to Sophie Radeke for steadfast assistance at the flow cytometer and to Hilde M. K. Stabell for making the clone library. We also thank Daniel Vaulot for fruitful discussions.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00191

Figure S1 | The abundance (cells mL−<sup>1</sup> ) of *Synechococcus* (red) and HNF (blue) plotted on the left y-axis during the first 5 days of fractionation experiments performed during the 5 cruises. The fractions <90 µm (A-E), <10 µm (F-J), <5 µm (K-O) and <3 µm (Q-U) are represented on each row. Exponential functions were fitted (lines) to the abundance providing the net growth rates (µ) given in the upper left corner for Synechococcus (red) and HNF (blue). The total prey (sum of Synechococcus, picoeukaryotes and heterotrophic bacteria) to HNF ratio is plotted for each triplicate on he right y-axis (open black circles), the black line connects the daily average prey:HNF ratio.

Figure S2 | The abundance (cells mL−<sup>1</sup> ) of *Synechococcus* (red) and picoeukaryotes (green) for all months within the upper 500 m, except for March where profiles are shown down to 1000 and 3000 m. Horizontal light blue lines mark the stations that were influenced by sea ice. Note the different x-axis for different months. Coordinates are given for each station above each graph.

Table S1 | Environmental from the cruises containing: dates (mm/dd/yy), latitude and longitude of stations (decimal degrees), depth (m), flow cytometer counts of *Synechococcus*, picoeukaryotes, nanophytoplankton, heterotrophic bacteria, and nanoflagellates (cells mL−<sup>1</sup> ), the growth rates *Synechococcus* and HNF (d−<sup>1</sup> ) from the <90 µm incubation, salinity, temperature and potential temperature (◦C), CTD-fluorescence (RUF), total chl a and the chl a fraction >10 µm (µg L −1 ), and nutrients (NH<sup>+</sup> 4 , NO<sup>−</sup> 3 , NO<sup>−</sup> 2 , PO<sup>+</sup> 4 , Si(OH)4 (µM). N.B. nutrients from January, May and August are not included here but will be available in Randelhoff et al. submitted.

Table S2 | Sequence ID of the members of each Operational Taxonomical Unit (OTU) defined for *petB* at 97% nucleotide sequence identity.


Li (Ottawa, ON: Canadian Bulletin of Fisheries and Aquatic Sciences), 71–120.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Paulsen, Doré, Garczarek, Seuthe, Müller, Sandaa, Bratbak and Larsen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Seasonal and Interannual Changes in Ciliate and Dinoflagellate Species Assemblages in the Arctic Ocean (Amundsen Gulf, Beaufort Sea, Canada)

#### Edited by:

George S. Bullerjahn, Bowling Green State University, USA

#### Reviewed by:

Rebecca Gast, Woods Hole Oceanographic Institution, USA Alison Clare Cleary, Independent Researcher, San Francisco, United States

#### \*Correspondence:

Connie Lovejoy Connie.Lovejoy@bio.ulaval.ca

#### Present Address:

André M. Comeau, Centre for Comparative Genomics and Evolutionary Bioinformatics-Integrated Microbiome Resource, Department of Pharmacology, Dalhousie University, Canada

†

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Marine Science

Received: 12 July 2016 Accepted: 11 January 2017 Published: 07 February 2017

#### Citation:

Onda DFL, Medrinal E, Comeau AM, Thaler M, Babin M and Lovejoy C (2017) Seasonal and Interannual Changes in Ciliate and Dinoflagellate Species Assemblages in the Arctic Ocean (Amundsen Gulf, Beaufort Sea, Canada). Front. Mar. Sci. 4:16. doi: 10.3389/fmars.2017.00016 Deo F. L. Onda1, 2, 3, Emmanuelle Medrinal 1, 3, André M. Comeau1, 3 †, Mary Thaler 1, 2, 3 , Marcel Babin1, 2 and Connie Lovejoy 1, 2, 3 \*

<sup>1</sup> Département de Biologie and Québec-Océan, Université Laval, Quebec, QC, Canada, <sup>2</sup> Takuvik, Joint International Laboratory, UMI 3376, Centre National de la Recherche Scientifique (CNRS, France) and Université Laval, Québec, QC, Canada, <sup>3</sup> Institut de Biologie Intégrative et des Systèmes, Université Laval, Quebec, QC, Canada

Recent studies have focused on how climate change could drive changes in phytoplankton communities in the Arctic. In contrast, ciliates and dinoflagellates that can contribute substantially to the mortality of phytoplankton have received less attention. Some dinoflagellate and ciliate species can also contribute to net photosynthesis, which suggests that species composition could reflect food web complexity. To identify potential seasonal and annual species occurrence patterns and to link species with environmental conditions, we first examined the seasonal pattern of microzooplankton and then performed an in-depth analysis of interannual species variability. We used high-throughput amplicon sequencing to identify ciliates and dinoflagellates to the lowest taxonomic level using a curated Arctic 18S rRNA gene database. DNA- and RNA-derived reads were generated from samples collected from the Canadian Arctic from November 2007 to July 2008. The proportion of ciliate reads increased in the surface toward summer, when salinity was lower and smaller phytoplankton prey were abundant, while chloroplastidic dinoflagellate species increased at the subsurface chlorophyll maxima (SCM), where inorganic nutrient concentrations were higher. Comparing communities collected in summer and fall from 2003 to 2010, we found that microzooplankton community composition change was associated with the record ice minimum in the summer of 2007. Specifically, reads from smaller predatory species like Laboea, Monodinium, and Strombidium and several unclassified ciliates increased in the summer after 2007, while the other usually summer-dominant dinoflagellate taxa decreased. The ability to exploit smaller prey, which are predicted to dominate the future Arctic, could be an advantage for these smaller ciliates in the wake of the changing climate.

Keywords: microzooplankon, 18S rRNA pyrosequencing, interannual variability, indicator species, Arctic ice loss

**75**

# INTRODUCTION

The climate driven changes now occurring across the Arctic are expected to affect marine phytoplankton communities in many ways, for example, by limiting nutrient supply, lowering surface salinities, and potentially by ocean acidification (Coupel et al., 2012, 2014; Riebesell et al., 2013; Thoisen et al., 2015). The loss of multi-year ice has implications for the sea-ice communities (Comeau et al., 2013), and consequently the Arctic food chain (Søreide et al., 2010). But less intuitively, the loss of multi-year ice can have an impact on the pelagic communities, for example by lengthening the open water season and influencing the depth of the upper mixed layer in this salinity stratified ocean. In the Canada Basin and elsewhere, changes in the timing and character of phytoplankton production have already been reported (Li et al., 2009; Ardyna et al., 2014). Less attention has been given to the microzooplankton (ciliates and dinoflagellates) that consume from 22 to 75% of the daily phytoplankton production in the Arctic and other cold water oceans (Paranjape, 1990; Landry and Calbet, 2004; Sherr and Sherr, 2009; Sherr et al., 2013). Ciliates and dinoflagellates are also influenced by top-down processes, where changes in macrozooplankton species are linked to the successional progression of the microzooplankton communities (Riisgaard et al., 2014).

Microzooplankton species, especially dinoflagellates, use a variety of techniques to capture prey (e.g., Nielsen and Kiørboe, 2015), with some ciliates being fastidious in terms of nutritional requirements (e.g., Johnson, 2011), and others having a narrow size specificity range (e.g., Park and Kim, 2010). Because of such selectivity, microzooplankton could have a top-down effect on phytoplankton and protist community composition (Irigoien, 2005), with implications for microbial food webs, nutrient, and carbon cycles, as well as zooplankton nutrition and higher food webs. In addition to their major role as phagotrophs, many microzooplankton are mixotrophic and able to carry out photosynthesis. In particular, about half of the described species of dinoflagellates are photosynthetic with integrated chloroplasts (Saldarriaga et al., 2001; Hansen, 2011). Other dinoflagellates, along with some ciliates, host photosynthetic symbionts, or are kleptoplasidic (Esteban et al., 2010). The range of trophic roles among microzooplankton suggest that the microzooplankton species composition could be used as indicators of differences in microbial food webs. Such an approach would require testing for associations among particular taxa and the conditions where they occur. However, there have been few studies focusing on occurrence patterns of ciliates and dinoflagellates in the Arctic (Nelson et al., 2014).

The use of 18S rRNA gene surveys has highlighted the prevalence of dinoflagellates and ciliates in the Arctic, but most studies have been restricted to summer and early fall (Comeau et al., 2011). Dinoflagellates and ciliates persist over winter darkness in the Arctic (Terrado et al., 2011; Marquardt et al., 2016), but their seasonal progression is poorly understood. To resolve temporal patterns at finer taxonomic resolution, and to test for trends consistent with environmental selection among Arctic microzooplankton, we investigated samples collected from November 2007 to July 2008 during the International Polar Year Circumpolar Flaw Lead Study (IPY-CFL; Barber et al., 2010, 2012). The IPY samples from both surface waters and the top of the Pacific halocline, where a subsurface chlorophyll maximum (SCM) usually develops (McLaughlin and Carmack, 2010; Monier et al., 2015). We hypothesized that the strong seasonality and stratification in Amundsen Gulf (Beaufort Sea) would be sufficient to lead to niche partitioning and select different species assemblages at the two depths and over time. Both DNA and RNA (converted to cDNA) were used as templates for targeted amplicon sequencing to compare the community with potential for active protein synthesis (from RNA) and communities from DNA, which could include free DNA and cells in state of dormancy or encystment (Jones and Lennon, 2010; Hunt et al., 2013). Specifically, we applied high throughput sequencing (HTS) targeting the V4 region of 18S rRNA (referred to as rRNA) and the 18S rRNA gene (referred to as rDNA) to determine species level composition in the communities. To test for interannual variability vs. trends over time, we reanalyzed the 8 years of ciliate and dinoflagellate amplicon data from Amundsen Gulf reported in Comeau et al. (2011). For both the seasonal and interannual studies, we classified the microzooplankton reads using an improved reference database (Lovejoy et al., 2016) and constructed robust phylogenies. We applied multivariate statistics to link changes in community structure and composition with prevalent environmental drivers, to provide insights on how microzooplankton communities respond to seasonal changes and inter-annual variability.

# MATERIALS AND METHODS

#### Sample Collection, Extraction, and Sequencing

Seasonal samples were collected onboard the Canadian Coast Guard Research Icebreaker CCGS Amundsen every 2–4 weeks from November 2007 to July 2008 from Amundsen Gulf and adjacent bays (**Figure 1A**). We targeted the polar mixed layer (PML) near the surface (5–20 m), and the top of a halocline (22– 80 m), which separates the PML from Pacific Summer Water (PSW). In late spring the Beaufort Sea SCM typically forms at this halocline (Bergeron and Tremblay, 2014). During winter and early spring, the halocline sample depth was identified by salinity and temperature profiles with salinities of 31–32, and from May to July confirmed by the deeper chlorophyll fluorescence peak.

Samples used to investigate interannual variability of ciliate and dinoflagellate communities in the SCM were collected during multiple missions to the Amundsen Gulf and Beaufort Sea between 2003 and 2010, aboard the CCGS Amundsen (2003- 2006, 2009-2010), Louis S. St. Laurent (2007), and Sir Wilfred Laurier (2008). Details of these missions have been reported elsewhere (Barber et al., 2010, 2012; Comeau et al., 2011; Terrado et al., 2011).

Except for 2010, all sample were preserved using the same methods, and following extraction, all were amplified and sequenced using the same protocols. Briefly, water samples for DNA (ca. 6 L) and RNA (ca. 4 L) were collected into cleaned acid rinsed carboys from 12-L (CCGS Amundsen) or

10-L (CCGS Louis St Laurent, Sir Wilfred Laurier) Niskintype bottles mounted on a Rosette system equipped with a conductivity, temperature and depth (CTD; Sea-Bird Electronics Inc., California), chlorophyll fluorescence (Seapoint Sensors Inc.), and relative nitrate from ultraviolet spectrophotometer (ISUS, Satlantic) profilers. Target depths were identified by fluorescence, salinity, temperature, and relative nitrate from the downcast and water collected on the upcast. All samples were filtered and preserved within 2 h of collection. Following prefiltration through a 50-µm mesh to remove macrozooplankton, samples were sequentially filtered using a peristaltic pump (Cole-Parmer, USA) onto a 3-µm pore-size 47-mm polycarbonate filter (PC, AMD Manufacturing), and 0.2 µm pore-size SterivexTM unit (Millipore) for the DNA and a 47-mm 0.2 µm poresize PC filter for the RNA. All PC filters were placed in 2 mL cryovials. Prior to 2010, DNA was preserved by adding 1.8 mL of lysis buffer to the Sterivex units (0.2–3 µm fraction) and cryovials (3–50 µm fraction), and RNA was preserved in RLT buffer (Qiagen) as in Terrado et al. (2011) and Thaler and Lovejoy (2015). Samples were either frozen immediately at −80◦C or placed in liquid nitrogen after adding buffer. Samples collected in 2010 were preserved using RNALater (Ambio). After 30–60 min in buffer, the filters were flash frozen in liquid nitrogen and then kept at −80◦C until processing in the laboratory. We note that preliminary tests on paired samples from both marine and freshwater environments showed no significant differences in DNA and RNA recovery or community compositions between the protocols (Lovejoy pers. comm.). For samples collected before 2010, DNA was extracted from the filters as in Terrado et al. (2011) using a salt based method (Aljanabi and Martinez, 1997) while RNA was extracted using the RNAEasy micro Kit (Qiagen). The sample collected in 2010 was extracted using the AllPrep DNA/RNA Mini Kit (Qiagen) with DNA and RNA from the same filters. Conversion of RNA to cDNA was carried out using the High Capacity Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA). The V4 region of 18S rRNA in both DNA and cDNA samples was amplified using eukaryotic specific forward primers E572F with the Roche A adapter and reverse primer E1009R as described by Comeau et al. (2011). For the seasonal samples, large and small fractions were pooled after normalization based on relative chlorophyll concentrations (Supplementary Table 1). Only the 0.2–3 µm fraction was used in the interannual study. Separate sequencing runs were carried out for the seasonal and interannual studies. For each run, equimolar concentrations of amplicons from each sample were pooled for multiplex sequencing using the GS FLX Titanium Roche 454 platform (Roche/454 Life Sciences, Branford, CT, USA) at IBIS/Université Laval Plateforme d'Analyses Génomiques (Quebec, QC, Canada). All reads have been deposited in the Short Read Archive of NCBI under accession codes: PRJNA283142 (IPY-CFL) and SRA029114 (Amundsen Gulf Time Series, Comeau et al., 2011).

#### Post-sequence Data Processing and Taxonomic Classification

The seasonal data reads were quality filtered using the Quantitative Insights into Microbial Ecology program (QIIME; Caporaso et al., 2010b). Short sequences and primers were identified in mothur and removed (Schloss et al., 2009). Data were then de-noised following Reeder and Knight (2010). Chimeric sequences were identified in UCHIME (Edgar et al., 2011) and removed. Interannual data was processed using mothur (Schloss et al., 2009) as described in Comeau et al. (2011). The two data sets were then processed for OTU picking at 98% similarity utilizing USEARCH (Edgar, 2010) against the Silva Reference Database v.102 (Pruesse et al., 2007). All OTU representative reads were aligned in PyNast (Caporaso et al., 2010a), manually curated in BioEdit v.7.2.5 (Hall, 1999) and used to construct a phylogenetic tree (FastTree; Price et al., 2010). Assignment of taxonomic identity was performed in mothur with a 0.8 confidence threshold against the Northern Reference Database v.1.0 (Lovejoy et al., 2016), which follows the Silva taxonomy but includes high-quality longer environmental sequences from the Arctic and North Atlantic (Terrado et al., 2009, 2011; Charvet et al., 2012; Monier et al., 2013; Dasilva et al., 2014). However, using this database, around 30–45% of the reads were still not classified beyond "Other Dinoflagellates" or "Other Ciliates."

To further increase taxonomic resolution, dinoflagellate sequences from Saldarriaga et al. (2001, 2004), Logares et al. (2007), and Potvin et al. (2013), and ciliate sequences from Dunthorn et al. (2014), along with Evolutionary Placement Algorithm, Randomized-Axelerated Maximum Likelihood- (EPA RAxML, Stamatakis, 2014) abundant Arctic-derived HTS reads were added to the v.1.0 of Northern Reference Database (v.1.1, Lovejoy et al., 2016). Taxonomic classifications of the dinoflagellate sequences used as references were verified in the AlgaeBase.org (Guiry and Guiry, 2015) and by literature searches, while ciliate taxonomic classification followed Lynn (2012). Using the v.1.1 of the database, taxonomy assignment was re-performed to generate the final OTU matrix (OTU reads per sample) with improved results (i.e., "Other Dinoflagellates" and "Other Ciliates" went down to 5–10%). OTUs matching metazoa, fungi, higher plants and reads that could not be classified to any taxonomic level below Eukaryota were removed and not analyzed further. This cleaned data set was considered as the total microbial eukaryotic community. To facilitate diversity comparisons, the OTU matrix was then rarefied, based on the sample with the fewest reads in the separate data sets, resulting in 4953 reads per sample for the seasonal study, and 7041 reads per sample for the interannual comparison.

To investigate the OTUs that contributed to the changes in July 2008 for the interannual data (see Section Results), the remaining abundant "Other Ciliate" OTUs were searched for in NCBI and their most similar nearly full length 18S rRNA gene sequences were used to construct reference trees as in Thaler and Lovejoy (2015). Then, the abundant but unclassified ciliate OTUs were then mapped back onto the reference trees using EPA RAxML v.8 (Stamatakis, 2014). We then inferred their probable taxonomic identity based on the nearest or most similar annotated reference sequences.

#### Statistical and Diversity Analyses

Statistical tests were carried out in the R environment v 3.0 (R Development Core Team, 2008). Spearman's Rank Correlation (rho) in the Vegan package (Dixon and Dixon, 2003) was used to test correlations between taxa and environmental parameters, and results were plotted in igraph (Csárdi and Nepusz, 2006). Only correlations with Spearman's rho ≥ 0.3 and significant at p < 0.001 were retained (Barberán et al., 2012). Speciesvariable relationships were then visualized by network analysis as target-source plots in Cytoscape 3.0 (Shannon et al., 2003; see Supplementary Methods 1). Analysis of variance (ANOVA) was applied to determine significant differences between samples and linear regression tests were used to infer relationships, both analysis were carried out using PAST v3.0 (Hammer et al., 2001).

Alpha diversity (Chao1 index) of the ciliate and dinoflagellate communities in each seasonal sample was estimated as implemented in QIIME (Chao and Shen, 2003; Caporaso et al., 2010b). Since the samples used for the interannual part of the study were collected in different months (Jul to Nov) over the 8 years, we tested for variability associated with the month of collection and found no significant differences in species richness of OTUs in ciliate or dinoflagellate abundance between summer and autumn (Supplementary Figure 1). Phylogenetic unweighted UniFrac dissimilarities (beta diversity) among the taxa were computed separately for seasonal and interannual datasets by the jackknife method and generalized UniFrac procedure (Lozupone and Knight, 2005; Chen et al., 2012). All dinoflagellate and ciliate OTUs in the seasonal study were included in the beta diversity measure. To avoid potential rarefication artifacts (Ramette, 2007) in the interannual data, rare OTUs that were defined as < 1% of the total ciliate or dinoflagellate reads, were not included in the beta diversity and remaining analysis. The pairwise dissimilarity matrices were used to generate dendrograms using UPGMA (Price et al., 2010). The weighted contributions of seasons, depths (surface or SCM-halocline) and templates (RNA or DNA) to community structuring were computed using ADONIS (Fierer et al., 2012). The checkerboard score (C-score) was used to test for non-random co-occurrence patterns under the null hypothesis parameter using the oecosimu function in the R library bipartite with 1000 simulations (Dormann et al., 2008).

# RESULTS

# Physico-Chemical Regimes of the Amundsen Gulf

For the seasonal study (**Figure 1A**), surface PML waters remained cold (−1.7 to 2◦C) from mid-January to mid-May and increased toward summer reaching 8◦C in July, when the region was ice-free. Phosphate concentration changed little over time with concentrations from 0.6 to 1.6 mg m−<sup>3</sup> . Over the 9 months, nitrate concentrations were significantly greater in the halocline compared to the surface (t-test, p < 0.001), with greatest concentrations from February to April (**Table 1**). Nitrate concentrations started to decrease beginning in early April in the surface and in May at the halocline. The chlorophyll a (Chl a) concentrations, as estimated from in situ fluorescence, were negligible until 9 April and were greater in the surface waters compared to halocline through 19 May. Concentrations were below 1 mg m−<sup>3</sup> in both the surface and halocline until July when a strong SCM was apparent at the halocline (**Table 1**).

The physical characteristics of the time series stations (**Figure 1B**) have been reported elsewhere (Comeau et al., 2011). Briefly, over the 8 years, nutrients in the SCM ranged from 0.09 to 8.22 mmol m−<sup>3</sup> for NO<sup>−</sup> 3 , 0.46–1.29 mmol m−<sup>3</sup> for PO<sup>4</sup> and the highest Chl a (total) was detected on September 2005 reaching 2.14 mg m−<sup>3</sup> (Supplementary Table 2). After binning the data from before and after 2007, total summer-fall Chl a values went from 0.43 mg m−<sup>3</sup> to 0.31 mg m−<sup>3</sup> in the SCM.

#### Community Clustering Based on 18S rRNA and 18S rRNA Genes

The seasonal rDNA and rRNA OTU communities separated into distinct surface and halocline clusters, with clear separation

TABLE 1 | Dates of collection 2007-2008 (Date), Stations (Stn), collection Latitude, and Longitude (Lat-Long), physico-chemical parameters, and chlorophyll a (Chl a) concentrations of the samples used for amplicon tag pyrosequencing.


Other column names refer to day length (DayL), depth (Z) of sampling, temperature (T), salinity (S), nitrate+nitrite (NO<sup>−</sup> 3 ), phosphate (PO3<sup>−</sup> 4 ), dissolved oxygen (DO), and Photosynthetically Active Radiation (PAR). na; not available.

between the rDNA and rRNA communities within the depth defined clusters (**Figure 2**). Within the template clusters samples separated by season as Autumn-Winter (Nov-April) and Spring-Summer (May-July) categories, except for one rRNA surface sample collected on June 14, which clustered with the winter rRNA samples (Supplementary Figure 2). The similarity was mainly driven by the presence of several OTUs that occurred in the rRNA surface community on Jan 3 and June 14 but were absent in the other summer samples (Supplementary Table 3). We note that, because of ship time constraints, the June 14 samples were collected from Darnley Bay, which is an extension

chlorophyll maxima/halocline (SCM; circles) layers representing different seasons including Autumn-Winter (AW) and Spring -Summer (SS), collected over the course of the IPY-CFL study.

of Amundsen Gulf. For all samples, the ADONIS test also showed individual contributions to clustering of around 14% (adj. R 2 = 0.148 and 0.146, p < 0.001) grouping by depth category and sample type, while season contributed the least to the variance observed (adj. R <sup>2</sup> = 0.073, p < 0.001). The C-score test further supported the non-random distribution of the microzooplankton communities (p < 0.001).

#### Seasonal Succession and Distribution

Since DNA can be retrieved from both living and senescent cells, and free DNA can persist in the environment (Torti et al., 2015), the RNA-sourced samples were used to investigate potential influence by local conditions. Based on rRNA reads, ciliates, and dinoflagellates were found in all samples (**Figure 3**, Supplementary Tables 4 and 5), with greater proportions during summer (ca. 40%) and lowest in spring (ca. 5%) when larger photosynthetic taxa (mostly diatoms, Joli et al., accepted) dominated the planktonic community. The combined mean seasonal alpha diversity (Chao 1 index) of both ciliates and dinoflagellates based on rDNA, decreased from 125 in winter to 83 after spring (Supplementary Figure 3). In terms of total relative abundance, ciliate proportions were similar in both depths during winter (**Figures 3A,C**) while dinoflagellates were better represented in the surface (**Figures 3B,D**). Ciliate read relative abundance increased in the surface starting early June (**Figure 3A**), while dinoflagellates increased in the halocline (**Figure 3D**).

Significant associations (Spearman's rho > 0.3, p < 0.001) among taxa and environmental variables were detected in the network analysis (**Figure 4**). The associations were consistent with the seasonal succession and depth (water mass) seen in the UniFrac clustering (**Figure 2**). For example, surfaceassociated ciliates: Strombidium, Pseudotontonia, Askenasia, and dinoflagellates Gymnodinium-like, Adenoides-like, Gyrodinium c.f. gutrula, and Scripsiella (**Figure 4**, taxa 1–7) were associated with higher DO, which was a characteristic of the surface PML. In addition, abundances of Amphidoma, "Arctic Clade 1," Blastodinium, Monodinium, and Novostrombidium (**Figure 4**, taxa 8–13) were higher when Chl a concentrations were greater, which was associated with longer day lengths and warmer water, suggesting characteristic surface summer taxa. Tintinnidium, Parauronema, Oligohymenophorea, Gyrodinium helveticum, and Amphidinium-like taxa (**Figure 4**, taxa 15–18) were more associated with the deeper, more saline and nutrientricher SCM or halocline layer. Although Laboea (**Figure 4**, taxon 14) was also strongly correlated with the SCM/halocline conditions, it was most abundant when the days were longer. In contrast, Woloszynskia, Gymnodinium sp., and "Other Gyrodinium" (**Figure 4**, taxa 19, 21, 22) were more abundant in the halocline during winter when ice cover was more extensive. These species, correlated with specific environmental variables, provide evidence for succession and seasonality within the two water masses. Binned OTUs of poorly defined groups that were not classified beyond their respective ranks, including "Other" Dinoflagellates, Ciliophora, Spirotrichea, Litostomatea, and Strombidium (**Figure 4**, taxa 23–30), did not correlate with any of the environmental variables tested.

#### Interannual Variability

Ciliate communities were consistently dominated by "Other Spirotrichea" with annual mean abundance of 9.4 ± 2% relative to the total microbial eukaryotic reads. "Other ciliates" (1.7 ± 1%), Strombidium sp., Pelagostrombidium sp., "Other Oligohymenophorea," "Other Litostomatea," Askenasia sp., and Monodinium sp. abundances ranged from 0.01 to 1.5% of total microbial eukaryote relative abundance (Supplementary Figure 4A). The most abundant dinoflagellate OTUs were from Gymnodiniales, Blastodinium-like, "Arctic clade 1," Gyrodinium, Adenoides-like, and unclassified Dinoflagellates, each accounting for 1–6% of the total microbial eukaryotic reads (Supplementary Figure 4B).

Here, using our refined taxonomic database we found that microzooplankton community composition also changed, especially in summer 2008. Specifically, based on abundant OTUs (>1%), the July 2008 sample showed the greatest dissimilarity from other communities (**Figure 5A**). This dissimilarity was driven by the July 2008 increase in both abundance and OTU richness of Strombidium, Laboea, Monodinium, and "Other Ciliates" (**Figure 5B**), with a significant correlation with decreasing salinity (r <sup>2</sup> = 0.82, p < 0.01). After 2008, the relative abundances of these groups returned to previous levels by 2010. Their relative abundances were also inversely correlated (multiple regression) with SiO<sup>3</sup> (r <sup>2</sup> = 0.29, p < 0.001), salinity (r <sup>2</sup> = 0.22, p = 0.1), and nitrate (r <sup>2</sup> = 0.18, p = 0.2). Phylogenetic placement of the "Other Ciliate" reads further revealed that these were in the Class Spirotrichea, and included sub-classes Oligotrichia and Choreotrichia. Most of these Spirotrichea clustered with environmental sequences, which we refer to as Environmental Clades 1, 2, 3, 4, 5, and 6 (Supplementary Figure 5). Aggregated read counts of Oligotrichia, and Environmental Clades 1, 4, 5, and 6 showed significant increase after 2007 (Krustal Wallis, p < 0.01), while Choreotrichia, and Environmental Clades 2 and 3 decreased.

#### DISCUSSION

Temporal and distributional studies of ciliates and dinoflagellates are usually restricted to morphologically recognizable species

FIGURE 5 | (A) The pairwise unweighted UniFrac community dissimilarity of combined ciliate and dinoflagellate communities. In the boxplots, one point (dot) represents the pairwise Unweighted UniFrac dissimilarity (Y-axis) of a particular sample against another sample. Thus, for each date there are 10 dots (overlapping dots are obscured) representing that date compared against 10 dates. The black line within the box, is the mean UniFrac distance of a particular sample against all the other samples. The broken line indicates the mean dissimilarity value among all samples. (B) Combined relative abundances of the major taxa contributing to the high July 8 dissimilarity, particularly of Strombidinium, Laboea, Monodinium sp. and other unclassified ciliates.

(Montagnes, 1996; Levinsen and Nielsen, 2002). The use of HTS coupled with the improved reference database provided higher resolution of potential species that suggests high diversity compared to previous reports from the Arctic. Overall, we discriminated around 30 taxa per 100 reads, with the total number of reads always higher for dinoflagellates than for ciliates. This translated into 251 dinoflagellates and 141 ciliate species based on OTUs defined as 98% similar as estimated using Choa1. These species estimates are considerably higher than the 55 ciliate species reported in the Western Canada Basin (Jiang et al., 2013) and the <20 dinoflagellate species reported in the Beaufort Sea (Okolodkov and Dodgeb, 1996) based on microscopy. The high proportion of environmental clades here suggests that there are likely ciliate and dinoflagellate species confined to the Arctic, as is the case for diatoms (Luddington et al., 2016) and prasinophytes (Lovejoy et al., 2007).

#### Interpretation of DNA and RNA-Derived Abundances and Diversity

Eukaryotic microbial communities from rDNA have been used to infer water mass history (Hamilton et al., 2008; Monier et al., 2013; Zhang et al., 2014), whereas taxa from rRNA are thought to more closely indicate active representatives of the community (Campbell and Kirchman, 2013; Hunt et al., 2013). Here, we found that for both DNA and RNA sourced samples, the surface communities always clustered together separate from the SCM communities. Earlier microscopy (Lovejoy et al., 1993; Okolodkov and Dodgeb, 1996; Jiang et al., 2013) and clone library studies using longer 18S rRNA sequences (Bachy et al., 2011; Lovejoy and Potvin, 2011) have also highlighted the differences between SCM and surface communities over the Arctic. Within the two water masses, communities separated by template suggests a pool of historic DNA, which could include dormant or less active stages, e.g., cysts (Verni and Rosati, 2011; Bravo and Figueroa, 2014), advected non-active species (Lovejoy and Potvin, 2011), preserved free DNA, or non-living material in marine snow (Nielsen et al., 2007; Boere et al., 2011). All of these sources can be transported by currents and persist over long distances (Heiskanen, 1993; Brocks and Banfield, 2009).

Although RNA-derived reads cannot be directly used to estimate biomass (Medinger et al., 2010; Blazewicz et al., 2013), the patterns of relative read abundance were consistent with specific environmental drivers operating on particular species or clades, with communities separating by season. One exception was the June 14 rRNA surface sample from Darnley Bay, which based on OTU composition clustered with the January 3 rRNA sample. The placement of the June sample in an otherwise winter cluster was consistent with deeper unmodified Pacific Winter Water appearing in Darnley Bay, which was evident in the physico-chemical sample clustering (Supplementary Figure 2). Upwelling and on-shelf transport of deeper Pacific Water from Amundsen Gulf has been reported previously (Garneau et al., 2006; Paulic et al., 2012). Unfortunately, CTD transects were not taken along the shelf break in June and the source of this water remains speculative. The biological clustering was driven by 15 OTUs belonging to "Arctic Clade 1," Gyrodinium, Gymnodinium, "Other Dinoflagellates," Askenasia, Tintinndinium, "Other Spirotrichea," and "Other Ciliates."

#### Environmental Influences on Microzooplankton

While the dependence on phytoplankton prey can generate annual bimodal or unimodal microzooplankton bloom patterns in some regions, with temperature having an influence (Levinsen and Nielsen, 2002; Godhantaraman and Uye, 2003), there is little understanding of species composition over annual cycles. Here we found that ciliate and dinoflagellate taxa showed distinct seasonal patterns that could be partially explained by salinity, temperature, and nutrient changes from winter to summer (**Figure 4**). Specifically, the proportion of ciliate reads were generally higher under ice free conditions, when the surface waters were less saline and at the end of the spring bloom when nitrate concentrations were drawn down by photosynthetic activity (Forest et al., 2010; Tremblay et al., 2011). Although many ciliates have a wide range of salinity tolerance (Montagnes, 1996), this has not been investigated for Arctic species. In contrast dinoflagellate proportions were higher in the surface in winter, but more abundant in the SCM toward summer. Dinoflagellates can produce the osmolyte dimethylsulfoniopropionate (DMSP), which allow cells to regulate internal homeostasis (see review by Stefels, 2000), which may provide an advantage relative to ciliates under conditions of slightly higher salinity.

During late spring and toward summer with increasing light availability, the putative plastid-containing ciliates and dinoflagellates increased proportionally in the rRNA OTUs, suggesting that photosynthesis may have provided additional energy for photosynthetic or mixotrophic taxa. In Amundsen Gulf, in summer, the higher sun angle and longer daylength mean that more light reaches the nutrient-rich halocline leading to the establishment of the SCM. We found that ciliate and dinoflagellate reads made up on average ca. 60% of the total microbial eukaryotic reads in early summer in the SCM. In contrast, a microscopy study in 2005-2006 reported that photosynthetic and non-photosynthetic dinoflagellates constitute only around 3–6% of the total mean phytoplankton cell counts in the SCM (Martin et al., 2010). The discrepancy is likely related to cell breakage during collection, misidentification with smaller dinoflagellates grouped with other flagellates, and loss during storage. Ciliates, can also be fragile and tend to be either ignored or under sampled using standard Lugol's techniques (Lovejoy et al., 1993). Ciliate and dinoflagellate dominance in 18S rRNA gene libraries has also been attributed to their comparatively larger genomes and multiple copies of the rRNA gene (Godhe et al., 2008). Irrespective of the actual cell abundance, our analysis clearly revealed otherwise undetected changes in species over seasons (Supplementary Tables 4 and 5). Our seasonal data ended in July, but the interannual data in the SCM showed that both alveolate groups remain through summer to late autumn.

### Ecological Functions

Season and depth were also the major factors that grouped ciliates and dinoflagellates using the network analyses. The seasonal changes in microbial composition also suggested shifts in dominant function within the two groups (**Figure 4**). For example, winter communities were mainly dominated by heterotrophic taxa. The surface winter communities would have access to small planktonic species such as Micromonas (Lovejoy et al., 2007), haptophytes, and smaller heterotrophic flagellates (Terrado et al., 2011). The relative scarcity of food resources during winter likely selected for particular groups as reported for Kongsfjorden, Svalbard (Seuthe et al., 2011), and Disko Bay (Levinsen et al., 2000; Levinsen and Nielsen, 2002). Dinoflagellates can also graze on smaller ciliates using pseudopodia or feeding veils (Jacobson and Anderson, 1986, 1996). Such trophic interactions, with dinoflagellates most likely preying on ciliates, could contribute to differences in relative abundances in different depths and times (Hansen et al., 1999; Møller et al., 2006; Seuthe et al., 2011). Additional trophic interactions were indicated in the summer when primary production was higher and zooplankton were more abundant in surface waters (Seuthe et al., 2007; Forest et al., 2011), with the appearance of the putative copepod parasite Blastodinium.

After spring and toward summer, the correlation of some ciliates and dinoflagellates with Chl a might also be associated with predation on phytoplankton. Ciliates and dinoflagellates consume from 30% (15.6 g C m−<sup>2</sup> ) of gross primary production in Amundsen Gulf to as much as 56% in some Arctic fjords (Seuthe et al., 2007; Forest et al., 2011). The availability of light in spring and summer could favor mixotrophy as well. For example, the obligate mixotrophic Laboea and most litostomatids, and the potentially plastidic Scripsiella-like, "Other Gyrodinium," "Other Gymnodiniales," and Adenoides further increased in abundance in the SCM over this period when both light and prey would be available.

# Interannual Microzooplankton and the Changing Arctic

The summer-autumn ciliate and dinoflagellate assemblages tended to be similar in the different years, with a consistent predictable re-establishment of communities most years, despite samples being collected from across a large geographical distance. The SCM in the Beaufort Sea forms near the upper surface of the Pacific Winter Water, which is transported as a distinct entity over long distances (Carmack and Macdonald, 2002). Previous studies consistently show that protist communities in the SCM of Beaufort Sea and Amundsen Gulf change little over geographical space and for a given year are strongly associated with their water mass of origin (e.g., Comeau et al., 2011; Lovejoy and Potvin, 2011; Monier et al., 2015). The strong association with water mass was evident in the two samples collected in July 2008 that were analyzed separately because of differences in sample preparation protocols between the seasonal study and the interannual study. The internannual study was only from the small fraction, while the seasonal study included amplicons from large and small fractions. Despite this, the dinoflagellate and ciliate OTUs from Amundsen Gulf collected 8 July, were similar to those collected further offshore in the Beaufort Sea on 21 July, consistent with both 2008 samples from the same water mass and similar conditions.

Lovejoy et al. (2002) showed in microcosm experiments in the Arctic that ciliates and dinoflagellates dominated when nutrients were depleted following blooms of large phytoplankton. Ciliates are able to exploit oligotrophic conditions, by grazing on smaller phytoplankton and bacteria either as strict heterotrophs or by way of mixotrophy. By the end of July 2008, the Beaufort Sea SCM was nitrate limited and had the lowest mean phosphate concentrations measured from 2003 to 2010. Overall primary productivity was low (Martin et al., 2013) with small cells (<3 µm) dominating the Chl a biomass (Comeau et al., 2011). The mean salinity at the SCM was lower in 2008 compared to the preceding years (2003-2007), which was due to the input of additional freshwater from multi-year ice melt in 2007. This event was part of the ongoing trend in decreasing summer ice extent that has been recorded since the onset of the satellite era. The continuing low ice volume over winter was associated with the 2008 ice break-up 2 weeks before the historic average. This precocious ice breakup stimulated an earlier pelagic spring bloom and earlier depletion of nutrients in the euphotic zone (Forest et al., 2011). These July 2008 conditions were reflected in the microzooplankton assemblage, characterized by the greater prevalence of Strombidinium, Laboea, Monodinium sp., other unclassified ciliates and relative decrease in most dinoflagellates, which was also noted in the seasonal data set. The relative abundance of ciliate taxa in the above community gradually returned to earlier levels along with the partial recovery of summer ice extent. At the OTU level, Spirotrichea showed different trends, before and after 2007. Spirotrichea ciliates are obligate mixotrophs, feeding on smaller prey and presumably able to use retained chlorophyte or prymnesiophytederived plastids for photosynthesis (Stoecker et al., 1988, 2009; McManus and Katz, 2009). Similar observations were reported by Jiang et al. (2013) in Western Arctic Ocean where a few ciliate groups became dominant during the 2012 record ice melt, supporting the notion that microzooplankton are highly sensitive to changing physico-chemical regimes, including those associated with low-ice or freshening events.

The mean vertical positions of the SCM and the nitracline in the Beaufort Sea and Amundsen Gulf deepened from 2003 to 2011 (Bergeron and Tremblay, 2014). The increasing distance between the depth at which light is still sufficient for photosynthesis and the nitracline will continue to accelerate under a changing climate regime (Yamamoto-Kawai et al., 2008; McLaughlin and Carmack, 2010; Krishfield et al., 2014), which could drive both the surface and SCM layers to become more nutrient-limited. Such an effect has already been reported for the Canada Basin with an increase in smaller phytoplankton and lower nitrate levels in the upper 200 m (Li et al., 2009). Under these new conditions ciliates and dinoflagellates that are better able exploit smaller prey would be selected for. Our results support recent mesocosm and modeling experiments that suggest that under future conditions, the microzooplankton and microbial loop will become more prominent as efficient intermediates between bacteria-picoplankton and the classical food webs (Skjoldborg et al., 2003; O'Connor et al., 2009; Montagnes et al., 2010; Lewandowska et al., 2014; D'Alelio et al., 2016).

In summary, this study shows that in Amundsen Gulf, ciliates and dinoflagellates exhibit complex temporal dynamics and are influenced by biological and physico-chemical regulators. Ciliate reads were more abundant in the surface when salinity was lower toward summer while dinoflagellates dominated in the SCM where nutrients and light were available for potential mixotrophic or photosynthetic activity. Despite this strong seasonality, the microzooplankton assemblages were also highly similar every summer from 2003 to 2010 except in 2008, following the 2007 summer ice minimum. Mixotrophic taxa increased drastically when nutrient concentrations were low and small prey were presumably available. Low nutrient-small cell conditions are predicted to occur more frequently in the Arctic, and this may lead to a change in dominant microzooplankton species that link to the classical food webs.

# AUTHOR CONTRIBUTIONS

DO and CL conceived the project. CL and EM collected samples. EM and AC carried out the laboratory work, and pre-processed the sequence data. DO, AC, and CL analyzed the data, DO and CL wrote the manuscript and prepared the figures. MB was responsible oceanographic analysis supervision. All authors commented on the text and agreed to this submission.

# FUNDING

The seasonal data was collected during Circumpolar Flaw Lead— International Polar Year (CFL-IPY) study supported by the Natural Sciences and Engineering Research Council of Canada (NSERC), and the Network of Centers of Excellence ArcticNet. NSERC Discovery grants and ArcticNet funding to CL facilitated completion of the study. DFLO received scholarships from Université Laval and the Canadian Excellence Research Chair— Remote Sensing of Canada's New Arctic Frontier (CERC) grant, and additional support from the Fonds de recherche du Québec Nature et Technologies (FRQNT) to Quebéc-Océan aided in this research. We also acknowledge support from Compute Canada.

#### ACKNOWLEDGMENTS

The authors acknowledge the help and assistance of staff at the Institute of Ocean Sciences, Department of Fisheries and

#### REFERENCES


Oceans Canada, the Captains and Crews of CCGS Amundsen, Louis S St. Laurent, and Wilfred Laurier. We are also grateful to Jonathan Gagnon and Jean-Éric Tremblay for the nutrient data, and to Ramon Terrado colleagues from ICM, CSIC Barcelona and Marianne Potvin in carrying out some of the sampling and laboratory work.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2017.00016/full#supplementary-material

improves color calling in SOLiD sequencing. Nat. Methods 7, 335–336. doi: 10.1038/nmeth.f.303


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Onda, Medrinal, Comeau, Thaler, Babin and Lovejoy. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Models of Plankton Community Changes during a Warm Water Anomaly in Arctic Waters Show Altered Trophic Pathways with Minimal Changes in Carbon Export

Maria Vernet <sup>1</sup> \*, Tammi L. Richardson<sup>2</sup> , Katja Metfies <sup>3</sup> , Eva-Maria Nöthig<sup>3</sup> and Ilka Peeken<sup>3</sup>

1 Integrative Oceanography Division, Scripps Institution of Oceanography, La Jolla, CA, United States, <sup>2</sup> Department of Biological Sciences and School of the Earth, Ocean, and Environment, University of South Carolina, Columbia, SC, United States, <sup>3</sup> Department of Polar Biological Oceanography, Alfred Wegener Institute, Bremerhaven, Germany

#### Edited by:

Connie Lovejoy, Laval University, Canada

Reviewed by: Robert McKay, Bowling Green State University, United States Douwe Maat, Royal Netherlands Institute for Sea Research (NWO), Netherlands

> \*Correspondence: Maria Vernet mvernet@ucsd.edu

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Marine Science

Received: 22 November 2016 Accepted: 12 May 2017 Published: 31 May 2017

#### Citation:

Vernet M, Richardson TL, Metfies K, Nöthig E-M and Peeken I (2017) Models of Plankton Community Changes during a Warm Water Anomaly in Arctic Waters Show Altered Trophic Pathways with Minimal Changes in Carbon Export. Front. Mar. Sci. 4:160. doi: 10.3389/fmars.2017.00160 Carbon flow through pelagic food webs is an expression of the composition, biomass and activity of phytoplankton as primary producers. In the near future, severe environmental changes in the Arctic Ocean are expected to lead to modifications of phytoplankton communities. Here, we used a combination of linear inverse modeling and ecological network analysis to study changes in food webs before, during, and after an anomalous warm water event in the eastern Fram Strait of the West Spitsbergen Current (WSC) that resulted in a shift from diatoms to flagellates during the summer (June–July). The model predicts substantial differences in the pathways of carbon flow in diatom- vs. Phaeocystis/nanoflagellate-dominated phytoplankton communities, but relatively small differences in carbon export. The model suggests a change in the zooplankton community and activity through increasing microzooplankton abundance and the switching of meso- and macrozooplankton feeding from strict herbivory to omnivory, detritivory and coprophagy. When small cells and flagellates dominated, the phytoplankton carbon pathway through the food web was longer and the microbial loop more active. Furthermore, one step was added in the flow from phytoplankton to mesozooplankton, and phytoplankton carbon to higher trophic levels is available via detritus or microzooplankton. Model results highlight how specific changes in phytoplankton community composition, as expected in a climate change scenario, do not necessarily lead to a reduction in carbon export.

Keywords: phytoplankton, flagellates, food web, carbon cycling, inverse model

# INTRODUCTION

The Arctic Ocean is one region where climate change is most pronounced, impacting the pelagic environment with observed effects on stratification, pH and currents. The consequences of these effects on phytoplankton are complex. Spatial shifts in latitude as well as timing of biological events affect phytoplankton bloom phenology, microalgal species distribution and trophic interactions (Aberle et al., 2012). A third major effect is the decrease in cell size distribution (Peter and Sommer, 2012), the focus of this study. A decrease in cell size can come about by direct effects of the environment on phytoplankton, e.g., higher temperature increasing metabolism, or indirectly, where environmental conditions alter grazing pressure on phytoplankton abundance, composition and cell size (Winder and Sommer, 2012). In the Arctic, warmer climate increases stratification, with warmer and less saline mixed layers, lower nitrate concentrations and higher picoplankton abundance (Li et al., 2009). Similarly, reduced sea ice cover in Lake Erie has been associated with smaller-sized cells that attain lower total biomass than during periods of ice cover that instead promotes chain-forming diatoms (Beall et al., 2016). Temperature could affect cell size of a given species or may facilitate larger vs. small species abundance, or both. However, this is not a given (Rüger and Sommer, 2012). Alternatively, it has been proposed that at higher temperatures grazing could intensify in a size-selective mode, affecting phytoplankton cell size distribution by top-down controls. Results are variable, with no cell size changes observed at higher temperatures (Rüger and Sommer, 2012 but see Daufresne et al., 2009) or grazing causing a reduction in cell size (Peter and Sommer, 2012). Although the importance of the grazers in the food chain is considered key to sedimentation (e.g., Reigstad et al., 2011) there is no large-scale consensus that small cells contribute substantially to sedimentation (but see Richardson and Jackson, 2007). Large zooplankton (e.g., Calanus spp.) feeding on the phytoplankton spring bloom, usually dominated by large cells, is known to produce a pulse of sedimentation through fecal pellet formation (Forest et al., 2010). Within this paradigm, it is expected that an absence of large cells, i.e., diatoms, will decrease the flux of material to the sediments (Wohlers et al., 2009).

In the Arctic, Atlantic water coming from the south becomes the West Spitsbergen Current (WSC); west of Svalbard, this current has a subsurface core at about 250 m depth and a surface expression. The current brings 6.6–8.5 Sv (or 10<sup>6</sup> m<sup>3</sup> s −1 ) with a northward flow (Beszczynska-Möller et al., 2011). A cooling occurs as the water moves north, losing heat at the surface in contact with the atmosphere as well as sub-surface cross-front exchange with fresher and colder water from sea ice and/or glacial melting (Rudels et al., 2005). The Atlantic water cooling and freshening as it is transported north has a 5-to-6-year cycle in its salinity and temperature properties. Temperatures >2 ◦C, with a mean temperature in the WSC of 3.1 ± 0.1◦C characterize the Atlantic water at these latitudes (Beszczynska-Moller et al., 2012). Only one third of the heat carried by the WSC is transported into the Arctic Ocean, the rest is lost in westward transport and sea surface cooling (Kawasaki and Hasumi, 2016). In the 1997–2010 period, the trend is one of increased temperature but no significant change in volume transport (Beszczynska-Moller et al., 2012). The Warm Water Anomaly in 2005–2007 was defined as a northward advance of Atlantic water, a warm tongue more than 350 km north, reaching the Fram Strait northwest of Svalbard with waters 1◦C higher than average (Walczowski et al., 2012).

The observed biological changes in eastern Fram Strait, and their implication for the Central Arctic Ocean, were tightly coupled with changes in the hydrography. Although the WSC shows pronounced inter-annual variability in primary productivity, phytoplankton and zooplankton abundance and composition (Wassmann et al., 2010; Carstensen et al., 2012; Kwasniewski et al., 2012), large changes in phytoplankton and zooplankton were associated with the warm water anomaly from 2005 to 2007 (Beszczynska-Moller et al., 2012; Nöthig et al., 2015; Soltwedel et al., 2016). This is best reflected in the longterm data set of the HAUSGARTEN observatory at 79◦N, 4◦E (Long-Term Ecological Research in the deep Arctic Ocean) that demonstrated a shift in phytoplankton community structure and in the composition of the sedimenting particulate carbon (Alcaraz et al., 2010; Lalande et al., 2013). The main diatoms found in the Atlantic waters of the WSC before the warm water event were large centric or chain-forming species, including Thalassiosira spp., Chaetoceros spp. (very often Chaetoceros socialis), chains of pennate diatoms of the genus Fragilariopsis spp., Navicula spp., Achnanthes taeniata and Fossula arctica in different proportions. Sometimes a few Rhizosolenia spp, Nitzschia/Pseudonitzscha sp., or Cylindrotheca sp., were observed (Degerlund and Eilertsen, 2010). At the time of the warm water pulse, higher phytoplankton biomass was observed in the water column, protistan plankton >3 µm changed in composition, and diatoms that dominated the period before the warm event switched to a dominance by coccolithophores in 2004, followed by Phaeocystis pouchetii dominance in 2006 (Nöthig et al., 2015). Several of these changes remained after the warm-water event, with Phaeocystis sp., still being prominent in the community (Metfies et al., 2016), although there has been a decrease in water temperature and in Phaeocystis sp. abundance from 97 to 48% from 2007 to 2011 (Soltwedel et al., 2016) whereas diatom concentration remained low and nanoflagellates increased to 43% (Nöthig et al., 2015). The ecosystem responded to the observed pelagic changes: there was an increase in food availability to the benthos in 2006–2007 when Phaeocystis sp., and flagellates dominated the overlying plankton community, which altered the abundance and community structure of the benthic bacteria and meiofauna, while macrofauna response lagged by a year (Jacob, 2014; Soltwedel et al., 2016).

Biological changes in the Fram Strait might foreshadow expected future changes in the Central Arctic, as this is the largest sub-Arctic water feeding the Arctic Ocean. In fact, what was observed in the Fram Strait during the warm period is seen throughout the Arctic Ocean and Arctic Seas: an increase of 20% in phytoplankton productivity due to more ice-free days during the growth season (e.g., Arrigo and van Dijken, 2011), a decrease in phytoplankton cell size associated with freshening and nitrate depletion in the mixed layer (Li et al., 2009), and changes in bloom phenology, both by an early sea ice retreat and late summer blooms (Kahru et al., 2011; Harrison et al., 2013; Ji et al., 2013; Ardyna et al., 2014).

In this study, we used a combination of linear inverse modeling and ecological network analysis to characterize and quantify the pathways of carbon flow through pelagic food webs of the eastern Fram Strait. We were particularly interested in how variations in phytoplankton community composition and in cell size before, during, and after the anomalously warm period of 2005–2007 affected the ecosystem trophic dynamics, including the transfer of carbon to higher trophic levels (planktivorous fish and cod) and export of carbon out of surface waters.

#### METHODS

#### Model Construction

We constructed food webs for the WSC region of the eastern Fram Strait (**Figure 1**) using published data for the late spring/summer of 2003 (before the warming event), 2006 (during), and 2010 (after), or as close as possible to the time period (but always within 1 year). The same model structure was used for each time period (**Figure 2**). Each web comprised 42 flows that represented carbon flows between two compartments, or from one compartment to a sink (**Table 1**). The structure of the food webs was based on the assumption that sizes of the producers and consumers were major determinants of the trophic dynamics of these systems, i.e., small grazers are restricted to small algae. Choices of compartments and trophic relationships were a compromise between achieving biological reality and keeping the total number of flows in the system reasonable. The living components included two phytoplankton compartments, three zooplankton compartments, one compartment for small planktivorous fish, one for cod, and one compartment for heterotrophic bacteria. The phytoplankton were divided into "small" (0.2 to ∼10 µm; assumed to be mainly picophytoplankton, coccolithophores, Phaeocystis sp., and small autotrophic flagellates) and "large" (> 10 µm; mainly diatoms and larger dinoflagellates; Kilias et al., 2014; Nöthig et al., 2015). Zooplankton size classes were the microzooplankton (20–200µm; ciliates and flagellates), the mesozooplankton (200 to ∼1,000 µm; mainly small copepods) and macrozooplankton (chaetognaths, euphausiids, and Calanus copepods >1,000 µm; Bamstedt et al., 1991; Hop et al., 2006; Blachowiak-Samolyk et al., 2007; Calbet, 2008; Pasternak et al., 2008; De Laender et al., 2010; Svensen et al., 2011; Monti and Minocci, 2013). Small planktivorous fish were assumed to be mainly capelin and herring but this compartment also includes carnivorous zooplankton, such as amphipods (Wassmann et al., 2006; Dalpadado et al., 2016). The top predator in the system was cod (Wassmann et al., 2015)

All living compartments contributed to a labile dissolved organic carbon (DOC) pool through excretion and to the detrital pool through mortality or defecation. Sloppy feeding was implicitly included as excretion to DOC. Detritus was transformed to DOC by chemically- or bacterially-mediated dissolution (Jumars et al., 1989). All living compartments lost carbon by respiration. Other forms of mortality (e.g., viral lysis or natural cell mortality) were implicitly included in flows to detritus and DOC. All non-respiratory losses from the system were represented by flows to an "external" compartment that served as a mathematical closure term. These losses included particulate organic carbon (POC) export by detrital settling, DOC loss by advection, and removal of cod through the fishery or via consumption by higher trophic levels.

#### Data

We used published data from Fram Strait to calculate input ("known") values for 7 of the 42 carbon flows: small and large phytoplankton primary productivity, bacterial productivity,

arrows); blue arrows indicate the outflow of polar water in the East Greenland Current in the western Fram Strait. The center of the star is at 79◦N and 4◦E indicating the source of the data used in this study, HG is HAUSGARTEN observatory. (Map was produced with ArcGIS 10.3 using GEBCO 08, modified by Laura Hehemann from the Alfred Wegener Institute, Germany).

microzooplankton grazing on small phytoplankton and flagellates, microzooplankton grazing on large phytoplankton, ingestion rates for the small fish and ingestion rates for cod (**Table 2**). Total primary productivity rates were taken from remote sensing estimates by Arrigo et al. (2008), Arrigo and van Dijken (2011, 2015) and daily production was estimated assuming annual primary production was evenly distributed over the period of open water. Contributions of flagellates vs. large phytoplankton to total primary productivity were assumed to be proportional to their size-specific contributions to biomass and were calculated from chlorophyll a measurements and phytoplankton community composition data of Nöthig et al. (2015, their Table 2). Accordingly, small phytoplankton and flagellates constituted 20% of the total phytoplankton biomass for the before models and 80% were considered large phytoplankton; Phaeocystis sp. and other flagellates accounted for 97 and ∼50% of the small phytoplankton biomass for the during and after models, respectively.

Conversions from chlorophyll a to carbon units were done using an average C:chl ratio of 53 (g:g) to avoid seasonal biases (Svensen et al., 2011). The C:chl ratio of 53 for phytoplankton was chosen as a compromise between spring and summer values and those for small and large phytoplankton cells as shown by mesocosm experiments in which C:chl ratio of diatoms, dinoflagellate and mixed composition were 95, 45, and 60, respectively (Svensen et al., 2011; Spilling et al., 2014). Similarly, phytoplankton in the Fram Strait in May and August 2014 had a C:chl ratio of 41 (Marit Reigstad, personal communication). The ratio of 53 was used when converting from chl a estimates in the field (in mg chla m−<sup>2</sup> ) to phytoplankton carbon (mg C m−<sup>2</sup> ). A different C:chl ratio would either increase or decrease phytoplankton biomass in the model constraints but has no effect where phytoplankton biomass is an unknown.

There is difficulty in obtaining reliable and consistent data in high latitude environments due to the effort and cost of such studies. The data from Nöthig et al. (2015) as well as other field campaigns are based on cruises of a few weeks length, in the June and July time period, with the exception of estimates of fish abundance, provided by year. In this way, it is possible to compare year-to-year summer variability. Data from cruises from other times of the year (either April–May or August–September) were not included in this study.



Flow symbols are used in Figure 1. Flows for which data were used directly (as knowns) are shown in bold; the inverse approach was used to calculate all other flows. Units are mg C m−<sup>2</sup> d −1 . Values presented are means ± standard deviations of 10,000 runs of each model.

Bacterial productivity values were calculated by multiplying bacterial abundance data by a C-specific production of 0.109 ± 0.89 d−<sup>1</sup> (L. Seuthe, personal communication) an integrated over 0–45 m during cruises to NW Spitsbergen in 2014 (n = 7), and a biomass of 10 fg C cell−<sup>1</sup> (Fukuda et al., 1998). Microzooplankton grazing rates were estimated from Verity et al. (1999, 2002) in the Barents Sea and Calbet et al. (2011) in the Fram Strait. For the before model, we assumed that microzooplankton grazing would be higher on small phytoplankton and flagellates (0.2 d −1 ) than on the larger phytoplankton (0.05 – 0.1 d−<sup>1</sup> ) based on Strom et al. (2001). In contrast, for the models of during and after the warm period, when Phaeocystis sp. dominated the phytoplankton community, we assumed that microzooplankton grazing rates on small phytoplankton and Phaeocystis sp. were also low (0.05 d−<sup>1</sup> for during and 0.1 d−<sup>1</sup> for after) and no more than 8% of the phytoplankton standing stock based on previous studies (see also Caron et al., 2000; Calbet et al., 2011). After the warm period, grazing of microzooplankton on non-Phaeocystis sp. was 0.2 d−<sup>1</sup> . Ingestion rates for the small fish and cod were calculated from annual fish biomass in ICES ASWG 2014 and a conservative C-specific ingestion rate of 0.017 d−<sup>1</sup> for cod (range of 0.017–0.057 d−<sup>1</sup> , De Laender et al., 2010) and an average Cspecific ingestion rate of 0.04 d−<sup>1</sup> for capelin and herring (range of 0.01–0.1 d−<sup>1</sup> , Ajiad and Pushchaeva, 1992; Megrey et al., 2007).

Sources of biomass for compartments are detailed in **Table 3**; these data are used to formulate the constraints used to set bounds on the flows predicted by the model (see Section Inverse Analysis below) (**Table 4**). The conversion factor of 0.132 was used to estimate carbon from wet weight in fishes (Sakshaug et al., 1994). Microzooplankton biomass was estimated from cell counts by the conversion factors of Verity and Lagdon (1984) and Menden-Deuer and Lessard (2000).

#### Inverse Analysis

The linear inverse modeling approach of Vézina and Platt (1988) was used in conjunction with the Monte Carlo solutions approach of Donali et al. (1999) for estimating the range of values for all flows in our constructed food webs. Model code was run in Matlab R2011b and was kindly provided by Dr. Nathalie Niquil (Centre National de la Recherche Scientifique, Caen, France). The approach taken assumes that biomass in any compartment is in steady state, i.e., the total flows entering any compartment are equal to the flows leaving it without any accumulation or decrease (with the exception of the "external" compartment), although modifications to the approach can be made to accommodate

TABLE 2 | Rates used as "known" flows for the inverse analysis, in units of mg C m−<sup>2</sup> <sup>d</sup> −1 .


Values were derived using information in the source materials according to the methods described in the text.

TABLE 3 | Biomass values (mg C m−<sup>2</sup> ) used for the formulation of constraint equations for the inverse analysis.


Values were derived using information in the source materials according to the methods described in the text. For the phytoplankton, chlorophyll values were taken from Nöthig et al. (2015) and were converted to carbon biomass using a C:chl ratio of 53 (g:g; Svensen et al., 2011). Size fractions were apportioned according to Nöthig et al. (2015) (see text for details).

#### TABLE 4 | Constraints on carbon flows for the inverse analysis.


GPP, gross primary productivity; NPP, net primary productivity; DOC, dissolved organic carbon. Values used for carbon content (W) were 6.3 fg C cell−<sup>1</sup> for bacteria (Kawasaki et al., 2011), 1.7 pg C individual−<sup>1</sup> for microzooplankton, 2214 pg C individual−<sup>1</sup> for mesozooplankton and 2.31 × 10<sup>8</sup> pg C individual−<sup>1</sup> for the macrozooplankton (Bamstedt et al., 1991). Temperatures were assumed to be 3.5, 5, and 4◦C for before, during, and after the warm anomaly, respectively (Beszczynska-Moller et al., 2012).

non-steady state scenarios by allowing residual flows to balance the system (e.g., Richardson et al., 2003).

As described above, data from the scientific literature were used to formulate 7 input equations. Combined with the 10 mass balance equations (one for each compartment; see **Table 5**), there were 17 equations available to describe the system with 42 flows. We reduced the number of possible solutions for this underdetermined system by applying a set of biological constraints (provided in **Table 4**). Allometric constraints based on published relationships incorporated available biomass data and provided upper and lower bounds on the rates and efficiencies of biological processes. For example, the respiration of all phytoplankton was constrained to be at least 5% but no more than 30% of the gross primary productivity (GPP) (Vézina and Platt, 1988). Growth efficiencies were assumed to be 25–50% of ingestion for the zooplankton groups (Straile, 1997). Bounds on assimilation efficiencies for all grazers were 50–90% of ingestion for the microzooplankton (Vézina and Platt, 1988; Straile, 1997) and 50–80% for the macrozooplankton (Straile, 1997). We also set a lower bound on the macrozooplankton grazing such that they consumed at least 30% of the large phytoplankton productivity (based on Wassmann et al., 2006). Other constraints are detailed in **Table 4**. We used temperatures of 3.5, 5, and 4◦C for the before, during and after models, respectively (Beszczynska-Moller et al., 2012).

Application of constraints reduces the range of possible solutions, but does not provide a unique solution. The Monte Carlo approach of Donali et al. (1999) (see also the review by Niquil et al., 2012) calculates 10,000 possible solutions for each set of flows, thus we were able to calculate both an average and a standard deviation for each flow in the food web.

#### Econetwork Analysis of Inverse Solutions

After food webs were constructed for the before, during and after warm water event, the structure and function of each web was assessed using EcoNetwork analysis software (available at https://www.cbl.umces.edu/~ulan/ntwk/network.html; see also Ulanowicz and Kay, 1991; Ulanowicz, 2004). Michaels and Silver (1988), Ducklow et al. (1989), and McManus (1991) used earlier versions of this program to examine flows of nitrogen and energy, respectively, through microbial food webs to higher trophic levels in planktonic systems. We chose a key index from the output, input/export vectors, to calculate how much of each input flow (i.e., primary production of the two phytoplankton groups) eventually was exported through the three possible routes of export (via cod, detritus, or DOC). This calculation allowed us to break down the export flows into the relative contributions by the flagellate (small) vs. diatom (large) phytoplankton.

#### RESULTS

Food webs constructed for before, during, and after the warm anomaly (**Figure 2**) differ substantially with respect to the input flows (contributions by the small phytoplankton and flagellates vs. large phytoplankton), trophic transformations, and predicted export pathways (**Figure 2**, **Table 1**). While diatoms (or large phytoplankton) dominated primary productivity before the warm anomaly in the WSC, Phaeocystis sp. accounted for 97% of the primary productivity during the warm years, and small phytoplankton (Phaeocystissp. and flagellates) continued to dominate for almost 4 years after the peak in water temperature. The dominance of Phaeocystis sp. and low grazing of this material by the microzooplankton during the warm water event (see also Calbet et al., 2011) resulted in the model prediction of more carbon from flagellates (or small phytoplankton) going to detritus, which then became an important source of food for mesozooplankton or macrozooplankton (**Figure 3**).

In general, the zooplankton diet in the model reflected the dominant phytoplankton community composition (**Figure 3**). When large phytoplankton dominated before the warm water event, the microzooplankton fed equally on small and large phytoplankton, but consumed mostly carbon originating from small cells and flagellates in the during and after periods. The mesozooplankton diet changed during and after the warm anomaly to rely more heavily on detritus than on the large phytoplankton, while the carbon flow increased

TABLE 5 | Mass balance equations (inputs–outputs = 0) for the inverse analysis.


Sph, small phytoplankton; Lph, large phytoplankton; mic, microzooplankton; mes, mesozooplankton; mac, macrozooplankton; fsh, small fish; cod, Cod fish; bac, bacteria; doc, dissolved organic carbon; res, respiration; det, detritus. gSp and gLp are the gross primary productivity of the small and large phytoplankton, respectively. Ext refers to the export of material to an external compartment (out of the ecosystem).

from <100 mg C m−<sup>2</sup> d −1 to >150 mg Cm−<sup>2</sup> d −1 . In contrast, macrozooplankton carbon flow was predicted to remain somewhat constant through the three periods albeit important changes in the quality of diet, from mostly feeding on large phytoplankton, to a mixed diet where 50% of the carbon originated from the abundant mesozooplankton, and an even more mixed diet after the warm water event, with approximately equal consumption of large phytoplankton, detritus, microzooplankton and mesozooplankton (**Table 1**, **Figures 2**, **3**).

Carbon in the form of detritus dominated the export fluxes, and was generally higher in the before period than during or after the warm event (**Figure 4**, **Table 1**). Carbon that originated from the diatoms dominated detrital export before the warm anomaly (86% of the total detritus export flux vs. 14% from the flagellates), but carbon from flagellates comprised the majority of the carbon exported as detritus during (96% small, 4% large) and after (89% small, 11% large) the warm water anomaly (**Figures 4**, **5**).

Calculated rates of respiration were dominated overall by the small phytoplankton and the microzooplankton (**Figure 6**). Microzooplankton and macrozooplankton respiration were maximum before, and mesozooplankton respiration rates were highest in during the warm water event. Only planktivorous fish and bacteria showed maximum respiration in after period. The largest changes were from before to during in small phytoplankton and mesozooplankton.

#### DISCUSSION

Carbon flow through pelagic food webs is impacted by the composition and biomass of the primary producers, the phytoplankton. It is expected that Arctic environmental change

in response to climate warming will lead to modifications in phytoplankton species composition, cell size and biomass (Daufresne et al., 2009; Winder and Sommer, 2012). Consequences of these changes must be elucidated via field studies and modeling. We selected a modeling approach to study possible consequences in carbon cycling in the WSC, Fram Strait ecosystem from observed phytoplankton changes (Nöthig et al., 2015). Inverse modeling can synthesize and test current understanding in a system as well as provide a first approximation of carbon flows for which data are sparse. It has been applied in high latitude waters, mainly the Barents Sea (De Laender et al., 2010), the Amundsen Bay in the Canadian High Arctic (Forest et al., 2010), and in the western Antarctic Peninsula (Daniels et al., 2006; Sailley et al., 2013). The results from the model presented here are considered hypotheses about how the phytoplankton carbon could cycle in Arctic regions subject to shifts in phytoplankton composition.

The changes in carbon flow predicted by the inverse model from before to after warm water conditions parallel the changes modeled by Rivkin et al. (1996) in food webs from the Gulf of St. Lawrence. There, the structure of the food web changed from the colder spring bloom period, when large phytoplankton and herbivory by mesozooplankton dominated, to a warmer summer, microbial-dominated food web, but the amount of carbon exported in the cold vs. the warm periods was not substantially different. In summer, the mesozooplankton switched from eating large phytoplankton to consuming mainly microzooplankton. Thus, while the pathway of carbon through the system changed, the POC export flux did not. The Fram Strait system responded similarly; grazing was replaced by omnivory during warm water periods in the absence of large diatoms (**Figure 3**). The generality of the response is more striking as the post-bloom condition in the Gulf of St. Lawrence and the eastern Fram Strait were different: the first one was dominated by dinoflagellates where in the latter, diatoms were replaced by Phaeocystis sp. and nanoflagellates.

The inverse model suggested a scenario predicted already by Weisse et al. (1994) for the North Sea Phaeocystis sp. blooms. This author speculated that mesozooplankton, especially small copepods like Acartia and Temora, may indirectly benefit from Phaeocystis sp. blooms by feeding on detrital particles or microzooplankton. Vast amounts of detritus appear in the form of marine snow during and after Phaeocystis sp. blooms that, when coated with bacteria and microheterotrophs, is considered nutritious (Heinle et al., 1977). This new scenario of trophic pathways is not commonly found in the literature, as very little is known about pelagic detritivory. It has been proposed that zooplankton ingestion of detritus breaks up marine snow particles, facilitating bacterial degradation that, in turn, increases the nutritional value of the organic matter (Mayor et al., 2014). Although these authors focus their hypothesis on the water column below the euphotic zone, similar processes could occur in the mixed layer.

#### Response of Phytoplankton to the Warming Event in Eastern Fram Strait

The prediction of the proliferation of small cells in a future warmer ocean is usually associated with flagellates, whereas larger cells are presumed to be diatoms (Li et al., 2009, but see Wright et al., 2010). In the WSC, the transition of phytoplankton communities exposed to elevated temperatures can be more complex: the prymnesiophyte Phaeocystis sp. is a flagellate that can form both large colonies as well as single cells (Rousseau et al., 2000). In the WSC, the warming event was associated with a proliferation of this microalga from the summer of 2005 onwards, i.e., different clades of Phaeocystis sp. were dominant during and after the warm water event from 2005 to 2007 (Nöthig et al., 2015). This species is not new to the Fram Strait, commonly found in the region in spring and summer (Smith, 1987; Hegseth and Tverberg, 2013; Saiz et al., 2013). P. pouchetti slowly decreased in concentration after 2007 with an average 45% concentration after the warm water event. A major consequence of this shift from diatoms to Phaeocystis sp. is the change in grazing pressure; Phaeocystis sp. experiences less grazing than other flagellates (Caron et al., 2000; Strom et al., 2001; Calbet, 2008; Calbet et al., 2011).

Blooms of single-cell Phaeocystis sp. are of widespread distribution, with individual cells of 3–7 µm (Vernet et al., 1996; Kozlowski et al., 2011; Metfies et al., 2016) although it is generally considered that Phaeocystis sp. blooms in its colonial form, with colonies in excess of 100 µm and up to 1000 µm (Schoemann et al., 2005; Lasternas and Agusti, 2010). Single-cell Phaeocystis sp. in the eastern Fram Strait was observed in 2012 after the warm event in the <3 µm phytoplankton size fraction (Metfies et al., 2016). The proportion of P. pouchetii in colonial or in singlecell form at the WSC in 2005–2007 is unknown; both forms were modeled in this study (Stelfox-Widdicombe et al., 2004).

The model provides realistic estimates of phytoplankton growth rates, approximated from C-specific primary production. Flagellates grew at an average 0.2, 0.13, and 0.28 d−<sup>1</sup> and large cells at an average 0.21, 0.39, and 0.28 d−<sup>1</sup> within the surface layer before, during and after the warm water event, respectively. These rates were not measured, the biomass originated from the field (**Table 3**) and primary production from the model output (**Table 1**), which carry the inherent approximation that both phytoplankton size fractions have equal photosynthetic efficiency.

#### Trophic Pathways during a Shift in Phytoplankton Composition

When diatoms dominated (i.e., cells > 10 µm), the model predicted that phytoplankton were consumed by meso- and macrozooplankton herbivores, with a major carbon flow from fecal pellets to detritus and eventual export (i.e., sedimentation, **Figure 2**). Carbon also flowed to detritus and microzooplankton, accounting for the rest of the large cells, but the contribution was minor based on the small grazing pressure of microzooplankton on diatom blooms (**Figure 3**, Sherr and Sherr, 2009). The inverse model suggested that when Phaeocystis sp. dominated pelagic photosynthetic communities, e.g., by contributing up to 97% of the autotrophic community, phytoplankton carbon goes directly to detritus, and to a lesser extent to microzooplankton and DOC production (**Figure 3**). Mesozooplankton fed mostly on the detrital carbon, originating from phytoplankton sinking and fecal pellets, while a large proportion of the detritus was also exported. In this way, the mesozooplankton role in the food web increased when flagellates dominated but macrozooplankton role stayed rather constant, as in the case of microzooplankton. Total export out of the system decreased by 15%, due to a 35% diminution in POC export while DOC export increased (**Figure 4**). When the phytoplankton community bounced back to more diatoms, but still with ∼45% Phaeocystis sp., overall export of particulate carbon also recovered while DOC contribution remained high (**Figure 4**). Additionally, the increase in microzooplankton and bacterial abundance during and after the warm water period indicate an increase in substrate, as expected from Kirchman et al. (2009a,b). The modeled sedimentation flux, where the mixed phytoplankton community composed of diatoms, Phaeocystis sp. and nanoflagellates exported as much carbon as diatoms alone, was also expected from the sediment trap data of Lalande et al. (2013) that showed that fluxes remained the same and only the quality changed. However, these model predictions are novel and will be discussed further.

#### Detritus Formation

The results of the inverse model suggest that when Phaeocystis sp. dominated the phytoplankton community a large proportion of the biomass was not consumed by grazers and was lost to other processes, mostly routed through detritus (e.g., marine snow), DOC production and respiration. The model did not predict high DOC production (**Table 1**, see below for further discussion) and the respiration changed based mostly on the amount of carbon cycling each through each compartment (compare **Figures 2**, **3** with **Figure 6**) and to lesser extent to higher ambient temperature; the remaining possibility was for the carbon to flow to detritus as marine snow. Most of the detritus originated from phytoplankton sinking or coagulating, from 148 mg C m−<sup>2</sup> d −1 from diatoms in the before conditions to 313–348 mg C m−<sup>2</sup> d −1 in the during and after conditions, mostly from Phaeocystis sp. (**Table 1**). Due to lack of data, the model does not have rigid constraints for this flow that at its highest reached 50% of the Phaeocystis sp. primary production in the during and after time periods (compare flows 5 and 1 in **Table 1**), whereas 26% of the diatom primary production was converted to detritus in the before conditions. These results suggest a doubling of the phytoplankton-detritus flow during the warm water event compared to the before (diatom) conditions, similar to observations of high concentration of marine snow in the North Sea during and after Phaeocystis sp. blooms (Lancelot and Mathot, 1987; Riebesell et al., 1993).

#### Microzooplankton Grazing

High microzooplankton grazing in summer/post-bloom/during conditions rich in flagellates has been well documented in the field (Vernet, 1991; Verity et al., 2002; Calbet and Saiz, 2005) and grazing efficiency of the microzooplankton can be lower when feeding on large phytoplankton cells (Strom et al., 2001). In the model, microzooplankton grazed on equal amounts of diatoms and small cells before 2003. During and after the warm water event, this compartment grazed mainly on Phaeocystis sp., maintaining their overall carbon intake (flow 4, **Table 1**). Microzooplankton consumed 235, 236, and 224 mg C m−<sup>2</sup> d −1 in the form of large and small phytoplankton and bacteria in the before, during, and after conditions (**Table 1**), corresponding to a grazing rate of ∼0.08, ∼0.055, and ∼0.1 d−<sup>1</sup> . These grazing rates are lower than what was observed in the Barents Sea for non-Phaeocystis phytoplankton (0.24 ± 0.1, 0.29 ± 0.13, 0.33 ± 0.11 d−<sup>1</sup> , Verity et al., 2002), and within the median value of Calbet et al. (2011) during a Phaeocystis sp. bloom (observed range of −0.04–0.14 d−<sup>1</sup> ) in the Fram Strait. The explicit grazing inhibition by Phaeocystis sp. in the model (see Methods) is found not only in the Fram Strait but also in Antarctica (Caron et al., 2000) and elsewhere (Strom et al., 2001). In their review, Nejstgaard et al. (2007, Table 4) report grazing rates of 0.0–0.36 d−<sup>1</sup> on solitary Phaeocystis sp. cells (3–8 µm). For field observations, the same authors report microzooplankton grazing was positive in April 2003 (0.21 ± 0.3 d−<sup>1</sup> ) and negative in May 2004 (−0.23 ± 0.34 d−<sup>1</sup> ). Without detailed knowledge of the factors affecting Phaeocystis sp. grazing in the WSC after 2004, the rates in the model are in the middle of the range found in the literature, and thus considered conservative. Further estimates of microzooplankton grazing, in particular in large phytoplankton and during Phaeocystis sp. blooms in the Arctic, are needed in order to improve our model parameterizations and our understanding of the fate of Phaeocystis sp. carbon through this compartment in the food web.

Inhibition of microzooplankton grazing by Phaeocystis sp. is similar to observations on other Ecosystem Disruptive Algal Blooms and Harmful Algal Blooms (EDABs and HABs). Acrylic acid, released by Phaeocystis sp. in the conversion from dimethylsulfoniopropionate (DMSP) to dimethylsulfide (DMS), is considered an antibiotic (Sieburth, 1960) and other growth and grazing inhibitors could be released by this species as well (Nejstgaard et al., 2007, but see Turner, 2015). The production of toxins by phytoplankton has lethal or sub-lethal effects on the microzooplankton, both for ciliates or tintinnids (Verity and Stoecker, 1982; Carlsson et al., 1990; Hansen, 1995). Rosetta and McManus (2003) concluded that ciliates may exert grazing pressure on HAB species early on, potentially contributing to the suppression and decline of Prymnesium minimum and P. parvum before they bloomed, but that ciliate grazing would be relatively ineffective once blooms (and toxicity) developed fully. In mixed diets, as long as non-toxic cells were available, ciliates survived and sometimes grew well at concentrations that otherwise would have killed them. Similar for rotifers, when exposed to a mixed diet of toxic and non-toxic phytoplankton species, these protists would tolerate and even acclimate to a toxic species (e.g., Karenia brevis), supporting the notion of low but positive grazing rates when Phaeocystis sp. was dominant (**Table 2**).

#### Grazing by Copepods

Grazing of meso- and macrozooplankton on Phaeocystis sp. depends on multiple environmental factors and it is not predictable. Nejstgaard et al. (2007) conclude in their review on grazing impacts on Phaeocystis sp. that small copepods cannot feed on colonies whereas macrozooplankton can. It has been observed that Arctic copepods do not avoid surface waters during Phaeocystis sp. blooms (Norrbin et al., 2009). However, Saiz et al. (2013) reported that under these conditions the copepod ingestion rate was low in spite of positive grazing rates, making a low impact on phytoplankton standing stocks. In this way, same as with microzooplankton grazing, Phaeocystis sp. seems to deter herbivory of larger zooplankton.

#### **Mesozooplankton grazing**

Grazing by mesozooplankton on diatoms, flagellates and detritus was set by the model within the constraints in this compartment on assimilation efficiency, respiration, excretion and growth gross efficiency (**Table 4**). Mesozooplankton consumed diatoms and detritus in the before conditions; in the absence of diatoms this compartment could decrease or consume more detritus. The model predicted detritivory, with an overall increase in mesozooplankton abundance (**Figure 3**, **Table 1**). Overestimation of detritivory with respect to other mesozooplankton feeding behavior by the model is possible due to the lack of constraints on this flow. As an alternative, mesozooplankton could consume more microzooplankton (Stoecker and Capuzzo, 1990; Rivkin et al., 1996). This pathway was not explicit in the model (**Figure 1**) as flows were limited to those identified as most important in the Fram Strait literature, where small copepods are considered of minor importance (Falk-Petersen et al., 2009; Nöthig et al., 2015). However, they might play a major role during the summer (Svensen et al., 2011). Results from the inverse model suggest that mesozooplankton could be an important carbon compartment in this region's food web and their role deserves further study and experimentation.

#### **Macrozooplankton grazing**

Macrozooplankton did not change ingestion on microzooplankton or detritus when Phaeocystis sp. was abundant, rather they increased predation on mesozooplankton. These results contrast with those of De Laender et al. (2010) that predicted higher trophic levels in food webs in the southern Barents Sea, flooded by Atlantic waters from another branch of the Norwegian Atlantic Current, could rely on the microbial loop as a source of carbon, with a doubling of microzooplankton as food source for Calanus spp. copepods. These authors argue that when small zooplankton is dominant during warmer periods, their feeding strategies are more suited to ciliate predation (e.g., Svensen and Vernet, 2016). In the Fram Strait model, macrozooplankton consumed 213, 179, and 196 mg C m−<sup>2</sup> d <sup>−</sup><sup>1</sup> before, during and after the warm water event from diatoms, microzooplankton, mesozooplankton and detritus (**Table 1**, **Figure 3**). In the absence of diatoms, large zooplankton switched their intake to 8x more mesozooplankton, 2x more microzooplankton, but remained rather constant on detritus consumption.

Results from grazing experiments do not present a clear picture on Phaeocystis-zooplankton interactions. In their review, Nejstgaard et al. (2007) found a large variability in grazing rates within the literature, attributed to differences in P. globosa and P. pouchetii strains, cell types, physiological state, etc. In addition to grazing, macrozooplankton has the ability to break up large marine snow aggregates into smaller ones, facilitating their decomposition and increasing their nutrition (Dilling and Alldredge, 2000). The grazing estimates in the inverse model compare well with recent experimental results in the Fram Strait: Hildebrandt (2014) reports an average concentration of 26.6 Calanus finmarchicus per m<sup>3</sup> with a grazing rate of 0.0028–0.014 µg chla h −1 for the summer of 2012; in a 45-m upper layer and assuming 24-h feeding during boreal summer the copepods could consume up to 24 mg C m−<sup>2</sup> d −1 . Similarly, average macrozooplankton grazing rates of 0.089, 0.205, and 0.137 d −1 for Calanus glacialis, C. hyperboreus, and C. finmarchicus, respectively, with an average rate of 0.15 d−<sup>1</sup> , were reported by Weydmann et al. (2014); these copepods could consume 349, 19, and 63 mg C m−<sup>2</sup> d <sup>−</sup><sup>1</sup> before, during and after the warm water event (based on phytoplankton biomass from **Table 3**). These calculations based on Fram Strait experiments and field data agree with results in the North Sea where of P. globosa was not considered a good food source for copepods (Gasparini et al., 2000). In spite of selecting for diatoms and microzooplankton, copepods suffered during a Phaeocystis sp. bloom; copepods consumed 27–50% of the copepod carbon weight per day during diatom dominance that decreased to 7–17% during the Phaeocystis sp. bloom and to 14–21% after the bloom.

#### **Detritivorous copepods**

Detritivorous copepods are usually considered to feed below the euphotic zone. Jackson (1993) suggested that this process could explain the decreased in POC sedimentation in the ocean where only a few percentage of primary production reaches the sediments. "Flux feeding" was proposed as a major carbon flow to complement bacterial degradation of sinking organic matter that could not explain all the carbon reduction with depth. Similarly, Reigstad and Wassmann (2007) measuring recycling of Phaeocystis sp. phytodetritus found that between 7 and 11% of Phaeocystis sp. biomass reaches 40 m depth and only 3 ± 2% reaches 100 m. Assimilation efficiency of zooplankton feeding marine snow in the California Current were 64–83% (Dilling et al., 1998). Similarly, 70% retention of copepod and euphausiid fecal pellet carbon was established in the mixed layers of the Barents Sea thru flux feeding (Wexels Riser et al., 2002), but no studies exist of Arctic copepods and other planktonic organisms consuming sinking phytodetritus (Turner, 2015). For the California Current, Graham et al. (2000) explained diel variability in marine snow concentration in the upper water column to nighttime consumption by vertically migrating zooplankton. There is evidence of high mesozooplankton abundance during periods of Phaeocystis sp. blooms in the North Sea (Fransz and Gieskes, 1984; Weisse et al., 1986). In the Arctic, there is an increasing awareness that small copepods have been undersampled due to large mesh sizes in zooplankton nets. Svensen et al. (2011) argue the best method to sample small copepods quantitatively is with water bottles (e.g., 30-L Niskin). Small copepods have a high growth rate and reproduce year around, are not restricted in their reproduction to the spring bloom as are large copepods and thus can be very abundant year around (Svensen et al., 2011). The model results highlight the possibility that if Phaeocystis sp. is not consumed at a high rate, the mesozooplankton could benefit (as predicted by Weisse et al., 1994). Potential detritivory and the role of mesozooplankton during periods of Phaeocystis sp. dominance are ideas that deserve further study.

#### Bacteria

The model predicts a higher bacterial abundance and activity in warmer periods in the Arctic, when the phytoplankton community is dominated by flagellates, resulting in a more active microbial food web (**Table 1**, **Figure 6**). The bacterial activation occurs parallel to detritus formation, during and after the warm water anomaly. Bacteria decomposition of this detrital material is accounted for in the model (flow 36 in **Table 1**), so bacteria can either benefit from phytoplankton excretion or lysis as DOC, other sources of DOC production, or particulate matter degradation (**Figure 2**). Bacterial production, based on abundance and a C-specific production of 0.1 ± 0.98 d−<sup>1</sup> (Seuthe, pers. commun.), is predicted at 22, 59, and 90 mg C m−<sup>2</sup> d −1 before, during, after the warm water event (flow 33 in **Table 1**). These estimates are within those observed in the region. In the productive waters of Kongsfjorden, a fjord in the western coast of Spitsbergen, Iversen and Seuthe (2011) reported that for 2006, integrated over 0–50 m depth, bacterial production was 105 mg mg C m−<sup>2</sup> d −1 and bacterial respiration 56 mg C m−<sup>2</sup> d −1 . For open waters close to the HAUSGARTEN station, surface bacterial production is highly variable, and was estimated at 2 mg C m−<sup>3</sup> d −1 (or 90 mg C m−<sup>2</sup> d −1 ) between 25 June and 20 July 2011 (Piontek et al., 2014, 2015).

#### DOC Production

In the model, DOC is produced by all living compartments via phytoplankton excretion, by bacterial activity (including detritus) and by zooplankton sloppy feeding (**Figure 1**, **Table 1**). DOC production increased during the flagellate periods, with more DOC produced in the during and after scenarios: from 57.3 to 116.1 mg C m−<sup>2</sup> d −1 and 143 mg C m−<sup>2</sup> d −1 as a result of bacterial activity (flow 36), of grazing by microzooplankton (flow 16), by mesozooplankton (flow 19) and by macrozooplankton (flow 23) (**Table 1**, **Figures 3**, **6**). Due to restrictions in the number of flows (see Methods) viruses and fungi, another potential source of DOC production, were not included explicitly as part of the microbial loop, but their activity was implicit in phytoplankton DOC production. The constraints for this compartment were chosen to allow for "excess" (beyond normal excretion) DOC production, up to 55% of the primary production (**Table 4**). The DOC rates from phytoplankton predicted by the model are toward the low end of this range, 2.1% to 14% for large phytoplankton, and 3–4.1% for Phaeocystis sp. (**Table 1**). These estimates are close to the 10% universal estimate on phytoplankton excretion even in the presence of mucilaginous colonies (Veldhuis et al., 1986), and lower than field measurements in the Arctic of up to 39% (Vernet et al., 1998; Matrai et al., 2007; Poulton et al., 2016). The extent of DOC production by viral lysis in the Arctic is not well characterized. In the North Atlantic, Mojica et al. (2016) reported elevated rates of DOC production from phytoplankton viral lysis, with a "striking reduction" toward high latitudes, where the ratio of viral lysis to grazing decreased by up to two orders of magnitude in comparison to lower latitudes. Fungi are reported to be abundant in sea ice, but have not been found in any quantity in seawater (Hassett and Gradinger, 2016). Metfies et al. (2016), analyzing data for this region, did not detect many fungi (OTUs) in the water column; they were found only occasionally in waters dominated by diatoms. Fungi seem to be mainly associated with marine snow (Bochdansky et al., 2017) and in sediment-trap material (Metfies, in prep.). Increasing the lower limit of DOC production in the model can force more carbon through the DOC compartment which might decrease other loss terms, such as detritus production during and after the warm water event, presumably channeling more carbon through the microbial loop. The complexity of the microbial food web in Arctic waters, including viruses and fungi, requires further experimentation. The model for this study (**Figure 2**) was structured to maximize all the pathways that contribute to carbon sedimentation out of surface waters in the WSC and is thus not the best vehicle to represent a complete picture of microbial processes in this region, which would probably require a model restricted to lower trophic levels only.

#### Carbon Export

What the inverse model in this study provides is a picture of possible carbon pathways within Arctic food webs that could explain how to maintain an important contribution of phytoplankton carbon to deep water in the absence of a diatom bloom. Export of carbon originating from non-diatoms is not surprising (**Figure 4**). Picoplankton and flagellates have been observed in sedimenting matter for the last 30 years: Synechococcus was a major contributor to fluxes to the deep sea (Lochte and Turley, 1988) and has been detected in the South Pacific (Waite et al., 2000), tropical Pacific (Stukel et al., 2013) and in equatorial Pacific, Atlantic and Indian Oceans (e.g., Lampitt et al., 1993). High abundance of Synechococcus has been recently reported for the eastern Fram Strait, included here as cells <10 µm (Paulsen et al., 2016). Similarly, the ubiquitous Micromonas sp. in Arctic waters has been detected in this region by sediment traps, and associated with increased <sup>234</sup>Thorium adsorption in the Central Arctic (Charles Bachy, pers. commun., Roca-Marti et al., 2016). Great quantities of Phaeocystis sp. were observed at the ocean bottom in the Ross Sea, at >500 m depth (DiTullio et al., 2000). In the Fram Strait, highest export was observed during a Phaeocystis sp. bloom, equivalent to the export efficiency, or % of primary production that sediments, by diatoms (Le Moigne et al., 2015). In the Gulf of St. Lawrence, the export efficiency increased from 10% during the diatom bloom to 10–5% in post-bloom conditions (Rivkin et al., 1996). In the Canadian Arctic, flagellates were associated with high export ratios of 0.38–0.69 (Lapoussiere et al., 2013). Similarly, the inverse model in the Fram Strait predicts an export efficiency of 51, 44, and 53.5% before, during and after the warm water event (**Table 1**). In this way, Phaeocystis sp. and flagellates can fuel the biological pump, transferring an important proportion of surface primary production to depth.

Carbon export in the model originated from either diatoms or detritus that during the warm period is overwhelmingly dominated by Phaeocystis sp. carbon (**Figure 5**). By which processes can flagellates contribute to export out of the surface layer? High biomass, stickiness, and presence of ballast all correlate with increased phytoplankton sedimentation by coagulation (e.g., Passow and Alldredge, 1999; Jouandet et al., 2014). In general, diatoms and coccolithophores are considered to sink faster than other phytoplankton and their silicon frustule (opal) or carbonate coccoliths are assumed to act as ballast for phytoplankton sinking and zooplankton fecal pellets, activating the biological pump (Armstrong et al., 2001; Klaas and Archer, 2002; Ploug et al., 2008). Ballast for phytoplankton could originate also from intracellular carbohydrates, minerals or carbonate precipitated within sea ice (Richardson and Cullen, 1995; Iversen and Ploug, 2010). Phaeocystis sp. blooms are reported to have very high sinking rates (Wassmann et al., 1990; DiTullio et al., 2000 but see Schoemann et al., 2005). DMSP, known to be elevated in Phaeocystis sp., has recently been suggested as ballast for this species (Lavoie et al., 2015, but see Boyd and Gradmann, 2002). Coagulation of cells in turbulent environments, in particular species with a sticky surface as observed in senescent Phaeocystis sp. blooms, generates marine snow; this process is considered a widespread venue of removing cells from the upper ocean (Passow and Wassmann, 1994; Logan et al., 1995). Mucus webs of pteropods are also known to be an efficient transport vehicle for pico-plankton particles (Noji et al., 1997). High stickiness in Arctic phytoplankton is expected; diatoms excrete large amounts of polysaccharides (Myklestad, 1995) and Arctic phytoplankton, both diatoms and flagellates, can excrete as much as 70% of their daily primary production as DOC (Vernet et al., 1998; Matrai et al., 2007; Poulton et al., 2016), which can be considered a source of stickiness (Schoemann et al., 2005). Marine snow is part of the detrital carbon, and is difficult to detect and quantify. TEP (transparent exo-polymers) is believed to comprise most of the marine snow and its sinking speed is also related to size, porosity and ballast usually provided by its constituents (Passow, 2002; Bach et al., 2016). For example, porosity of marine snow is lower when flagellates dominate in comparison with diatom-rich aggregates, thus providing another mechanism by which non-diatom aggregates can export carbon (Bach et al., 2016).

The changes in carbon export predicted by the model when Phaeocystis sp. dominated agree in large extent to the observations from sediment traps in the eastern Fram Strait. Flux of (POC) at 179–280 m depth from 2002 to 2008, with 20 sampling cups per year collecting material from 59 days in winter and 7 days in summer, showed POC sedimentation associated with biogenic silica (bSi) pulses (Lalande et al., 2013). Before the fall of 2004, these pulses occurred in spring (April to June), sometimes associated with the ice edge and in the late summer (August to October) due to atmospheric heating of the upper water column. Before 2004 the pulses ranged from 30 to 50 mg C m−<sup>2</sup> d −1 and 10–30 mg bSi m−<sup>2</sup> d −1 . From late 2004 to the summer of 2008, during the warm water event, the consistency of the spring and late summer bSi pulses disappeared, with a few peaks in sedimentation in either May or August (∼10 mg bSi m−<sup>2</sup> d −1 ) remaining and the rest of the time sedimentation was <5 mg bSi m−<sup>2</sup> d −1 (Figure 2f, Lalande et al., 2013). The pulses of POC remained unchanged throughout this period, both in magnitude and time of the year (Figure 2e, Lalande et al., 2013).

Any differences between model predictions and sediment trap data on sedimentation rates are expected, as export in the model represents carbon loss out of the surface layer (<100 m) while the sediment traps were deployed at ∼250 m depth. In the field, the changes in flux of bSi correlated with other important changes in the nature and quality of the sedimenting matter: lower fecal pellet carbon and an increase in the amount of small fecal pellets, attributed to dominance of smaller zooplankton (Lalande et al., 2013). The inverse model predicts the change in quality of sedimenting matter can be attributed to the dominant phytoplankton community, diatoms vs. Phaeocystis sp. and to the changes in trophic pathways in the food web (**Figure 5**, see Section Discussion. for a detailed discussion on carbon flow through the modeled food web). The predicted changes in the biomass of the different compartments are reflected in higher respiration during and after the warm water event for small phytoplankton, mesozooplankton (small copepods) and bacteria and lower respiration from large phytoplankton (**Figure 6**).

Our model results present an alternative hypothesis that warming and flagellates could bring increased heterotrophy to the Arctic, expressed as the ratio of respiration to primary production (**Table 1**, **Figure 6**). The paradigm that flagellates will decrease sedimentation corresponds to a scenario of higher retention of organic carbon in surface waters and higher respiratory losses (Forest et al., 2010; Vaquer-Sunyer et al., 2010). Similarly, predictions of lower sedimentation are associated with an activation of the microbial loop (Kirchman et al., 2009b). High export in absence of diatoms in the Arctic is also possible as shown in this study, or when dinoflagellates replace diatoms (Rivkin et al., 1996); high sedimentation by flagellates has been observed in the field even when diatoms are abundant in surface waters (Amacher et al., 2013). A high export is possible if a link between the classical and microbial food webs develops through the consumption of microzooplankton and detritus by copepods.

#### CONCLUSIONS


#### REFERENCES


The quality of the results is based on previous knowledge of the food web and trophic interactions as well as the availability of rate processes at critical times through the growth season (April to September). Technological advances will help provide a better understanding of inter-annual and intra-annual variability in Arctic systems. As more data becomes available the quality of model predictions, as well as their accuracy, will increase.

#### AUTHOR CONTRIBUTIONS

MV organized the study, was responsible for searching the majority of data used in the model, participated in model interpretation and was in charge of the writing. TR was in charge of the model development, participated in interpretation of results and writing the manuscript. IP, EN, and KM provided data for the model and participated in food web construction, data interpretation and writing.

#### ACKNOWLEDGMENTS

This work was made possible through a fellowship from Hanse-Wissenchaftskolleg, Delmenhorst, Germany to MV and we thank them for their hospitality. We thank Natalie Niquil for sharing inverse modeling code and Marit Reigstad and Lena Seuthe for data on POC, chla and bacterial abundance and production. We also thank Marit Reigstad, Lena Seuthe, Camilla Svensen and Elisabeth Halverson for discussions on Arctic food webs, Anya Waite and George Jackson for discussions on phytoplankton sedimentation and Alexandra Kraberg for sharing her knowledge on C:Chla rations in the Arctic. The Polar Biological Oceanography Group of the Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research within Polar Regions and Coasts in the changing Earth System (PACES I and II) provided support for EN and IP in the Plankton Ecology and Biogeochemistry in a Changing Arctic Ocean (PEBCAO) group. A United States National Science Foundation grant PLR-1443705 and the Carbon Bridge project No. 226415, Polar Program under the Research Council of Norway, provided partial funding for MV.


thresholds for Arctic plankton community metabolism. Biogeosciences 10, 357–370. doi: 10.5194/bg-10-357-2013


organisms. Limnol. Oceanogr. 34, 1290–1299. doi: 10.4319/lo.1989.34. 7.1290


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Vernet, Richardson, Metfies, Nöthig and Peeken. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Protist Communities in Moored Long-Term Sediment Traps (Fram Strait, Arctic)–Preservation with Mercury Chloride Allows for PCR-Based Molecular Genetic Analyses

#### Edited by:

Marcelino T. Suzuki, Sorbonne Universities (UPMC) and Centre National de la Recherche Scientifique, France

#### Reviewed by:

David Walsh, Concordia University, Canada Susanne Neuer, Arizona State University, United States Anna Vader, University Centre in Svalbard, Norway

> \*Correspondence: Katja Metfies katja.metfies@awi.de

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Marine Science

Received: 30 August 2016 Accepted: 04 September 2017 Published: 21 September 2017

#### Citation:

Metfies K, Bauerfeind E, Wolf C, Sprong P, Frickenhaus S, Kaleschke L, Nicolaus A and Nöthig E-M (2017) Protist Communities in Moored Long-Term Sediment Traps (Fram Strait, Arctic)–Preservation with Mercury Chloride Allows for PCR-Based Molecular Genetic Analyses. Front. Mar. Sci. 4:301. doi: 10.3389/fmars.2017.00301 Katja Metfies 1, 2, 3 \*, Eduard Bauerfeind<sup>1</sup> , Christian Wolf <sup>1</sup> , Pim Sprong<sup>1</sup> , Stephan Frickenhaus <sup>1</sup> , Lars Kaleschke<sup>4</sup> , Anja Nicolaus <sup>1</sup> and Eva-Maria Nöthig<sup>1</sup>

<sup>1</sup> Alfred -Wegener- Institut Helmholtz-Zentrum für Polar-und Meeresforschung, Bremerhaven, Germany, <sup>2</sup> Jacobs University Bremen, Bremen, Germany, <sup>3</sup> Helmholtz Institute for Functional Marine Biodiversity, Oldenburg, Germany, <sup>4</sup> Institute of Oceanography, University of Hamburg, Hamburg, Germany

Here we present a pilot study demonstrating, that preservation with mercury chloride allows the application of PCR-based molecular methods for the characterization of marine protist communities collected with moored long-term sediment traps. They can provide information on pelagic protist communities by collecting sinking plankton from the upper water column all year-round, even in remote polar oceans. Assessment of small protist species from the nano- and picoplankton fractions in sedimented material by microscopy is extremely challenging or almost impossible. Hence, comprehensive studies of variability in protist community composition in moored long-term sediment traps are scarce. Considering that marine nano- and picoeukaryotes are ecologically very important, new approaches are urgently needed to investigate protists in the smallest size-fractions of moored long-term sediment trap samples. We applied the quick and cost-effective Terminal Restriction Length Polymorphism (T-RFLP) on a set of selected samples that were collected between 2000 and 2010 in September at a depth of ∼300 m in the area of the "LTER (Long-Term Ecological Research) site HAUSGARTEN" in the eastern Fram Strait (Arctic). The results of these analyses suggest a change in the trapped protist community after 2002 in this area. A comparison of 18S sequences obtained via 454-pyrosequencing from samples collected in the water column and mercury chloride preserved sediment traps in 2009 and 2010 suggests, that sediment traps might reflect the pelagic eukaryotic microbial biodiversity qualitatively. Furthermore, we have indication that preservation with mercury chloride does not severely change the nucleotide composition of 18S rRNA genes in long-term sediment traps. Overall, we suggest that preservation with mercury chloride is a key to open the door for molecular genetic analyses of long-term sediment trap samples, and that PCR-based molecular methods have a strong potential to become an important tool for comprehensive taxonomic analyses of protist- and bacterial communities in moored long-term sediment traps.

Keywords: mercury chloride, long-term sediment traps, 454-pyrosequencing, protist communities, terminal restriction length polymorphism, molecular analyses

#### INTRODUCTION

The deployment of moored sediment traps provides valuable long-term information on particle export and composition of sedimenting particles. It facilitates an understanding of plankton dynamics in the upper water column all year-round. It has to be kept in mind, however, that there are uncertainties related to the use of these tools, such as trapping efficiency that might compromise the validity of the results (Butman, 1986; Gust et al., 1994; Buesseler et al., 2007). Nonetheless, they are an appropriate tool for gaining insights in vertical particle flux patterns. At present they are also the only tool that allows the continuous collection of sinking particles for further microscopic chemical and biochemical analyses over larger temporal scales. Timeseries measurements represent an excellent approach to evaluate implications of environmental change on ecosystems, including the Arctic pelagic ecosystem (Glover et al., 2010; Wassmann et al., 2011). In 1999, the Alfred-Wegener-Institute Helmholtz Centre for Polar and Marine Research established the LTER (Long-Term Ecological Research) observatory HAUSGARTEN (Hausgarten) to carry out regular observations of the ecosystem in the eastern Fram Strait (Soltwedel et al., 2005, 2016). It is expected that global warming and the ensuing sea ice melt will strongly alter the Arctic pelagic environment. This could eventually result in modification of unicellular plankton species composition and biomass with changes in matter fluxes within the entire pelagic system. Thus, it is necessary to attain information about the temporal occurrences of phyto- and protozooplankton species and to understand their variability in relation to different environmental conditions. In the area of the "Hausgarten," investigations involve year-round observation of vertical particle flux by using sediment traps. Measurements of bulk parameters like sedimenting matter and its components combined with light microscopy can provide an estimate of the pelagic protist community and its fate in the catchment area above the traps (Bauerfeind et al., 2009). However, assessment of small protist species from the nano- and picoplankton fraction in this way is almost impossible owing to their small size and simple morphology. As a general consequence, information on changes in protist communities collected with sediment traps is mainly limited to shell-forming taxa. Considering the ecological relevance of the nano- and picoplankton fractions, new approaches are needed that provide comprehensive information on changes in protist communities collected with long-term sediment traps.

Molecular methods have become an indispensable tool box in marine biology over the past three decades. These techniques allow comprehensive evaluation of marine protists, including the nano- and picoplankton fraction, because they are independent of organism size or morphology (Ebenezer et al., 2012). Among many others, these approaches include molecular fingerprints, such as terminal restriction lengths polymorphism (T-RFLP). Molecular fingerprinting approaches are well-established molecular genetic tools for quick comparative analyses targeting the rRNA-coding genes of microbial communities (Dunbar et al., 2000; Danovaro et al., 2006; Joo et al., 2010). Ribosomal genes, being universally present in all cellular organisms are well-suited for molecular surveillance of marine biodiversity (Woese, 1987). Over recent years, a considerable number of marine surveys have taken advantage of ribosomal sequence information to broaden our understanding of protist diversity and community structure (e.g., Medlin et al., 2006; Metfies et al., 2010). In this context, the assessment of the diversity of protist communities via T-RFLP fingerprints is based on sequence heterogeneity e.g., within the 18S rRNA gene of different taxa. The 18S rRNA gene is amplified with a fluorescently-labeled primer and digested with restriction enzymes (Marsh et al., 2000). The composition of differently sized restriction fragments in a sample acts as a characteristic fingerprint of a microbial community that allows qualitative comparisons of community composition. However, it does not provide information on species composition, abundance, or diversity in a sample (Bent et al., 2007). Nevertheless we have chosen this method as a representative PCR-based molecular method for our pilot study. It is quick and cost-efficient in comparison to other molecular methods, such as next generation sequencing of ribosomal genes. Thus, in case of a failure of our pilot study, financial loss would have been relatively small. In contrast, Next Generation Sequencing (NGS) of ribosomal genes, e.g., via 454-pyrosequencing or Illumina sequencing is more costly, but it allows high resolution, taxon-specific assessments of protist communities, including their smallest size fractions and the rare biosphere (Kilias et al., 2014; de Vargas et al., 2015).

Overall, molecular analyses of moored sediment trap samples are potentially promising tools for comprehensive long-term information on change in sinking protistan communities, even for the smallest protists. Molecular methods have been used to study microbial community composition in unpreserved samples collected with sediment traps (e.g., Amacher et al., 2009). However, time-series studies based on the deployment of long-term (annual) sediment traps require treatment of the collected particles with well-established and efficient fixatives like mercury chloride to prevent microbial degradation (Knauer et al., 1984). To our knowledge, to date there is no publication that describes the application of molecular methods to assess microbial community composition in preserved sample material collected with long-term sediment traps.

In this pilot study, we assessed the applicability of 18S T-RFLP fingerprinting and 454-pyrosequencing for analysis of mercury chloride-preserved protist communities collected with moored long-term sediment traps deployed in the eastern Fram Strait (Arctic). We used T-RFLP and 454-pyrosequencing as two methods among various other PCR-based molecular approaches, such as quantitative PCR or molecular sensors. The manuscript is based on the assumption that binding of mercury ions to nucleic acids is a reversible process that does not alter the sequence of the DNA (Yamane and Davidson, 1960). A further goal of this study was to address if differences in the observed fingerprinting patterns might reflect variability in the environmental conditions observed in the catchment area of the sediment traps.

#### MATERIALS AND METHODS

# Sample Collection

#### Sediment Traps

Sedimenting particles including protist cells were sampled by modified automatic Kiel sediment traps, with a sampling area of ∼0.5 m<sup>2</sup> , and 20 liquid-tight collector cups (Zeitzschel et al., 1978; Kremling et al., 1996). Here, we present results from the shallowest (∼200–300 m below sea surface) sediment traps at the central station of the "Hausgarten" (79◦ N, 4◦ E; water depth 2,550 m) (**Figure 1**). The sampling time for the ten samples analyzed in this study varied between 8 and 20 days (one collector cup) during the first 2 weeks of September for ten years (**Table 1**). A gap of data in 2003 results from an electronic failure. The collector cups were filled with filtered sterile North Sea water. Salinity was adjusted with NaCl to 40 psu. The liquid in the collector cups (250 or 400 ml, depending on the sediment trap used) was spiked with mercury chloride (0.14% final concentration). After recovery of the moorings (∼10 month after the collection period) the samples were stored refrigerated until further processing in the laboratory. A first batch of samples (2000–2008) was processed in 2010, while a second batch (2008–2010) was processed in 2013. The sample collected in 2008 was analyzed as a replicate in both batches. Samples were split by a wet splitting procedure after removal of zooplankton (swimmers) >0.5 mm, which were manually removed under a dissecting microscope at a magnification of 20 and 50. Subsequent molecular analyses are based on 1/64 splits of the original sediment trap sample. We collected cells for isolation of DNA by filtration of a fraction of the original sample onto a 0.2µm Isopore GTTP membrane filter (Millipore, Schwalbach, Germany). Filters were washed with sterile North Sea water (∼50 ml). The sterile sea water was applied and pumped over the filter while it was still kept in the filtration unit. This washing step was included in the protocol to remove residual mercury chloride from the samples. PCR-amplification from sediment trap material failed if this step was not included in the protocol (personal communication Stephan Thiele). Because of limited sample material we were not able to carry out a series of optimization experiments of this step. Nonetheless, the sediment trap sample collected in 2008 was analyzed twice (including the filtration step) to assess the reproducibility of the method.

FIGURE 1 | Map of the "Hausgarten" of the Alfred Wegener Institute for Polar and Marine Research in the Eastern Fram Strait. Black dots in the map mark the location of the stations. The central station (HG4) of the "LTER (Long-Term Ecological Research) observatory HAUSGARTEN" is labeled in the map as "IV" (copyright Soltwedel, AWI).



"Deployment Period" describes the complete period during which the sediment trap was collecting and "Collection Period" describes the time period during which sedimenting particles were collected in one of the collection cups of the sediment trap.

Furthermore, we aimed to elucidate if prolonged storage would affect the results of T-RFLP.

#### Water Column

Samples were taken in July during research cruises ARKXXIV (2009) and ARKXXV (2010) of RV Polarstern to Fram Strait. Water samples were collected during the upcasts at the vertical maximum of Chl a fluorescence determined during the downcasts. The sampling depths varied between 10 and 50 m. 2 liter subsamples were taken in PVC bottles from the Niskins. Particulate organic matter for molecular analyses was collected by sequential filtration of one water sample through three different mesh sizes (10µm, 3µm, 0.4µm) on 45 mm diameter Isopore Membrane Filters at 200 mbar using a Millipore Sterifil filtration system (Millipore, USA).

#### Microscopic Counting of Phytoplankton

A split (1/8) of the original sediment trap sample was stored refrigerated in brown glass bottles in the fridge until microscopic counting. Aliquots of 3–10 ml were settled for 8–24 h, and a minimum of 50–100 cells of the dominant species or groups were counted with an inverted microscope at four different magnifications (100–400x) using phase contrast (Utermöhl, 1958).

#### DNA-Extraction

Genomic DNA was isolated from the samples with the E.Z.N.A Plant DNA Kit (Omega Bio-Tek Inc., Norcross, USA). We modified the original isolation protocol by insertion of an additional washing step using buffer "SPW wash buffer". Following DNA extraction 2.4µl of the genomic DNA [∼15 ng/µL] isolated from the sediment traps was amplified with the REPLI-g Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol.

#### T-RFLP in silico Analyses

Protist communities collected from the depth of the chlorophyll a maximum (chl a max) during the expedition ARK XXVII-2 of RV Polarstern in 2010 to the Fram Strait were analyzed by 454-pyrosequencing the 18S V4-region (Kilias et al., 2013). Restriction sites of HaeIII (GGCC) were mapped in silico on the abundant 18S sequence reads (>1%) of all observed reads in the sample) observed in this data set. Representative sequences for these taxa were downloaded from Genbank. The mapping was carried out using MapDraw as implemented in the DNAstar software package (Lasergene, USA).

#### T-RFLP in vitro Analyses PCR-Amplification

T-RFLP analysis is based on the amplification of ribosomal sequences via PCR (Treusch et al., 2009). A fragment of the small subunit ribosomal RNA gene was amplified with a universal primer set targeting marine protists. The amplification was based on the universal primer set 690F-FAM 5 ′ -TCAGAGGTGAAATTCTTGGAT-3′ (Metfies and Medlin, 2008) and 1055R 5 ′ -CGGCCATGCACCACCACCCAT-3′ (Metfies et al., 2007). We used this primer set because it amplifies parts of the 18S V4-region, which is the most variable region of this gene (Nickrent and Sargent, 1991). This primer set was chosen over other primer sets because other universal primers e.g., 528F 5′ -GCGGTAATTCCAGCTCCAA-3′ (Medlin et al., 2006) that include the V4 region in the PCR-product have no mismatches against key metazoans that had to be excluded from the analyses. The forward primer (690F-FAM) has two mismatches to metazoan species like Calanus finmarchicus or Themisto libellula, which are frequently found in sediment traps deployed in Fram Strait. The 690F-FAM primer was labeled with phosphorarmidite fluorochrome 5-carboxy-fluorescein (FAM). The PCR cocktail contained 1µL of a 1:20 dilution of the amplified genomic DNA as template, 1x amplification buffer (5 Prime, Hamburg, Germany), 0.2µM dNTPs (5 Prime, Hamburg, Germany), 0.2µM of each primer, and 1 U of Taq DNA polymerase (5 Prime, Hamburg, Germany). The amplification protocol was 5 min at 94◦C, 1 min at 94◦C, 2 min at 54◦C, and 2 min at 72◦C for 35 cycles and an extension for 10 min at 72◦C. For some samples (2009 and 2010) the amplification efficiency was improved by addition of 0.016% BSA (Sigma, Hamburg, Germany) and 2.5% Polyethylenglycol (Sigma, Hamburg, Germany) to the PCR cocktail.

#### Restriction Analyses

Subsequently ∼500 ng of the resulting 18S fragment was subjected to a restriction analysis with the frequent cutter HaeIII (New England Biolabs, Ipswich, USA). We have chosen this restriction enzyme because it is a frequent cutter that cuts the 18S PCR-fragment in the motif GGCC downstream of a variable region that contains a number of insertion and deletions. The addition of another frequent cutter, such as Sau3aI (GATC) would have decreased the resolution of the analysis (**Supplementary Figure 1**). The restriction reaction was carried out in a reaction volume of 15µL (500 ng DNA; 1.5µg/µL BSA; 1x restriction buffer; 2 U HaeIII) for 1 h at 37◦C. The restriction enzyme was inactivated by 20 min incubation of at 80◦C. The sizes of the 18S terminal restriction fragments were determined by analysis with a capillary sequencer (ABI 3130XL, Applied Biosystems). The T-RFLP analysis was carried out in three technical replicates for each sample.

#### Statistical Analyses

In an 18S T-RFLP analysis the community is characterized by its composition (presence/absence) of differently sized DNA fragments, which are a result of a sequence specific restriction of the amplified 18S fragment with the restriction fragment HaeIII. Presence/absence matrices, reflecting the community profiles in the samples, were generated by binning the original data obtained after size separation with the capillary sequencer using the "Interactive Binner" (Ramette, 2009) (http://www.mpi-bremen.de/en/Software\_2.html). Differences in the T-RFLP community profiles were estimated by calculating the Jaccard index. The Jaccard index is a statistical method used for comparing the similarity and diversity of sample sets. It determines the similarity between samples. The result of the analysis is a distance matrix of the samples in the data set. We visualized the resulting distances using multidimensional scaling (MDS) implemented in the vegan- software package (http://r-forge.r-project.org/projects/vegan/). Groups in the MDS plot were determined a priori. The significance of the grouping was tested by analyzing the similarity between groups with an ANOSIM analysis (Clarke, 1993). ANOSIM is a multivariate, non-parametric statistical method used for comparing community compositions among groups of samples. The correlation of environmental conditions and T-RFLP patterns was tested with a Mantel-test, comparing distance matrices of the environmental conditions and the T-RFLP profiles also using the vegan-software package. All statistical analyses were carried out within the R-package (R Development Core Team, 2011). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria, URL http://www.R-project.org/).

#### 454-Pyrosequencing and Sequence Analyses

For 454-Pyrosequencing, a ∼670 bp fragment of the 18S rRNA gene containing the hypervariable V4 region was amplified separately from each filter fraction with the primer set 528F (GCG GTA ATT CCA GCT CCA A) and 1055R (ACG GCC ATG CAC CAC CAC CCA T). Details about the procedure, the PCR reaction mixture and the reaction conditions were described previously (Wolf et al., 2013). Pyrosequencing was performed on a Genome Sequencer FLX system (Roche, Germany) by GATC Biotech AG (Germany).

#### Data Analysis of Water Column

Raw sequence reads were processed using the analysis pipeline Quantitative insights into Microbial Ecology Version 1.8.0 (QIIME) (Caporaso et al., 2010). Reads with a length under 250 bp were excluded from further analysis to ensure including the complete V4 region in the analysis and to get rid of short reads. The quality score was set to 25 and eight homopolymeres and two primer mismatches were allowed. Chimeric sequences in the remaining data set were eliminated from further analyses based on an assessment using the software UCHIME (Edgar et al., 2011) within QIIME. The resulting high quality reads were grouped into operational taxonomic units (OTUs) at the 97% similarity level using Uclust (Edgar, 2010). OTUs comprised of less than 4 sequence reads were removed from the analysis. The remaining sequences were aligned using the SILVA reference database (SSU Ref 119). The raw sequences were deposited at the Sequence Read Archive of the European Nucleotide Archive (ENA) under Accession PRJEB21238.

#### Data Analysis of Sediment Traps

Raw sequence reads were processed as described previously (Wolf et al., 2014) to obtain high quality reads. Briefly, reads shorter than 300 bp and longer than 670 bp, reads with more than one uncertain base (N), chimeric reads, and reads belonging to metazoans were removed. The remaining high quality reads of all samples were clustered (furthest neighbor algorithm) into operational taxonomic units (OTUs) at the 97% similarity level using the software Lasergene 10 (DNASTAR, USA). OTUs comprised of only one sequence (singletons) were removed. Consensus sequences were generated for each OTU and their taxonomical affiliation was determined by placing them into a reference tree, containing about 1,250 high quality sequences of Eukarya from the SILVA reference database (SSU Ref 111), using the PhyloAssigner pipeline (Vergin et al., 2013). The compiled reference database is available on request in ARB-format. The raw sequences were deposited at the Sequence Read Archive of the European Nucleotide Archive (ENA) under Accession ERP024120.

#### Phylogenetic Sequence Comparison

Abundant OTUs obtained from water column samples were compared with OTUs from sediment traps via phylogenetic sequence comparison. Sequence alignment construction and phylogenetic tree calculation was done with the open software package MEGA (Kumar et al., 2016).

#### Environmental Parameters

Data on aerial sea ice extent were obtained from daily special sensor microwave imaging data. The number of days of ice coverage at the central station of the "Hausgarten" during the summer months (May–September) was calculated from these data (Bauerfeind et al., 2009). Average sea surface temperatures for the same period in the area of the "Hausgarten" were taken from the "Averaged Time Series, MODIS-Terra.R1.1 data set" (https://giovanni.gsfc.nasa.gov/giovanni/).

# RESULTS

#### T-RFLP Analyses

Prior to the in vitro analysis, we carried out an in silico restriction of ribosomal sequences originating from taxa that dominated 454-pyrosequencing libraries obtained from pelagic samples collected in the study area during a summer cruise (Kilias et al., 2013). The restriction fragments obtained in silico from the reference sequences were 383 bp for Gyrodinium sp. (JQ692033), 379 bp for Phaeocystis pouchetii (AJ278036), and 382 bp for the Arctic strain of Micromonassp. (AY954999) and the 18S sequence of Bathycoccus sp. (JF794058).

The DNA-concentration obtained from the sediment trap samples was significantly correlated with concentrations of particulate organic carbon in the traps (R <sup>2</sup> = 0.76). Subsequent to successful amplification of a ∼400 bp PCR product (**Figure 2**), the T-RFLP analysis resulted in a total of 61 different 18S T-RFLP fragments in vitro. The average number of fragments per sample was rather low at 17 fragments. Eleven fragments were found mainly in the earlier phase of the observation period (2000– 2002), while six fragments were mainly found after that period (**Supplementary Table 1**). All samples contained at least one T-RFLP fragment in the size range of 379–383 bp deduced in silico for the dominant pelagic taxa collected in Fram Strait in July 2010 (Kilias et al., 2013). Moreover, all samples contained a restriction fragment of 382 bp (Arctic Micromonas sp. CCMP 2099) and one of 383 bp (Gyrodinium sp.). The distances between the different T-RFLP profiles were visualized in a multidimensional scaling plot (metaMDS plot) (**Figure 3**). The plot suggests that the different community profiles segregate into four groups. The community profiles within a group are more similar to each other than to the community profiles of the other groups. An ANOSIM analysis to test the significance of the grouping resulted in an R-value of 0.9256 and p-value of 0.001, indicating highly significant differences between the groups. Group one is composed of samples collected in 2000–2002, group two of samples collected in 2004–2005, group three consists of samples collected in 2006–2007, and group four of samples collected in 2008 to 2010. The metaMDS-plot suggests that the protists communities in 2006–2010 are more similar to each other than to those of the years before. This observation might point toward a change in the protist community structure after 2002.

# Correlation of Environmental Conditions with Molecular Fingerprints

Summer ice coverage at the sediment trap locations between 2000 and 2010 varied from a minimum total absence of ice (0 days) in 2006 to a maximum 113 days of ice coverage in 2008. Minimum and maximum average surface temperatures in the area of the central station "HGIV" (**Figure 1**) ranged from 1.33◦C in 2008 to 3.37◦C in 2006 (**Table 2**). Overall, we observed that the mean surface temperature during summer (May–September) was warmer during those years with less days of ice-coverage than in those years with longer periods of ice-coverage. We used the mean surface temperature and ice-coverage data from May to September for the statistical analyses, to ensure that the data cover the whole Arctic growth period. Based on a Mantel-test we did not discover a statistically significant correlation between the difference patterns of water temperature or ice-coverage and the difference patterns from the protist community fingerprinting profiles (data not shown).

#### Phytoplankton Counts

Sedimentation patterns of phytoplankton cells encompassed with the light microscope reveal that during the first 3 years (2000– 2002) high cell numbers of diatoms were found among the phytoplankton cells in the samples. In the following years the abundance of diatoms significantly decreased and they were present in only fairly low numbers in the sediment traps (**Figure 4**).

FIGURE 2 | Picture of an agarose gel (1.5% [w/v]) to visualize the outcome of the amplification of a 18S DNA fragment via PCR. The size of the PCR-product was determined in comparison to a 1 kb size marker (M).

FIGURE 3 | Multidimensional scaling plot (MDS-plot) to illustrate differences between the 18S T-RFLP patterns of the different protist communities in the sediment traps. The MDS-plot is based on the calculation of Jaccard indices for the data set derived from amplification of a 18S PCR fragment and subsequent digestion with the restriction enzyme HaeIII.

TABLE 2 | Days of ice coverage over the sediment trap at the "Central Hausgarten Station" during the summer of the collection period and the average surface temperature in the "Hausgarten" during summer in the year of the collection period.


#### Comparison of Sequence Assemblages Obtained from Unpreserved Pelagic Samples and Mercury Chloride Preserved Sediment Trap Samples

It would only make sense to use archived sediment trap samples to elucidate changes and variability in pelagic eukaryotic microbial biodiversity above the sediment traps if it is known to which degree the biodiversity found in the sediment traps is representative for pelagic biodiversity. We compared the abundant biosphere (>1% of total sequences) in 18S sequence libraries obtained from samples collected directly from the deep chlorophyll a maximum (DCM) with the abundant

biosphere in 18S libraries generated from the mercury chloride preserved sediment trap samples. In this study the abundant biosphere represented for both sample sets 60–90% of all sequences identified in the individual samples. Sampling in the water column took place at four different stations within the observation area (HG4; S3; N4, and HG1) during summer of 2009 and 2010, and sediment traps were deployed at HG4 (**Figure 1**). For these years, the number of abundant operational taxonomic units (OTUs) in the two different sample types (water column, trap) was in a similar range. In the DCM samples we observed 45 abundant OTUs, while we detected 39 abundant OTUs in the sediment trap samples. At higher taxonomic level the OTUs of the DCM samples could be assigned to Mamiellophyceae, Phaeocystis sp., Arthropoda, other metazoan, Ciliophora, Dinophyceae, Syndiniales, MAST (Marine Stramenopiles), Bolidomonas sp. and diatoms (**Table 3**). The OTUs found in the sediment traps could be assigned to Mamiellophyceae, Chrysochromulina sp., Ichtyosporea, Metazoa, Fungi, Syndiniales and Rhizaria (**Table 3**). 18S sequences of Ciliophora were exclusively found in the water column sequence assemblages, while Ichtyosporea sequences were only found in the sediment trap sample of 2010. All other taxonomic groups were found in both, pelagic samples and sediment trap samples. Moreover, at the level of OTUs more than half of the abundant OTUs observed in the water column samples were also abundant in the sediment trap samples (24/39). This included sequences affiliating with Phaeocystis sp., Micromonas sp., diatoms, alveolates and copepods (**Figure 5**). Mamiellophyceae (mainly Micromonas sp.) and Syndiniales (alveolates) were significantly abundant in both sample sets. There was even agreement between the sample sets that the relative abundance of Micromonas sp. was higher in 2010 than in 2009. In contrast, OTUs assigned to Phaeocystis sp. were major contributors to the abundant biosphere of the DCM, while they were only found in the rare biosphere of the sediment trap samples. The data indicate the same trend for Dinophyceae that dominated the eukaryotic microbial sequence assemblage of the DCM, while they were found in the rare biosphere of the sediment traps. Only a third of the OTUs assigned to Dinophyceae observed in the abundant biosphere of the sediment traps were also detected in the abundant biosphere of the water column. Rhizaria and fungi dominated the eukaryotic microbial sequence assemblage of the



The relative sequence abundance depicted for the DCM are mean values calculated from the relative abundances observed for a respective taxon in at four different stations (S3; HG1; HG4; N4) in the area of the "Hausgarten".

sediment traps, while they were only found in the rare biosphere of the samples collected in the DCM above the sediment traps. Overall, most of the taxa found in our study were present in both kinds of samples, but their relative contribution to the eukaryotic microbial community differed between the DCM samples and the sediment trap samples.

The impact of preservation with mercury chloride on the nucleotide composition within the 18S rRNA gene of individual taxa was evaluated by comparing sequences of taxa that were found in the abundant biosphere of both, sediment trap and water column samples. In addition, one representative sequence for each taxon was downloaded from GenBank and included in the analyses. Overall, sequences retrieved from the sediment traps were highly similar to sequences obtained from the water column or GenBank. Any sequence that was observed in the abundant biosphere of the mercury chloride preserved sediment trap sample had max. 0.036% difference to either a sequence of this taxon observed in the water column or to a GenBank sequence (**Figure 5**). Most dissimilarity was in the range of 0.01%. Overall, the range of dissimilarity between the sequences of the abundant biospheres observed in unpreserved DCM samples and mercury chloride fixed sediment trap samples was in a range between <0.01 and 0.036%.

#### DISCUSSION

#### Mercury Chloride Preservation

In this pilot study, we demonstrate for the first time the applicability of molecular methods on mercury chloridepreserved cells obtained by means of moored sediment traps. The use of long-term sediment traps means analyzing preserved organic material, which is often a challenging task. Prior to the molecular analyses the genomic DNA isolated from the preserved material was randomly amplified using a DNA-amplification kit in order to ensure that we would have sufficient amounts of genomic DNA for PCR-optimization and potential future molecular analyses. This method has proven to be applicable for reproducible random amplification of genomic DNA (Han et al., 2012), while highly GC-rich regions might be amplified with slightly less efficiency. The region amplified in this study is not particularly GC-rich. Therefore, the random amplification of this genomic region should not be affected by these problems. Other publications have noted difficulties with amplification of nucleic acids from tissues preserved with fixatives that contain mercury chloride stating that mercury chloride could be deposited and remain bound in tissues. This could inhibit Taq DNA polymerase activity during PCR (Oleary et al., 1994). Our protocols for the preparation of sample material for molecular analyses and nucleic acid isolation each involved an additional step of washing. For most of the samples this vigorous washing of the sample material and the isolated nucleic acids appeared to be sufficient to remove residual mercury chloride that could inhibit the PCRamplification. However, for some samples (2009–2010) it was necessary to add the PCR enhancer bovine serum albumin (BSA) and polyethylenglycol (PEG) for successful amplification of an 18S PCR product. BSA is an agent used to increase DNA polymerase stability. BSA has also been shown to overcome the inhibitory effects on RT-PCR (Giambernardi et al., 1998). PEG has been shown to enhance PCR reactions in the presence of high concentrations of polysaccharides, which inhibit DNA polymerase efficiency (Pandey et al., 1996). These observations suggest that our DNA-extraction method does not always completely clean the DNA from potential PCR inhibitors present in the sediment trap material such as polysaccharides. However, it is not uncommon that PCR-amplification from sediment samples is hampered by inhibitors present in environmental samples e.g., from river sediments (Arbeli and Fuentes, 2007). Thus, individual optimizations of amplification protocols for single samples might be necessary for future PCR-based molecular analyses of long-term sediment trap samples.

#### PCR-Based Methods

As examples of PCR-based methods, T-RFLP analysis and metabarcoding are prone to biases related to the PCR-amplification of ribosomal genes. Ideally, the relative contribution of taxa in a protist community should remain unchanged by PCRamplification. However, there are indications that PCR in mixed communities leads to alterations in the original DNA-ratios and target gene compositions (Amacher et al., 2011) and the addition of the PCR-enhancer PEG might impact molecular fingerprinting patterns (Pandey et al., 1996). Amplification efficiency of the target gene in a sample is impacted by the fact that some taxa are amplified by a certain primer set with greater efficiency than others (Suzuki and Giovannoni, 1996; Polz and Cavanaugh, 1998). We used T-RFLP analysis and metabarcoding to carry out qualitative comparisons of community compositions in sediment traps and the DCM. Thus, for this study it is probably of minor relevance that the 18S of some

species is amplified with more efficiency than that from others. In any case, the PCR amplification for T-RFLP analysis in our study suggests reproducibility of the methodology. The eukaryotic microbial communities collected in 2008 were analyzed twice, including repetition of the filtration step and the amplification of genomic DNA. Amplification of the 18S from the samples of 2009 and 2010 was only possible in the presence of BSA and PEG. Therefore, the repetition of the analysis of the 2008 sample was also done in the presence of BSA and PEG. In the metaMDS-plot, the replicates are located in the same position. This result is of further interest, as the T-RFLP analyses were carried out in two batches. The first batch (2000–2008) was analyzed in 2010, while the second batch (2008–2010) was analyzed in 2013. This approach allowed us to elucidate if longer storage of the samples alters the T-RFLP results. The longest stored samples from 2000 to 2002 are most different from samples collected in later years. These data are supported by microscopic analyses that suggest a change in the community structure after 2002, reflected by a significant decline of diatoms (**Figure 4**). Therefore, it is likely that the differences in the T-RFLP profiles could be related to changes in the community structure, rather than to effects of loss of cells by filtration, washing steps, the addition of PCR-enhancers or long-term storage on the DNA composition. The superposition of the replicate samples in the metaMDS-plot suggests that neither the addition of PCR-enhancers, nor prolonged storage over three years alter the T-RFLP patterns obtained from the sediment trap samples. Certainly, based on the current data set and the limited amount of replicated samples, we cannot completely exclude the possibility that long-term storage impacts nucleic acid composition in the sediment trap samples. Therefore, we suggest repeating the analysis on a 5 years basis in order to monitor the impact of long-term storage on biodiversity patterns obtained from mercury chloride preserved protist communities collected in long-term sediment traps. Finally, the results also suggest that the amplification of genomic DNA prior to the analyses does not change the protist community structure in the sample.

# T-RFLP Fragments and Species Compostion

Prior to T-RFLP analysis, expected fragment lengths were determined by in silico restriction of publically-available 18S sequences from taxa that dominated 454-pyrosequencing libraries obtained from pelagic samples collected in the eastern Fram Strait in summer 2010 (Kilias et al., 2013). These analyses were carried out to get a first impression of the T-RFLP fragment size spectrum that might be expected from in vitro analyses. We assumed that T-RFLP fragments in the size range determined in silico for abundant sequences (>1% of all sequence reads in a sequence library) observed in the water column should contribute to the protist communities in the sediment trap. The results of the in silico restriction were in agreement with the T-RFLP and meta-barcoding data. This could be an indication that the T-RFLP patterns obtained from the sediment traps might indeed reflect the protist composition in the traps. However, we cannot completely exclude the possibility that the presence of the restriction fragments could be based on the presence of a closelyrelated taxon with the same 18S restriction pattern. A T-RFLP fragment could represent the presence of a single species or of a variety of taxa (**Supplementary Figure 1**). Thus, T-RFLP is of very limited benefit for species identification and will probably be replaced by NG-sequencing in the future.

# Natural Pelagic Community and Mercury Chloride Preserved Trap Samples with Ecological Implications

In this study we used 454-pyrosequencing to evaluate to what extent the eukaryotic microbial biodiversity in the upper water column is reflected by the biodiversity of this group of organisms observed in mercury chloride preserved sediment trap samples. Furthermore, we addressed if preservation with mercury chloride severely impacts nucleotide sequence composition of the 18S rRNA gene of different taxa in the sediment traps. 454 pyrosequencing provides high resolution taxon-specific information on variability in sequence composition in the samples (Wolf et al., 2013). The untreated samples from the DCM were collected in 2009 and 2010 at four different stations of "Hausgarten" in order to acknowledge that the catchment area of the sediment-traps is large and very variable because of the highly dynamic oceanographic environment in Fram Strait. The integration of four different sampling sites was supposed to give a representative insight into the protist community composition in the DCM in the catchment area of the traps. The taxon composition of those sequences found in the abundant biosphere of the DCM reflects very well what was observed previously in large-scale biodiversity studies focusing on Arctic pelagic protists (Kilias et al., 2014; Metfies et al., 2016). In accordance with these studies, sequences affiliating with Micromonas sp., Dinophyceae and Phaeocystis sp. contributed significantly to the abundant biosphere of the 18S libraries generated from the DCM. Surprisingly the picoeukaryote Micromonas sp. was also significantly contributing to the abundant biosphere of the sediment trap samples, while Dinophyceae (Gymnodiniphycidae) and Phaeocystis sp. were only present in the rare biosphere of the sediment trap samples. Thus, the sequencing data are in accordance with the T-RFLP data of this study that suggest the presence of Micromonas sp. and Gyrodinium sp. (Gymnodiniphycidae) in the sediment trap samples. Overall, most taxa, and even picoeukaryote taxa were present in both sequence sample types. This finding suggests that most species of the photic zone might be exported to deeper water layers, while the mechanisms of transport remain to be elucidated. However, there are most probably differences in the export efficiency of different taxa due to a variety of factors and processes impacting the export efficiency of each taxon differently (Amacher et al., 2009). Thus, molecular based information on eukaryotic microbial biodiversity obtained from mercury chloride preserved sediment trap samples might provide valuable qualitative information on pelagic eukaryotic biodiversity in the catchment area, but it is most probably from limited value in respect to the relative contributions of taxa to the eukaryotic microbial communities above the traps. In this study we observed significant differences in the relative sequence contribution of some taxa. Our study suggests limited export of Phaeocystis sp., which is in agreement with a previous study that also reports limited sinking of Phaeocystis sp. in the Southern Ocean (Wolf et al., 2016). In contrast, sequences assigned to fungi were mainly found in the sediment traps. Marine fungi can dominate biomass on marine snow in the bathypelagic realm (Bochdansky et al., 2017). Thus, the marine fungi sequences observed in the sediment traps might originate from marine snow floating in the water column between the DCM and the sediment trap.

In this study the 18S rRNA gene sequences obtained from the mercury chloride preserved sediment trap samples were highly homologs to sequences obtained from the DCM or retrieved from a public database. The range of intra specific sequence variability observed in this study was in the range of 0.01– 0.036 %, which is reasonable considering natural intra-specific variability and sequencing errors of NGS approaches (Behnke et al., 2011; Miranda et al., 2012). Overall, our data suggest that preservation with mercury chloride does not significantly impact the nucleotide composition of 18S sequences or qualitative information on community composition determined with molecular methods. Thus, molecular data obtained from mercury chloride preserved sediment trap material might be suited to provide meaningful information on quantitative changes and variability in pelagic and exported eukaryotic microbes, including organisms of the smallest size fraction such as Micromonas sp.

# Time-Series Aspects

The value of the new approach was confirmed by testing the approach in the framework of LTER "Hausgarten." In our data set the T-RFLP profiles point toward a shift or at least strong variability in protist community composition in the sediment trap samples during the observation period 2000–2010. The data suggest that in September the trapped protist communities in the latter part of the study are significantly different than those trapped at the beginning of the observation period in 2000–2002. Based on the choice of samples we cannot exclude that differences in bloom timing, plankton phenology or local patchiness might be the reason for the observed differences in the T-RFLP patterns in September during the observation period. A more comprehensive ecologically focused NGS-study that includes additional parameter, and all samples collected all year round at different locations and depths in the area of "Hausgarten" has to elucidate this uncertainty. Nonetheless, our findings are supported by microscopic analyses that suggest a change in the community structure after 2003, reflected by a significant decline of diatoms (**Figure 4**), and all year round studies of other parameters for example zooplankton composition and fecal pellets similarly imply a shift during the observation period (Kraft et al., 2013; Lalande et al., 2013). Considering these results, and assuming trophic interactions, it seems that T-RFLP analysis is well suited to identify long-term variability and shifts in protist communities collected with moored sediment traps. Admittedly, it does not provide information on the extent or nature of the variability, but this was not the scope of this study. Statistical analyses suggest that the observed variability cannot be significantly explained by differences in summer sea surface temperature and ice-coverage over the trap at the "LTER observatory HAUSGARTEN." Long-term measurements across Fram Strait showed an anomalous incursion of warm Atlantic water into Fram Strait during 2005–2007 (Beszczynska-Möller et al., 2012). At the same time, summer sea ice-coverage in the area is determined by sea ice export from the central Arctic Ocean (Smedsrud et al., 2011). The failure to correlate the variability of T-RFLP patterns obtained from the sediment trap samples with differences in sea surface temperature and ice-coverage in eastern Fram Strait may not be too surprising given their complex causes, and given that they are independent processes.

#### Comments and Recommendations

The successful application of T-RFLP analysis and 454 pyrosequencing on our long-term sediment trap samples suggests that mercury chloride is appropriate to preserve longterm sediment trap samples for molecular genetic investigations that have the potential to provide meaningful information on pelagic eukaryotic biodiversity above the sediment traps. We also introduce T-RFLP as a cost-efficient, informative and reliable tool for quick qualitative assessments of protist communities. Finally, we suggest that other PCR-based methods, like quantitative PCR (Toebe et al., 2013) and molecular sensors (Metfies and Medlin, 2007; Diercks et al., 2008) could also be applied to analyze the composition of microbes (eukaryotes and prokaryotes) in moored long-term sediment traps. The application of next generation sequencing techniques, as demonstrated in this publication has the potential to provide qualitative taxon specific information on eukaryotic microbial biodiversity (Stoeck et al., 2010; Wolf et al., 2013). In contrast, quantitative PCR and molecular sensors could provide information on the presence and abundance of selected taxa in the sediment trap samples (Metfies et al., 2007). Currently, for various reasons these techniques are not applied routinely in marine ecology and long-term observation programs. But we expect, that technical progress make it feasible that some of these techniques might be applied on a routinely basis for the analysis of marine protist communities in the future. Overall, we expect that molecular methods become an important tool for comprehensive taxonomic analyses of protistand bacterial communities in moored long-term sediment trap.

#### AUTHOR CONTRIBUTIONS

KM and EN had the idea for the project and designed the experiments. AN carried out the practical work related to

#### REFERENCES


the analyses. KM analyzed the data and carried out statistical tests in collaboration with SF. EB was responsible for the sampling of protists via long-term moored sediment traps and contributed significantly to the evaluation of the data in the context of environmental conditions. LK contributed data on the ice-coverage for the observation periods. EN contributed microscopic counts of diatoms to the manuscript. CW contributed the 454-pyrosequencing data for the sediment traps. PS contributed the 454-pyrosequencing data for the water column. KM wrote the manuscript and all other authors contributed by critically reading and optimizing the manuscript.

#### ACKNOWLEDGMENTS

This work was supported by institutional funds of the Alfred -Wegener- Institut Helmholtz-Zentrum für Polar-und Meeresforschung, Germany and funds of the Helmholtz Association for financing the Helmholtz-University Young Investigators Group Planktosens (Grant VH-NG-500) and within the framework of project Frontiers in Arctic Marine Monitoring (FRAM). We greatly thank the crew of R.V. Polarstern for excellent support during the work at sea. Furthermore we thank Annika Schröer and Christiane Lorenzen for great technical assistance. Annegret Müller and Uwe John are acknowledged for excellent technical support of the fragment analysis.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2017.00301/full#supplementary-material

Supplementary Table 1 | Presence/Absence matrix of T-RFLP fragments. Fragments that appear mainly 2000–2002 are labeled in yellow, while fragments that appear mainly after 2006 are labeled in green.

Supplementary Figure 1 | Alignment of a selected set of protist taxa. Restriction sites of HaeIII, Sau3aI, and the binding sites of the primer used in this study are indicated.

pyrosequencing of hypervariable SSU rRNA gene regions. Environ. Microbiol. 13, 340–349. doi: 10.1111/j.1462-2920.2010.02332.x


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Metfies, Bauerfeind, Wolf, Sprong, Frickenhaus, Kaleschke, Nicolaus and Nöthig. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A Decadal (2002–2014) Analysis for Dynamics of Heterotrophic Bacteria in an Antarctic Coastal Ecosystem: Variability and Physical and Biogeochemical Forcings

Hyewon Kim1, 2 \* and Hugh W. Ducklow1, 2

<sup>1</sup> Department of Earth and Environmental Sciences, Columbia University, New York, NY, USA, <sup>2</sup> Division of Biology and Paleo Environment, Lamont-Doherty Earth Observatory, Columbia University, New York, NY, USA

#### Edited by:

Ingrid Obernosterer, Observatoire Océanologique de Banyuls-sur-Mer (CNRS), France

#### Reviewed by:

Michael R. Twiss, Clarkson University, USA Meinhard Simon, University of Oldenburg, Germany

> \*Correspondence: Hyewon Kim hyewon@ldeo.columbia.edu

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Marine Science

Received: 01 August 2016 Accepted: 14 October 2016 Published: 04 November 2016

#### Citation:

Kim H and Ducklow HW (2016) A Decadal (2002–2014) Analysis for Dynamics of Heterotrophic Bacteria in an Antarctic Coastal Ecosystem: Variability and Physical and Biogeochemical Forcings. Front. Mar. Sci. 3:214. doi: 10.3389/fmars.2016.00214 We investigated the dynamics of heterotrophic bacteria in the coastal western Antarctic Peninsula (WAP), using decadal (2002–2014) time series of two bacterial variables, bacterial production (BP) via <sup>3</sup>H-leucine incorporation rates and bacterial biomass (BB) via bacterial abundance, collected at Palmer Antarctica Long Term Ecological Research (LTER) Station B (64.8◦S, 64.1◦W) over a full austral growing season (October–March). Strong seasonal and interannual variability in the degree of bacterial coupling with phytoplankton processes were observed with varying lags. On average, BP was only 4% of primary production (PP), consistent with low BP:PP ratios observed in polar waters. BP was more strongly correlated with chlorophyll (Chl), than with PP, implying that bacteria feed on dissolved organic carbon (DOC) produced from a variety of trophic levels (e.g., zooplankton sloppy feeding and excretion) as well as directly on phytoplanktonderived DOC. The degree of bottom-up control on bacterial abundance was moderate and relatively consistent across entire growing seasons, suggesting that bacteria in the coastal WAP are under consistent DOC limitation. Temperature also influenced BP rates, though its effect was weaker than DOC. We established generalized linear models (GLMs) for monthly composites of BP and BB via stepwise regression to explore a set of physical and biogeochemical predictors. Physically, high BP and large BB were shaped by a stratified water-column, similar to forcing mechanisms favoring phytoplankton blooms, but high sea surface temperature (SST) also significantly promoted bacterial processes. High BP and large BB were influenced by high PP and bulk DOC concentrations. Based on these findings, we suggest an increasingly important role of marine heterotrophic bacteria in the coastal WAP food-web as climate change introduces a more favorable environmental setting for promoting BP, with increased DOC from retreating glaciers, a more stabilized upper water-column from ice-melt, and a baseline shift of water temperature due to more frequent delivery of warming Upper Circumpolar Deep Water (UCDW) onto the WAP shelf.

Keywords: heterotrophic bacteria, bacterial production, bacterial biomass, phytoplankton, the western Antarctic Peninsula, Palmer LTER, climate change

# INTRODUCTION

Heterotrophic bacteria utilize and remineralize dissolved organic carbon (DOC) and organic nutrients, mobilize carbon for upper trophic levels via microzooplankton grazing, and ultimately affect carbon fluxes and cycling in the ocean (Azam et al., 1983; Ducklow, 1983). Antarctic coastal waters are characterized by low bulk DOC concentrations compared to other coastal ecosystems due to the lack of terrestrial and allochthonous inputs of DOC. Therefore, bacteria must ultimately rely on in situ DOC produced by phytoplankton (Baines and Pace, 1991; Nagata, 2000) or from other trophic levels and processes such as zooplankton grazing, sloppy feeding, excretion, and cell lysis (Strom et al., 1997). Bacterial coupling with phytoplankton has been demonstrated in the coastal WAP (Ducklow et al., 2012; Saba et al., 2014), consistent with other observations from Antarctic waters (Morán et al., 2001; Morán and Estrada, 2002). Saba et al. (2014) demonstrated that bacterial blooms occurred in years of positive chlorophyll (Chl) anomalies. Ducklow et al. (2012) showed that the degree of coupling between phytoplankton and bacterial properties was variable depending on the space and time scales analyzed.

The coastal WAP is a very productive ecosystem with high phytoplankton productivity and biomass accumulations (i.e., high Chl) as spring ice-melt drives stratification of the upper water column and subsequently alleviates light limitation to cells (Smith et al., 2008; Vernet et al., 2008). The WAP reflects strong seasonal and interannual variability in a variety of marine ecological and biogeochemical functions as a result of large-scale climate variability including the Southern Annular Mode (SAM) and El Niño-Southern Oscillation (ENSO) and their impacts on local physical drivers such as sea ice, upper ocean physics, and local wind patterns. The teleconnection of ENSO with the WAP has similar effects as the SAM, where El Niño causes fewer storms and colder air temperatures, inducing winter sea ice growth. The negative SAM also creates favorable conditions for winter sea ice growth as a result of increased cold southerly winds blowing across the WAP (Stammerjohn et al., 2008). It was demonstrated that positive Chl anomalies occurred every 4–6 years in a stabilized water column induced by large sea ice extent during preceding winters as a consequence of the negative phase of winter SAM and by reduced wind speeds in the following spring as a result of positive spring SAM (Saba et al., 2014). Biological drawdown of dissolved inorganic nutrients in nearshore waters at Palmer Station is also regulated by a similar suite of climate and physical forcing factors (Kim et al., 2016). However, impacts of these physical forcing factors on other components of food webs, such as marine heterotrophic bacteria or microbial loop dynamics, have not been examined. The coastal WAP ecosystem is ideal for exploring these relationships given its well-established relationships with physical and environmental drivers.

The ratios of bacterial production (BP) to PP (BP:PP) are low in polar waters indicating that only a small portion of PP supports bacterial carbon consumption (Ducklow et al., 2001, 2012; Kirchman et al., 2009a). It has long been hypothesized that bacterial growth in Antarctic waters is inhibited by low temperature (Pomeroy and Wiebe, 2001), low terrestrial fluxes of biologically available dissolved organic material (DOM; Kirchman et al., 2005, 2009a), and strong top-down control via bacteriovore grazing and viral lysis (Bird and Karl, 1999). Recently, Brum et al. (2015) suggested that temperate bacteriophage infection imposed a strong top down control on BP. The review by Kirchman et al. (2009b) suggested that BP:PP ratios significantly increase with temperature, but only below 4◦C, possibly due to reduced bacterial uptake rates of organic carbon in cold waters without a significant decrease of PP at the same time. Ducklow et al. (2012) showed that temperature per se does not seem to regulate BP, but that high BP was rather associated with phytoplankton blooms. Their analysis was constrained to limited duration cruises, providing snapshots of microbial loop dynamics in the middle of summer (January). Given that not only the degree of phytoplanktonbacterial coupling but also overall bacterial responses may change across different time and space scales, it is important to revisit this issue with seasonal datasets, and over longer time spans.

Here, we report variability of bacterial dynamics in the coastal waters off Palmer Station in the WAP, using a decadal time series (2002–2014) of two bacterial properties, BP and bacterial biomass (BB), over the phytoplankton growing season (October-March). We use the term bacteria to denote heterotrophic bacteria due to a very small fraction of autotrophic bacterioplankton (usually <1% of total bacterial count) and also a low fraction of archaea in the upper 100 m of the WAP in summer (Church et al., 2003). Based on the two bacterial indices, we addressed seasonal and interannual variability of bacterial dynamics, the degree of bacterialphytoplankton coupling, and of resource (bottom-up) control on bacterial growth, and how these couplings vary seasonally and interannually. Next, we investigated a potential suite of physical and biogeochemical forcing factors influencing each bacterial variable, using time series of the relevant climate and local physical and biogeochemical parameters. We hypothesized that active bacterial processes are shaped by similar physical forcing factors as phytoplankton blooms (e.g., Saba et al., 2014), however, with complicating effects from temperature in the coastal WAP.

# METHODS

# Site Description

The study area is located at Palmer Station B (64.8◦ S, 64.1◦W) in the coastal WAP (**Figure 1**, adopted from Kim et al., 2016). Palmer Station B is relatively shallow (∼75 m) and located inshore adjacent to the coastal Marr Glacier, which has retreated significantly over the past few decades.

#### Sampling

Samples for bacterial and other ecological and biogeochemical properties (e.g., DOC, PP, Chl) were collected at Palmer Station B on a twice-weekly basis over the 2002–2003 (hereafter 2002– 2003) to the 2014–2015 Antarctic field seasons, with a Go-Flo bottle cast from a Zodiac boat. Samples were not collected during the 2007–2008 field season due to logistical reasons. Usually 4–6

samples were obtained in the upper 50 m. Initial date of sampling was determined by sea ice retreat allowing small boat access to sampling sites. Termination of each season was determined by vessel scheduling. Typically, sampling lasted from late October to late March.

#### Measurement of Bacterial Production via <sup>3</sup>H-Leucine Incorporation

The <sup>3</sup>H-leucine incorporation rate of the collected water samples was measured to derive BP rates. We followed a modified protocol of the original leucine assay proposed by Smith and Azam (1992), as described in Ducklow et al. (2012). The <sup>3</sup>H-leucine incorporation rate (pmol l <sup>−</sup><sup>1</sup> h −1 ) was converted to BP rates (mgC l−<sup>1</sup> h −1 ) using the factor 1.5 kgC mol−<sup>1</sup> leucine incorporated (Ducklow et al., 2000). We recognize that conversion factors may vary (Kirchman et al., 2009b and references therein); but adopt a constant factor to facilitate comparison with other studies (e.g., Ducklow et al., 2012).

#### Measurement of Bacterial Abundance via Flow Cytometry

BB was derived from bacterial abundance measurements. Samples for bacterial abundance were analyzed within 6-h of collection by flow cytometry following the protocol of Gasol and Del Giorgio (2000) on an Accuri C6 (Becton-Dickinson). Total bacteria concentrations were determined by adding 1 µm microspheres (Polysciences, Warrington, PA) and 5 µM final concentration of SYBR-Green stain to 0.5 ml of samples. The incubated samples were measured for 2-min at a low flow rate. Total bacterial cells and beads were enumerated in cytograms of side scatter (SSC) vs. green fluorescence (FL1). The absolute concentration of stained cells was calculated using the total sample volume analyzed, as determined by the count of the added microspheres. Bacterial abundance was then converted to BB values using 10 fgC cell−<sup>1</sup> (Fukuda et al., 1998). Thus, BB in our study simply reflects bacterial abundance integrated in the upper water-column, not incorporating the effect of cell volume changes over time and space. The time series data for both BP and BB values (Dataset 47) are archived at the Palmer LTER Datazoo: (http://oceaninformatics.ucsd.edu/ datazoo/data/pallter/datasets). Other complementary datasets for the purpose of comparing with bacterial properties are also available at the Palmer LTER Datazoo: Chl (Dataset 126), <sup>14</sup>C-PP or PP (Dataset 127), and DOC (Dataset 70) and full descriptions on their analysis and measurements are found in Supplementary Material.

#### Calculations of Bacterial Standing Stocks

Due to a weak variability of bacterial indices below 50 m and the shallow depth of Palmer Station B, <sup>3</sup>H-leucine incorporation rates and bacterial abundance were depth-integrated to 50 m. This is also the depth at which Chl and PP are typically at or near zero and at which macronutrients first reflect their maximum water-column concentrations (Kim et al., 2016). Thus, depthintegrated standing stocks for bacterial properties (mgC m−<sup>2</sup> d −1 for BP and mgC m−<sup>2</sup> for BB) provide a consistent basis of comparison across months, seasons, years, and other datasets. Prior to depth-integration, replicate values at the same depth were averaged and the values at the shallowest depth (<5 m) were extrapolated to the surface if the surface samples were missing. The values at 50 m were linearly interpolated if not sampled. The depth-integrated standing stocks were calculated only when the values for at least 3 discrete depths were measured in the upper 50 m. Otherwise, the depth-integrated stocks were estimated using a linear regression between surface and depth-integrated values (r <sup>2</sup> = 0.64–0.79, all P < 0.001).

# Empirical Orthogonal Function (EOF) Decomposition

We performed EOF analysis to investigate intraseasonal (i.e., variations within each growing season, 2002–2003 to 2014– 2015 field seasons) to interannual covariability (i.e., interannual variability of seasonal dynamics) of BP and BB. More details regarding input data matrices and seasonal grids are found in Text S1. Basically, the sample covariance (**C**ˆ) was estimated by computing biased covariance between time series from all pairs of grid cells, including mutually occupied time steps only. A standard EOF analysis is performed on this estimated covariance matrix to compute eigenstructure, where first 4–5 dominant EOFs (4 EOFs for BP accounting 86.6% of total variability and 5 EOFs for BB accounting 97.9% of total variability after visually examining eigenvalue and scree plots) are combined to provide a smooth, reduced-space interpolant data across the grid in seasons and years (i.e., reduced space optimal analysis (RSOA); Kaplan et al., 1997). More details on EOF computations are available in Text S1.

### Generalized Linear Models (GLMs) by Stepwise Regression

We established GLMs for BP and BB by stepwise regression to investigate large-scale climate and local physical and biogeochemical forcing factors predicting high BP and BB, respectively. The predictor (independent) variables tested for model selection are summarized in Table S1 and Text S2. When performing stepwise regression analysis for GLMs, it is critical that a response (dependent) variable being modeled is normally distributed. To ensure each data set comes from as much like a normal distribution as possible, the BP dataset was log<sup>10</sup> transformed and the BB dataset was square root transformed.

Model selection of GLMs was based on the corrected Akaike Information Criterion (AICC). Typically, an ordinary, uncorrected version of AIC is used to determine the significance of each predictor variable in the model, where the residual sum of squares is penalized by twice the number of parameters times the residual mean square of the initial model (Akaike, 1974). The usage of the AIC<sup>C</sup> allows, additionally, the selection of the model with the closest to optimal balance between goodness of fit and parameterization, while making a correction for small sample size typically as in the case of ecological time series (Burnham and Anderson, 2004). Predictor variable terms (Text S2, Table S1) were added and removed in order of greatest reduction in the AIC<sup>C</sup> value in a stepwise manner (i.e., a combination of backward elimination and forward selection). Then, the final GLMs were determined from a sequence of the steps, which minimizes the AICC, associated statistics, and has a fewer number of predictor variables, if possible, to avoid overfitting in the model. The models tested include multiple types, including linear models (e.g., constant, linear models) and non-linear models (e.g., quadratic, pure quadratic, and interactions). Finally, statistical stability of the established BP and BB GLMs was assessed based upon two criteria: (1) normality of the residuals of the selected GLMs and (2) a low degree of multicollinearity of the used predictor variables. We confirmed based on normality tests (e.g., Jarque-Bera, Lilliefors, and χ 2 goodness-of-fit) that the residuals between modeled and observed BP and BB are from a normal distribution at a 95% confidence interval (CI). This indicates that linear relationships between the predictor and response variables are not affected by outliers or highly skewed predictor variables. Multicollinearity, a statistical phenomenon where predictor variables in a multiple regression model are highly correlated, leads to a reduction in the ability to detect reliable effects of correlated variables. To examine how much the variance of the coefficient estimates are being inflated by multicollinearity in each bacterial GLM, we calculated the variance inflation factors (VIFs) for each predictor variable used in the GLMs. The selected GLMs and associated statistics are summarized in **Tables 1**–**2**.

#### RESULTS

#### Depth Distributions of Bacterial and Phytoplankton Properties

Depth contours of temperature and salinity showed seasonal evolution in the upper water column with significant interannual variability in terms of the timing. Seasonal warming of the water column started from early to late December (day 340–365) with corresponding decreases of salinity (i.e., freshening) from surface of the water column (**Figures 2A,B**). High <sup>3</sup>H-leucine incorporation (>50 pmol l−<sup>1</sup> h −1 ) or large bacterial abundance (>10<sup>9</sup> cells l−<sup>1</sup> ) generally followed high phytoplankton values (>100 mgC m−<sup>3</sup> d −1 ; >5–10 mg Chl l−<sup>1</sup> ) with varying lags from a few days (days of year 335–350 in 2012–2013; **Figures 2C–E**) up to a month (days of year 410–425 in 2005–2006 and days of year 365–395 in 2012–2013; **Figures 2C–E**), though not always. For example, there was anomalously high <sup>3</sup>H-leucine incorporation in 2010–2011 (day of year 365–410; **Figure 2E**), but with only a slight increase of Chl (**Figure 2D**) and a lack of high PP (**Figure 2C**). This discrepancy was also observed in the year 2011–2012. Conversely, high Chl and PP did not result in high <sup>3</sup>H-leucine incorporation events as shown in December of 2003–2004. Furthermore, high <sup>3</sup>H-leucine incorporation did not always co-occur with large bacterial abundance as shown in years 2010–2011, 2011–2012, and 2012–2013 (**Figures 2E,F**).

# Bottom-Up Control on Bacterial Growth

To assess the degree of bottom up (resource or substrate) control on bacterial growth, we examined slope values from significant log-log regressions of BP on BB (Billen et al., 1990; Ducklow, 1992) both for seasons and individual years (**Figure 3**). For annual slope values (**Figure 3A**) slope values were calculated from BP-BB regressions by pooling all monthly data together in each individual year. For seasonal slope values (**Figure 3B**), slope values were calculated from BP-BB regressions by pooling all yearly data together in each particular month. The slope value plots from BP-BB regressions are also overlaid with slope values from SST-BP regressions to compare bottom-up control with the strength of temperature control on bacterial activity (Morán et al., 2001). When pooled for all seasons and years, the slope value from BP-BB regressions was 0.35 (r 2 = 0.26, P < 0.001, n = 333; **Figure 3C**), while the slope value from SST-BP regressions was 0.14 (r <sup>2</sup> = 0.20, P < 0.001, n = 314; **Figure 3D**), which is slightly lower than the BP-BB regression.

TABLE 1 | Summary of generalized linear models (GLMs) and associated statistics for monthly composites of BP during summertime only (A) and during entire growing seasons (B).

#### A. BP (DJF)

Log10BPDJF = a<sup>1</sup> + a<sup>2</sup> (Sea ice extent)July + a<sup>3</sup> (MLD)ND + a<sup>4</sup> (Log10PP) + a5 (Log10DOC)


Distribution: Normal


(Continued)

Annual slope values (only from significant relationships) of BP-BB regressions ranged from 0.23 in 2003–2004 to 0.58 in 2009–2010, showing a considerable interannual variability in the degree of bottom-up control (**Figure 3A**). Unlike significant bottom-up control in most years, temperature control was significant only for half of our study years. Temperature control was never stronger than bottom-up control in any years (**Figure 3A**).

Seasonally, the regression slope values for BP-BB were very consistent with a value near 0.32 (**Figure 3B**). Most seasons showed a significant bottom-up control except for October. However, temperature control was only significant for less than the half the growing seasons, only evident in December and February (**Figure 3B**). The strength of temperature control on BP was higher in February than in December.

#### Bacterial-Phytoplankton Coupling

To assess the degree of bacterial coupling with phytoplankton processes, we examined slope values from significant log-log regressions between PP or Chl (as an independent variable, X) and BP (as a dependent variable, Y). The regressions were assessed seasonally and annually. When pooled for all seasons and years, the slope value from log-log regression of Chl-BP was 0.32 (r <sup>2</sup> = 0.12, P < 0.001, n = 271; **Figure 4C**), while the slope

#### TABLE 1 | Continued

#### B. BP (ONDJFM)

#### Log10BPONDJFM = a<sup>1</sup> + a<sup>2</sup> (SAM)July + a<sup>3</sup> (SST) + a<sup>4</sup> (Sea ice extent)July + a5 (MLD)ND + a<sup>6</sup> (Density gradient)ND+ a<sup>7</sup> (Log10PP)

Estimated Coefficients


Distribution: Normal

AIC<sup>C</sup> = −24.8

Observations = 38

Error degrees of freedom = 31

Estimated dispersion = 0.023

F-statistic vs. constant model: 16.7 <sup>P</sup> <sup>=</sup> 1.7×10−<sup>8</sup>

Variance Inflation Factors (VIFs)\* SAMJuly = 1.35 SST = 1.09

Sea ice extentJuly = 1.83 MLDND = 1.17 Density gradient ND =1.80 Log10PP = 1.88

The monthly composites of summertime (December-February or DJF, Table 1a) and of entire growing seasons (October-March or ONDJFM, Table 1b) BP were expressed as a linear sum of different physical and biogeochemical predictor forcing variables as shown in the equations below. The tables present regression coefficients (a1–a5) for a particular independent forcing variable and their standard errors (S.E.), t-statistics, and P-values. Larger absolute values of t-statistics indicate a larger influence of a particular predictor variable on the given dependent variable. The best model selection was via achievement of the lowest AIC<sup>C</sup> values in a stepwise manner. Statistical stability of the selected GLMs was tested based upon the normality of residuals and the low degree of multicollinearity. <sup>+</sup>Following a typical threshold value of 4.0 at which the multicollinearity problem can be ignored for a particular predictor variable (O'Brien, 2007), we confirmed that all VIFs in each model are <4.

\*All the VIFs in each model are <4.

value from log-log regression of PP-BP was 0.17 (r <sup>2</sup> = 0.07, P < 0.001, n = 230; **Figure 4D**). The BP:PP ratio **Figure 5B**) was calculated from annually integrated PP and BP (days 290–440; **Figure 5A**). The 11-year mean BP:PP ratio was 0.04, showing considerable interannual variability in the ratio. Annual-scale PP-BP regressions were largely absent except for 2005–2006 and 2010–2011 with slope values of 0.57 and 0.46, respectively (**Figure 4A**). In contrast, significant Chl-BP regressions were more frequent with slope values ranging from 0.33 to 0.56 (years 2005–2006, 2009–2010, 2010–2011, 2013–2014, and 2014–2015; **Figure 4A**) showing a considerable interannual variability in the degree of Chl control on BP.

TABLE 2 | Summary of generalized linear models (GLMs) and associated statistics for monthly composites of BB during summertime only (A) and during entire growing seasons (B).


Distribution: Normal

AIC<sup>C</sup> <sup>=</sup> 90.9 Observations = 20 Error degrees of freedom = 15 Estimated dispersion = 3.6 F-statistic vs. constant model: 8.01 P = 0.001

Adjusted r <sup>2</sup> = 0.60

Variance Inflation Factors (VIFs)<sup>+</sup> Wind speed = 1.14 SST = 1.36 Density gradientND = 1.50 Log10PP = 1.72

B. BB (ONDJFM)

BB1/<sup>2</sup> ONDJFM <sup>=</sup> <sup>a</sup><sup>1</sup> <sup>+</sup> <sup>a</sup><sup>2</sup> (Wind speed) + a<sup>3</sup> (SST) + a<sup>4</sup> (MLD)ND +


Distribution: Normal

AIC<sup>C</sup> <sup>=</sup> 169.6

Observations = 38

Error degrees of freedom = 31

Estimated dispersion = 5.41 F-statistic vs. constant model: 11.4

<sup>P</sup> <sup>=</sup> 8.59 <sup>×</sup>10−<sup>6</sup> Adjusted r<sup>2</sup> = 0.54

Variance Inflation Factors (VIFs)<sup>+</sup> Wind speed = 1.40 SST = 1.33 MLDND = 1.06 Density gradientND = 1.09

<sup>+</sup>All the VIFs in each model are <4.

Seasonally, PP-BP regressions were only significant in October, January, and February, while significant Chl-BP regressions were more frequently observed (November to February; **Figure 4B**). The PP-BP regression slope values (only significant relationships) ranged from 0.26 to 0.41. The slope values from significant Chl-BP regressions ranged from 0.25 to 0.38, showing a similar seasonal variability as PP-BP regressions.

#### Climatology of Seasonal Bacterial Dynamics

The individual year plots indicate strong seasonal and interannual variability of BP and BB values (**Figure 6**). From a climatological (2002–2014) perspective, on average, there was a gradual, seasonal increase of BP rates from a seasonal minimum of 11.4 mgC m−<sup>2</sup> d −1 to maximum at 46.4 mgC m−<sup>2</sup> d −1 in mid-January (**Figure 7A**). BP declined to ∼109% of its starting value (seasonal minimum). In contrast, climatological BB showed somewhat a different pattern than BP, with a sustained increase as the mean growing season progressed, reaching a maximum of 419 mgC m−<sup>2</sup> at the end of the sampling period (**Figure 7B**). Thus, we recognize the possibility of increasing BB values after the end of our sampling period, but an analysis for extended seasons until Austral winter is beyond the scope of our study. PP (**Figure 7C**) and Chl climatologies (**Figure 7D**) showed similar patterns to the BP climatology, suggesting a close phytoplankton-bacterial coupling.

#### Interannual Variability of Seasonal Bacterial Dynamics

This section describes seasonal variability patterns (i.e., EOF) of BP and BB, and how these patterns vary interannually (i.e., PC). We report first the two dominant EOF modes (EOF1-2; **Figures 8A,D**) and their PC time series (PC1-2; **Figures 8C,F**) to consider major seasonal patterns (phenology) of bacterial processes. Next, we present anomalies relative to the 11-year climatology of the given seasonal time bins (RSOA anomalies; **Figures 8B,E**).

BP mode 1 (42% of total BP variability) was characterized by a gradual increase of BP rates starting with a slight increase in late October, a peak in January, and then decreases after February (BP EOF1, **Figure 8A**), similar to the climatology. The seasonal maximum of BP occurs from December to February (DJF) and is therefore termed "DJF bacterial productivity." The years of high DJF bacterial productivity were represented as high PC1 values (years 2005–2006, 2010–2011, 2012–2013, and 2013–2014; **Figure 8C**). BP mode 2 (20% of total BP variability) mainly captured the rest of the variability left out from BP mode 1, showing elevated BP rates during November to the first half of December. The years of this high November-early December, "spring bacterial productivity" were observed as high PC2 values (years 2004–2005 and 2012–2013; **Figure 8C**).

BB mode 1 (49% of total BB variability) represented high BB throughout growing seasons with slightly higher biomass during

year 290 to 440 or mid-October through March. Black triangle marks on the x-axis indicate sampling time points.

DJF than other months (BB EOF1, **Figure 8D**). This overall, seasonally increased BB pattern was captured by large PC1 (BB PC1, **Figure 8F**; years 2005–2006, 2013–2014, and 2014–2015). Relatively speaking, BB mode 2 (22% of total BB variability) represented large BB especially in the beginning of the season (October to first half of November) and corresponded to large PC2 (BB PC2, **Figure 8F**; years 2002–2003, 2010–2011, and 2012–2013).

#### Physical and Biogeochemical Forcing Factors of Bacterial Dynamics

Finally, based on our results from bacterial GLMs (BP and BB GLMs) we explored a potential suite of physical and biogeochemical forcing factors responsible for the observed variability of each bacterial property. We first examined monthly summer (DJF) composites of BP and BB since they show shared strong positive responses from the EOF plots during

values (A), the slope values for that year were calculated by pooling all seasonal months' data in that particular year. For seasonal slope values (B), the slope values for each month were calculated by pooling all individual years' data in that particular month. When pooled for all seasons and years (C,D), the regression slope value for BP-BB was 0.35 (r <sup>2</sup> = 0.26, P < 0.001, n = 333) and the regression slope value for SST-BP was 0.14 (r <sup>2</sup> = 0.20, P < 0.001, n = 314). Error bars indicate standard deviation of the slope value at 95% Confidence Interval (CI).

these months (**Figures 8A,D**). To see if bacterial responses to physical and biogeochemical forcing factors differ outside of the DJF period, we additionally examined BP and BB GLMs over entire phytoplankton growing seasons (October to March or ONDJFM). Thus, bacterial GLMs (2002–2014) here explain an occurrence of years with high monthly composites of BP and BB standing stocks during summertime (DJF) periods (**Tables 1A**, **2A**), as well as over entire growing seasons (**Tables 1B**, **2B**).

During summertime (DJF), significant physical and biogeochemical forcing variables shaping high BP included shallow MLD, high bulk DOC, and high PP (in order from stronger to weaker forcing; **Table 1A**). If extended to entire growing seasons, high SST was additionally observed as a significant physical predictor variable for high BP rates: high BP rates during ONDJFM were influenced by high PP, high SST, and shallow MLD (stronger to weaker forcing; **Table 1B**).

The BB GLMs during summertime (DJF) showed that large BB was influenced by high salinity-driven density gradient and high PP (stronger to weaker forcing; **Table 2A**). SST was also a significant forcing factor for the BB GLM, but its effect is comparatively lower than other forcing variables based on its lower absolute t-statistic value (**Table 2A**). As observed in the BP GLM during entire growing seasons, high SST was also shown as a significant predictor variable (in this case, as a "stronger" forcing variable based on a higher absolute t-statistic value than BB GLM for DJF) for large BB if extended to entire growing seasons: large BB was shaped by high salinity-driven density gradient and high SST (stronger to weaker forcing; **Table 2B**).

#### DISCUSSION

Studies have investigated dynamics of heterotrophic bacteria in the WAP and adjacent Antarctic ecosystems, with a focus

<sup>2</sup> = 0.07, P < 0.001, n = 230). Error bars indicate standard deviation of the slope value at 95% CI.

on a variety of processes, including bacterial responses to environmental gradients, coupling with phytoplankton, and community structures (e.g., Karl and Tien, 1991; Karl et al., 1991; Morán et al., 2001; Morán and Estrada, 2002; Ducklow et al., 2012; Luria et al., 2014, 2016; Nikrad et al., 2014; Bowman and Ducklow, 2015). However, these studies are mostly short-term and usually constrained to just 1 or 2 years. Our seasonally extended (October–March) and decadal time series analysis of bacteria highlights a set of patterns that have not been identified in the previous work, including time-varying temperature control on BP and consistent DOC limitation throughout the phytoplankton growing season. Our discussion begins with a broad overview of bacterial dynamics using the depth contour plots for each season and year (see below).

#### Overview of Bacterial Dynamics Based on Depth Profiles

To first order, depth profiles showed that high <sup>3</sup>H-leucine incorporation rates generally followed large phytoplankton blooms and were also almost always found in the upper water column coinciding with high PP, thereby indicating a frequent, but variable coupling between bacterial and phytoplankton processes in the coastal WAP waters. Bacteria and phytoplankton appeared to be coupled (i.e., high bacterial events following large phytoplankton events), with varying degrees of lag between the two. The different degrees of bacterial-phytoplankton coupling could be attributed to different production rates of DOM from phytoplankton assemblages and its rate of subsequent bacterial utilization, which is often a function of different bacterial clades (Straza et al., 2009; Nikrad et al., 2014). It was shown in Antarctic waters that up to a month was needed for bacterial blooms to respond to diatom blooms due to delayed production and utilization of macromolecular, polymeric DOM (Billen and Becquevort, 1991; Lancelot et al., 1991), compared to shorter lags of a few days at lower latitudes (Billen et al., 1990). If bacterial blooms are dominated by fast growing bacteria such as gammaproteobacterial clades (e.g., Ant4D3, Arctic96B-16), there may be a relatively short lag between high BP and

regression slope value for PP-BP was 0.17 (r

phytoplankton bloom events, in response to low-molecularweight, readily utilizable DOM than when dominated by slowly growing bacterial clades such as SAR86 (Nikrad et al., 2014). Conversely, if bacterial metabolism is mainly based on highmolecular-weight DOM, it likely leads to a longer lag. Luria et al. (2016) demonstrated that different individual bacterial taxa were dominant over the course of the bacterial bloom in the coastal WAP, which contributed to different time lags in the bacterial responses to DOM. On the other hand, varying lags of bacterial-phytoplankton coupling might also happen if bacteria rely not only on LDOC but also on semi-labile DOC (SDOC) from previous months' or years' phytoplankton accumulations (Kirchman et al., 2001; Davis and Benner, 2005), which leads to a seemingly uncoupling or a coupling with a long lag (more detailed discussion on SDOC is found in Section Biogeochemical Forcing Factors of Bacterial Dynamics). In the Ross Sea, it was shown that bacterial growth was strongly dependent on previously accumulated SDOC, at different rates over the course of growing seasons, but SDOC was eventually entirely consumed by the end of the season (Ducklow et al., 2001; Ducklow, 2003). Using 16S rRNA gene phylotype-derived prediction of microbial metabolic pathways, Bowman and Ducklow (2015) also showed that bacterial communities presented a differential distribution for pathways associated with the degradation of DOC pools along the WAP, which likely implies different bacterial clades might be responsible for utilization of different DOC pools in our study.

In contrast, BP was occasionally decoupled from PP or Chl (e.g., December of 2003–2004; **Figures 2C–E**) where high phytoplankton processes did not lead to high BP events. This implies that some other environmental factor prevents high bacterial activity despite availability of organic carbon sources for bacteria. One possibility is low temperature, which has long been hypothesized to cause low BP to PP ratios in polar waters (Pomeroy and Wiebe, 2001). In our depth profiles, bacterial accumulations were rarely observed in early spring when SST was still low (< ∼0.5◦C; **Figures 2A,E,F**), implying that there might be at least some degree of temperature control on bacterial activity in the coastal WAP waters. However, presumably due to small sample size (n = 9) during the time of the observed decoupling (December of 2003–2004; **Figures 2C–E**), SST was not significantly correlated to <sup>3</sup>H-leucine incorporation rates, thereby precluding a test of a direct effect of temperature. Regulation of diverse bottom up control factors (e.g., Chl, PP) on BP is often manifested at larger scales (e.g., Dufour and Torreton, 1996; Ducklow et al., 2012). Thus, by pooling all years' data in each seasonal month, we subsequently investigated the degree of temperature control on BP with greater sample size, using the slope values from SST-BP regressions seasonally (See Section Bottom-Up Control on Bacterial Abundance).

Our observations also showed that high BP did not always lead to large BB accumulations (e.g., 2010–2011, 2011–2012, and 2012–2013). Here, top down control factors might play a significant role in causing BP-BB decoupling, given that BB could only increase if bacterial growth exceeds removal by grazers or bacterivore grazers are not capable of balancing the production of enlarging bacterial cells (Gasol et al., 1999). Antarctic waters are often characterized with a high grazing pressure of protozoan grazers on bacteria, which strongly limits large bacterial accumulations (Bird and Karl, 1999). In the coastal WAP, microzooplankton were specifically shown to exert a substantial grazing pressure on bacteria, often removing >100% of BP (Garzio et al., 2013). Viral infection and lysis were also shown as critical factors for causing significant bacterial mortality in Antarctic waters (Guixa-Boixereu et al., 2002; Brum et al., 2015). In the following Section Bottom-Up Control on Bacterial Abundance, we discuss in detail the effect of bottom-up control on bacterial abundance from our data, primarily using the slope values from BP-BB regressions.

# Bottom-Up Control on Bacterial Abundance

In a steady-state system where the bacterial utilization rate of DOM is coupled to its rate of supply, BP may be considered a proxy for DOM input flux and thus indicative of bottomup control (Billen et al., 1990). Traditionally, predominance of bottom-up control on biomass of a certain trophic level has been explored from log-log relationships between abundance (biomass) of that trophic level and presumed controlling factors (McQueen et al., 1986). Following this approach, the strength of bottom-up control on bacterial abundance has been assessed in diverse ecosystems using the slope of significant log-log regressions of BP (DOC)-BB (Shiah and Ducklow, 1995; Dufour and Torreton, 1996; Gasol et al., 2002; Hale et al., 2006; Garneau et al., 2008; Morán et al., 2010; Calvo-Díaz et al., 2014;

Ducklow, 1992). Higher slope values (i.e., stronger degree of BB dependence on BP) indicates stronger responses of BB to increases in substrate (DOC) supply or BP, implying that bacteria experience substantial DOC limitation. Conversely, if BB is not responsive to increases in BP, bacteria are not DOC-limited; or possibly are efficiently removed by grazers and viruses.

The regression slope values from our BP and BB data (**Figure 3B**) indicate a very consistent bottom-up control across seasons with an average slope value of around 0.32. This value reflects a weak to moderate bottom-up control on bacterial growth (Ducklow, 1992). The slope value was never below the threshold slope value of 0.2 (Ducklow, 1992), implying that

FIGURE 7 | Climatology of BP (A), BB (B), PP (C), and Chl (D). Climatology (2002–2003 to 2014–2015) of depth-integrated (0–50 m) BP (A), BB (B), PP (C), and Chl (D) for a full growing season. The data in the calculation are RSOA-interpolated data. Error bars indicate 1 standard error.

bacteria in this system were always under some sort of resource limitation over entire growing seasons. Resource-wise, bacteria could be limited by DOC, dissolved inorganic nutrients (e.g., nitrate and phosphate) or trace metals such as iron. However, there is no evidence of macronutrient and iron limitation on phytoplankton at this site (Kim et al., 2016 Sherrell, personal comm.). Experiments by Ducklow et al. (2011) also suggested that the summertime bacterial community is not nitrogen limited in the WAP coastal waters, given that bulk bacterial properties (e.g., BP and abundance) significantly increased in response to experimental additions of glucose, but not to ammonium enrichment. Calvo-Díaz et al. (2014) suggested a seasonal switch from temperature to substrate supply as a major controlling factor for bacterial abundance in the coastal Bay of Biscay, based on their observations of decreasing slope values from temperature-BP regressions and increasing slope values from BP-BB regressions as seasons progress. The WAP system differs substantially from their observations from estuaries (e.g., Shiah et al., 2003). In our study, temperature control on BP was only significant in December and February, while consistent degrees of bottom-up control on BB were evident over all months (**Figure 3B**). Similar patterns were observed from individual years' slope values (**Figure 3A**). Only the half of years showed a weak to moderate degree of temperature control on BP, while most years were under weak to moderate DOC limitations. When pooled for all years and seasons, temperature control was considerably weaker than bottom-up control (**Figures 3C,D**). Based on these observations, we suggest **(**1) that bacteria in coastal WAP waters are always under some degree of DOC limitation and **(**2) that both yearly and seasonally temperature also influences BP but the effect may not be as large and consistent as DOC limits bacterial processes.

#### Bacterial-Phytoplankton Coupling

In the coastal WAP ecosystem heterotrophic bacteria may feed principally on in situ photosynthetic products released from phytoplankton to fuel BP (Baines and Pace, 1991; Nagata, 2000) due to minimal inputs of terrestrial DOM and low bulk DOC concentrations (40–50µmol C l−<sup>1</sup> ) in the water column (Ducklow et al., 2012). The mean BP:PP ratio across all years and seasons indicated that BP was equivalent to only 4% of PP, similar to the greater WAP region and other polar areas (Kirchman et al., 2009b; Ducklow et al., 2012). The degree of the dependence of bacteria on phytoplankton is often assessed as BP:PP ratios with the reasoning that higher values mean tighter coupling. However, we based our discussion regarding the covariation between bacteria and phytoplankton or bacterial-phytoplankton coupling on the slope values of significant regressions between bacterial and phytoplankton variables (Morán et al., 2001; Morán and Estrada, 2002), not the absolute value of the connection. When pooled for allseasons and years, the PP-BP slope value (i.e., the degree of bacterial-phytoplankton coupling or PP control of BP) was 0.17 in our study (**Figure 4D**). This is equivalent to 0.004 if recalculated from the untransformed (linear) regression, not significantly different from the regression slope value reported from Ducklow et al. (2012) for the peninsula-wide offshore region during January.

BP was coupled with PP only in October, January, and February, while BP was mostly coupled to Chl with stronger degrees of control (i.e., higher regression slope values; **Figure 4B**). Most years showed weak to strong coupling between Chl and BP with strong interannual variability in the degree of Chl control on BP (**Figure 4B**), while only two years showed a significant direct control of PP on BP. On average (over all seasons and years), the degree of Chl control on BP (i.e., slope value of 0.32) also appeared to be stronger than that of PP control on BP (i.e., slope value of 0.17). Thus, these results consistently suggest an overall tighter coupling of BP with Chl than with PP. This implies that overall accumulations of phytoplankton biomass might support BP, but not necessarily carbon flux from PP directly; i.e., release of DOC from active phytoplankton with the timescale of the <sup>14</sup>C-PP measurements (∼1 day). This point was also raised by Ducklow et al. (2012) to support their observation of tighter coupling between Chl and BP as well, which they attributed to bacterial utilization of DOC from other sources like zooplankton grazing and sloppy feeding on the timescale of phytoplankton stock turnover (∼10 days).

Our analysis indicates that only a very minimal portion (i.e., 4%) of fixed PP supports BP in the coastal WAP, corresponding to other studies in polar oceans (Ducklow et al., 2001, 2012; Kirchman et al., 2009a). In the Ross Sea, Carlson et al. (1998) showed that up to ∼90% of phytoplankton-production is partitioned to the POC pool, rather than routed through the DOC pool, which accounts for only about 10% of total carbon fixed. Low accumulations of DOC were previously reported after phytoplankton blooms in the WAP region (Tupas et al., 1994). These findings support our conclusion that the limited supply of DOC is a primary reason for low BP:PP ratios. Such a low BP fraction of PP begs a question about the fate of the large amount of unused PP or POC within the food-web of the coastal WAP ecosystem. Using in situ O2:Ar measurements at Palmer Station, Tortell et al. (2014) showed there was a period of intense net heterotrophy during the crash of a phytoplankton bloom. They implicated increased respiration by zooplankton as a major fate of bloom production. Stukel et al. (2015) also demonstrated with <sup>15</sup>NO<sup>−</sup> 3 uptake and <sup>234</sup>Th measurements that new production exceeded export production by a factor of five at Palmer Station. Hence, it is likely that a majority of PP is respired by intense heterotrophic respiration in the upper water column, rather than appearing as bacterial production or being directly exported to depth.

# Climatology of Seasonal Bacterial Dynamics

Consistent with strong seasonal and interannual variability of other ecological and biogeochemical properties along the WAP (Ross et al., 2008; Smith et al., 2008; Vernet et al., 2008; Ducklow et al., 2012; Saba et al., 2014; Kavanaugh et al., 2015; Kim et al., 2016), BP and BB also showed strong seasonal and interannual covariability (See Section Interannual Variability of Seasonal Bacterial Dynamics). Accordingly, the climatologies did not reflect occurrence of high BP or bacterial blooms for any one particular year. Both BP and BB values commenced with very low values, 4-fold and 2.8-fold lower than seasonal maxima, respectively, consistent with very low BP values during austral winter seasons (i.e., leucine incorporation rates of typically <5 pmol l−<sup>1</sup> h −1 ).

Notably, seasonal progression patterns were largely different between BP and BB where the annual BP maximum occurred in the middle of the growing season, while the BB maximum was observed at the end of or possibly even after our growing seasons due to its sustained increase throughout the growing season. The slope of the increase of climatological BB over time (our sampling period from October to March) was ∼1.8 mgC m−<sup>2</sup> per day, <5% of the maximum production rate, indicating the longer-tem balance between bacterial growth and removal.

### Interannual Variability of Seasonal Bacterial Dynamics

From the BP and BB EOF plots, the first two dominant modes mostly captured high BP and large BB from the start to nearly end of our sampling period. However, the leading modes (EOF1) of BP and BB did not show the same dominant seasonal patterns between the two properties, in accord with our aforementioned observations (Section Overview of Bacterial Dynamics Based on Depth Profiles), climatologies and individual year plots (Section Climatology of Seasonal Bacterial Dynamics). It was previously shown that the leading EOF modes of two different biogeochemical variables might become alike, if similar physical or biological processes drive variability of each variable (e.g., nitrate and phosphate EOF patterns in Kim et al., 2016). There is no a priori reason that BP needs to have the same EOF patterns with BB if the two properties are shaped by different regulatory causes, again supporting such factors' roles (e.g., temperature, SDOC, DOC from other sources, grazing) described in previous Sections Overview of Bacterial Dynamics Based on Depth Profiles and Bacterial-Phytoplankton Coupling. The interplay of these factors in influencing the interannual variability of BB and PP is considered in the next section using GLMs via stepwise linear regression.

# Physical and Biogeochemical Forcing Factors of Bacterial Dynamics

It is well demonstrated that large-scale climate and associated local physical forcing (e.g., sea ice and ice melt drivenstratification of the upper water column) control the timing and extent of phytoplankton blooms and PP along the WAP (Smith et al., 2008; Vernet et al., 2008; Venables et al., 2013; Saba et al., 2014) and adjacent regions (Schloss et al., 2012). Thus far, we have shown evidence of close coupling between bacteria and phytoplankton. In accord with this, we hypothesized that due to a close coupling between bacteria and phytoplankton, similar but possibly slightly different regulatory mechanisms to those of phytoplankton blooms may influence active bacterial processes as a consequence of complicating effects from other controlling factors.

#### Physical Forcing Factors of Bacterial Dynamics

A stabilized upper water-column via shallow MLD or high salinity-driven density gradient was a significant physical predictor in driving high BP and large BB during summertime (DJF) as well as over entire growing seasons. This is similar to regulatory factors for large phytoplankton accumulations in inshore waters off Palmer Station (Saba et al., 2014; Kim et al., 2016). In the coastal WAP, phytoplankton blooms are primarily induced once they are free from light limitation when the upper MLD begins to shoal (Mitchell et al., 1991; Venables et al., 2013). In turn, shoaling of the upper MLD is controlled by seasonal seaice dynamics (e.g., timing of retreat, winter ice extent) since ice melt water plays an important role in stratifying the upper water column (Smith et al., 1998).

It is noteworthy that high SST was also a significant physical predictor for both high BP and large BB, if extended from summertime to entire growing seasons. However, SST neither shaped high BP during summertime alone (DJF) nor played an important role for large BB during DJF as it did over entire growing seasons. It is possible that simply due to small sample size, a correlation between SST and BP (or BB) was absent (or low) during DJF or that a response to temperature only manifested when looking at a longer temporal scale, in this case the seasonal scale, due to a more important role of DOC.

Besides temperature effects, other physical and climate forcings could directly control BP, not modulated through phytoplankton processes. Large winter (July) sea ice extent was a significant physical predictor although its effect was lower than other stronger physical forcings (e.g., MLD, SST). This winter sea ice effect seemed to be somewhat stronger when extended to entire growing seasons (BP GLM for ONDJFM) compared to summertime (BP GLM for DJF). Observations of bacterial and phytoplankton properties were not available in early spring (October-November) of 2004–2005, but RSOAinterpolated anomalies indicated that BP was characterized by strong positive anomalies in contrast to negative anomalies for Chl and PP during that time (**Figure 8B**, Figure S1). The RSOAinterpolated plot for 2004–2005 also showed the similar patterns (**Figure 6A**, Figure S2). Thus, there was a high BP event during early spring months of 2004–2005, where bacteria did not depend on phytoplankton-derived fresh LDOC. Notably, the year 2004 was an anomalously heavy sea-ice year compared to the 11 year climatology during our field seasons, with large winter (July) sea ice extent (193,073 km<sup>2</sup> ) and total ice cover (12,066 % days) as well as late retreat (day 395). Winter sea ice had not yet totally retreated at the time of the high BP event. One possibility in support of such high BP rates under high sea ice conditions is an additional supply of DOM from ice melt. It was shown by several studies that DOC could be captured in sea ice during ice formation and utilized by bacteria as ice melts (Kähler et al., 1997; Fortier et al., 2002; Müller et al., 2013).

#### Biogeochemical Forcing Factors of Bacterial Dynamics

High PP was observed as a common biogeochemical predicting factor for both high BP and large BB. Unlike BP GLMs where PP increased BP regardless of seasons, BB increased in response to PP only during summertime. Bulk DOC was a significant predictor for high BP only if extended to entire growing seasons, not during summertime, even though our data showed no significant relationship between bulk DOC concentrations and Chl or PP (data not shown). Semi-labile DOC (SDOC, turnover time of seasons) is biologically available for bacterial incorporation in polar waters (Kirchman et al., 2001; Davis and Benner, 2005). In the Ross Sea, the entire SDOC pool was utilized for bacterial metabolism by the end of growing season (Ducklow, 2003). This result implies that bacteria in the Ross Sea were under DOC limitation even during phytoplankton blooms, similar to our results of consistent DOC limitation yearly and throughout seasons. Given possible turnover times of some SDOC fractions of up to several years (Kirchman et al., 2001), and the lack of correlations between bulk DOC and Chl or PP from our datasets, it is likely that SDOC produced from previous years or months support bacterial growth in the WAP system similar to the Ross Sea (Carlson et al., 2000; Ducklow, 2003).

### Implications for Climate Change along the Coastal Antarctic Ecosystem

The WAP has responded to regional climate change with significant winter atmospheric warming, ocean warming, reduced sea ice duration, and retreat of glaciers and ice sheets (Liu et al., 2004; Cook et al., 2005; Vaughan, 2006; Stammerjohn et al., 2008). The Marr Glacier, located adjacent to Palmer Station B, has retreated, potentially releasing L- or SDOC available to heterotrophs in coastal waters (Cook et al., 2005; Hood et al., 2009, 2015). The heat content of Upper Circumpolar Deep Water (UCDW), regularly spilled into WAP shelf waters, has also increased (Martinson et al., 2008), and its delivery is further expected to increase as a consequence of an increasing trend in the positive SAM phase (Lubin et al., 2008). This may contribute to an overall baseline shift in temperature or increasing UCDW events on the shelf, which may influence BP rates.

In our study, the results from bacterial GLMs showed that bulk DOC was a significant forcing predictor for BP. Together with the results from bottom-up control (Section Bottom-Up Control on Bacterial Abundance), this implies that bacterial growth in the coastal WAP is still under-saturated "resourcewise" (i.e., DOC-limited). The current trend of climate change will enhance water-column stratification through higher surface water temperature and increased freshwater flux from sea ice and glacial melts. As a result, LDOC pools may significantly increase, resulting in more efficient activity of heterotrophic bacteria in the WAP food-web. Recent model findings demonstrated that instead of dominance by the efficient diatom-krill-penguin food chain, the WAP presents a strong interannual variability in food-web dynamics, with an important role of microbial food-webs initiated by small cells such as cryptophytes (Sailley et al., 2013). Retreating glaciers as a result of warming could also increase abundances of freshwater favorable cryptophytes (Moline et al., 2000; Garibotti et al., 2005), further contributing to a shift to a microbial food web dominated system, where heterotrophic bacterial roles in carbon and biogeochemical recycling in the coastal ecosystem also become more prominent.

# CONCLUSIONS

In conclusion, our seasonally-extended analysis of bacterial dynamics using a decadal (2002–2014) time series of BP and BB revealed a set of patterns which have not been identified previously due largely to the short-term, temporally limited focus of most previous studies. Our major findings highlight strong seasonal and interannual variability in the degree of bacterial coupling (or decoupling) with phytoplankton properties, with varying lags presumably due to different bacterial clades' roles in DOM uptake and bacterial reliance on SDOC in the system. BP was more tightly coupled with Chl than with PP, suggesting that bacteria in the WAP system also rely on DOC produced from diverse trophic levels and processes, not solely on LDOC of recent phytoplankton origin. The degree of bottom-up control on bacterial growth showed that bacteria in this system are under a consistent degree of DOC limitation during entire growing seasons, with a great interannual variability in the strength of such control. Temperature also seems to play an important role, but not as much as DOC in causing low BP:PP ratios (i.e., only ∼4% of PP supports BP) in coastal WAP waters. Top-down effects of microzooplankton grazing and viral lysis also seem to exert important roles in decoupling of BB from BP. All these aforementioned factors accordingly appeared as significant regulatory physical and biogeochemical forcing factors in bacterial GLMs. Given the current state of heterotrophic bacteria in the WAP system, which are under substantial DOC and temperature control, an ongoing trend of climate change along the WAP may increase BP rates as a consequence of a potential increase of DOC from retreating coastal glaciers and an increasing trend of temperature, ice melt, and ocean heat content of UCDW and their subsequent impacts on the upper water-column.

# AUTHOR CONTRIBUTIONS

Substantial contributions to the conception or design of the work (HK, HD); The acquisition, analysis, or interpretation of data for the work (HK, HD): Drafting the work (HK); Revising the work critically (HD); Final approval of the version to be published (HD).

# FUNDING

Palmer LTER program was supported by U.S. National Science Foundation awards OPP 0217282, OPP 0823101, and GEO-PLR 1440435. HK was supported in part by the Department of Earth and Environmental Sciences at Columbia University, and by a subcontract to Lamont-Doherty Earth Observatory from NASA ROSES award NNX14AL86G to Scott Doney (Woods Hole Oceanographic Institution).

# ACKNOWLEDGMENTS

We thank Jeff Bowman and XAG Morán for helpful discussion as well as Palmer LTER field assistants for collection of bacterial samples and sample analyses.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00214/full#supplementary-material

# REFERENCES


during the 1986–87 austral summer. Deep Sea Res. A 38, 1029–1055. doi: 10.1016/0198-0149(91)90095-W


experiments of 14C incorporation. Deep Sea Res. II 49, 769–786. doi: 10.1016/S0967-0645(01)00123-0


Peninsula: I—Sea ice, summer mixed layer, and irradiance. Deep Sea Res. II 55, 2068–2085. doi: 10.1016/j.dsr2.2008.05.021

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Kim and Ducklow. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Seasonal Succession of Free-Living Bacterial Communities in Coastal Waters of the Western Antarctic Peninsula

#### Catherine M. Luria<sup>1</sup> , Linda A. Amaral-Zettler2,3 \*, Hugh W. Ducklow<sup>4</sup> and Jeremy J. Rich<sup>5</sup> \*

<sup>1</sup> Department of Ecology and Evolutionary Biology, Brown University, Providence, RI, USA, <sup>2</sup> The Josephine Bay Paul Center for Comparative Molecular Biology and Evolution, Marine Biological Laboratory, Woods Hole, MA, USA, <sup>3</sup> Department of Earth, Environmental and Planetary Sciences, Brown University, Providence, RI, USA, <sup>4</sup> Department of Earth and Environmental Sciences, Lamont-Doherty Earth Observatory of Columbia University, Palisades, NY, USA, <sup>5</sup> School of Marine Sciences and Darling Marine Center, University of Maine, Walpole, ME, USA

#### Edited by:

Eva Ortega-Retuerta, Spanish National Research Council, Spain

#### Reviewed by:

James T. Hollibaugh, University of Georgia, USA Meinhard Simon, University of Oldenburg, Germany

#### \*Correspondence:

Linda A. Amaral-Zettler amaral@mbl.edu Jeremy J. Rich jeremy.rich@maine.edu

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Microbiology

Received: 22 July 2016 Accepted: 17 October 2016 Published: 03 November 2016

#### Citation:

Luria CM, Amaral-Zettler LA, Ducklow HW and Rich JJ (2016) Seasonal Succession of Free-Living Bacterial Communities in Coastal Waters of the Western Antarctic Peninsula. Front. Microbiol. 7:1731. doi: 10.3389/fmicb.2016.01731 The marine ecosystem along the Western Antarctic Peninsula undergoes a dramatic seasonal transition every spring, from almost total darkness to almost continuous sunlight, resulting in a cascade of environmental changes, including phytoplankton blooms that support a highly productive food web. Despite having important implications for the movement of energy and materials through this ecosystem, little is known about how these changes impact bacterial succession in this region. Using 16S rRNA gene amplicon sequencing, we measured changes in free-living bacterial community composition and richness during a 9-month period that spanned winter to the end of summer. Chlorophyll a concentrations were relatively low until summer when a major phytoplankton bloom occurred, followed 3 weeks later by a high peak in bacterial production. Richness in bacterial communities varied between ∼1,200 and 1,800 observed operational taxonomic units (OTUs) before the major phytoplankton bloom (out of ∼43,000 sequences per sample). During peak bacterial production, OTU richness decreased to ∼700 OTUs. The significant decrease in OTU richness only lasted a few weeks, after which time OTU richness increased again as bacterial production declined toward pre-bloom levels. OTU richness was negatively correlated with bacterial production and chlorophyll a concentrations. Unlike the temporal pattern in OTU richness, community composition changed from winter to spring, prior to onset of the summer phytoplankton bloom. Community composition continued to change during the phytoplankton bloom, with increased relative abundance of several taxa associated with phytoplankton blooms, particularly Polaribacter. Bacterial community composition began to revert toward pre-bloom conditions as bacterial production declined. Overall, our findings clearly demonstrate the temporal relationship between phytoplankton blooms and seasonal succession in bacterial growth and community composition.

**139**

Our study highlights the importance of high-resolution time series sampling, especially during the relatively under-sampled Antarctic winter and spring, which enabled us to discover seasonal changes in bacterial community composition that preceded the summertime phytoplankton bloom.

Keywords: 16S rRNA gene, ecological succession, Antarctica, bacterial production, bacterial community composition, Polaribacter, pelagibacter ubique (SAR11), Rhodobacteraceae

#### INTRODUCTION

fmicb-07-01731 November 1, 2016 Time: 17:3 # 2

Strong seasonal patterns in the marine ecosystem west of the Antarctic Peninsula (WAP) provide a natural experiment to assess how bacteria respond over time to changes in both biotic and abiotic factors. The Antarctic spring brings about a cascade of environmental changes, including light-driven modification of dissolved organic matter (DOM), sea ice melting and retreat, warmer water temperatures and stratification of the water column. These physical changes trigger phytoplankton blooms that support large stocks of upper level consumers (Smetacek and Nicol, 2005) and represent a potentially important sink for atmospheric CO<sup>2</sup> (Arrigo et al., 2008).

In the water column, some bacterial taxa are adapted to colonize and attach to particles through surface adhesion and gliding motility, while other taxa are more adapted toward life as free-living cells (Dang and Lovell, 2016). This is reflected in differences in community composition between the particle attached and free-living communities (DeLong et al., 1993; Ortega-Retuerta et al., 2013; Rieck et al., 2015). Particle attached bacteria play an important role in the initial degradation of particulate matter, hydrolyzing polymers and releasing smaller molecules that can diffuse away from particles and be utilized by free-living bacteria (Stocker, 2012; Dang and Lovell, 2016). While particle attached bacteria may have higher specific activity, freeliving cells generally contribute more to overall bacterial activity in the water column due to greater overall cell abundance, but there are exceptions (Iriberri et al., 1987; Turley and Mackie, 1994; Rieck et al., 2015). Therefore, studies that focus on freeliving cells describe an important part of the bacterial community but not the entire community (Teeling et al., 2012; Williams et al., 2013; Sunagawa et al., 2015).

Bacterial interactions with phytoplankton contribute to ecosystem function in multiple ways in both the WAP and the global ocean (Cole, 1982; Croft et al., 2005; Sher et al., 2011). Sustained primary production is partly dependent upon the microbial loop, in which particle attached and free-living heterotrophic bacteria degrade DOM and are consumed by bacterivores, thereby recycling nutrients (Azam et al., 1983). Studies in marine ecosystems indicate that bacterial growth is frequently dependent on phytoplankton-derived DOM (Church et al., 2000; Morán et al., 2001, 2002; Piquet et al., 2011; Ducklow et al., 2012; Kim et al., 2014). Phytoplankton derived carbon likely influences the succession of bacterial communities as various bacterial taxa differ in their ability to degrade phytoplankton derived DOM and particulate detritus (Kerkhof et al., 1999; Pinhassi et al., 2004; Teeling et al., 2012). Lower bacterial diversity, in terms of both richness and evenness, accompanies seasonal changes, reflecting an increase in abundance of relatively few bacterial taxa (Gilbert et al., 2012; Ladau et al., 2013).

Certain groups of bacteria (e.g., Flavobacteria and Rhodobacteraceae) increase in abundance during phytoplankton blooms, while other groups such as Pelagibacter are better adapted to the free-living state and non-bloom conditions (Williams et al., 2013; Buchan et al., 2014; Voget et al., 2015). Flavobacteria, described as 'first responders' to phytoplankton blooms, break down complex organic matter by direct attachment and exoenzymatic attack of phytoplankton cells and phytoplankton-derived detrital particles (Williams et al., 2013). An abundant genus of Flavobacteria in the bacterioplankton is Polaribacter, which possess traits consistent with a life-strategy of particle attachment and polymer degradation (Fernández-Gómez et al., 2013). However, Polaribacter is metabolically flexible and also abundant in the free-living community, indicating that its ecological niche extends beyond particle attachment (Smith et al., 2013; Williams et al., 2013). Flammeovirgaceae, a family within Bacteroidetes, has been associated with the degradation of algal-derived polysaccharides (Chan et al., 2015; Liu et al., 2015). Members of Rhodobacteraceae are often found in close association with phytoplankton blooms in either the particle attached or free-living part of the community. They generally use small molecular weight substrates, including the degradation products produced by Flavobacteria (Pinhassi et al., 2004; West et al., 2008; Wemheuer et al., 2015).

Studies of bacterial seasonal succession have emphasized the role of bacteria as degraders of labile organic matter (Teeling et al., 2012; Moran, 2015; Needham and Fuhrman, 2016). However, there is increasing evidence that bacterial interactions with phytoplankton may influence the development of phytoplankton blooms themselves through bacterial production of key vitamins, chelating agents, or hormones that stimulate or impede phytoplankton growth (Amin et al., 2012, 2015; Prieto et al., 2015; Wang et al., 2016). Bacterial succession prior to the onset of phytoplankton blooms has been arguably under-studied (Moran, 2015; Needham and Fuhrman, 2016).

Previous analyses in the WAP, based either on community fingerprinting techniques (i.e., denaturing gradient gel electrophoresis) over one or more seasons, or on highthroughput DNA sequencing from only few mid-winter and mid-summer sampling dates, hint at a relationship between bacterial community succession and phytoplankton blooms similar to that observed in more temperate regions (Murray et al., 1998; Murray and Grzymski, 2007; Grzymski et al., 2012; Luria et al., 2014). However, the intervening time period between winter and summer is severely under-sampled in the WAP,

as it is throughout the Southern Ocean, making comparisons to other systems difficult. Our objective was to obtain a new high-resolution seasonal time-series of bacterial properties to determine how free-living bacterial succession proceeds during the dynamic Antarctic winter to summer transition. We hypothesized that a phytoplankton bloom would trigger bacterial succession as has been demonstrated previously in temperate regions (e.g., Teeling et al., 2012). Our findings on bacterial community succession provide new evidence of coupling between phytoplankton and bacterial blooms.

# MATERIALS AND METHODS

#### Field Sampling and Contextual Data

Seawater samples were collected from coastal surface waters at Palmer Station, on the west coast of the Antarctic Peninsula at one to 2 week intervals beginning in the austral mid-winter (July 2013) and ending in late summer (March 2014). Samples were drawn directly from a seawater intake located at a depth of 6 m, 16 m from the station. Triplicate 20-L samples were collected in acid-washed Nalgene carboys and immediately transferred to the laboratory for processing. All processing took place in a 0◦C cold room to maintain initial water temperature.

Each 20-L carboy was sub-sampled for dissolved nutrients (phosphate, silicate, and nitrate), particulate organic carbon and nitrogen (POC and PON), chlorophyll a (chl a), and bacterial abundance and bacterial production measurements. Samples were collected and processed according to Palmer LTER standard protocols<sup>1</sup> . Briefly, nutrient samples were filtered through combusted 0.7-µm glass fiber filters (Whatman, GE Healthcare Life Sciences, Piscataway, NJ, USA) and frozen at −80◦C until analysis on a SEAL AutoAnalyzer 3 (data available at doi:10.6073/pasta/e893d71c5586769731875d49fde21b1d; Ducklow, 2016). POC and PON samples were collected on combusted 0.7-µm glass fiber filters from 1 to 3 L of seawater and were frozen at −80◦C until analysis via combustion using a Perkin Elmer 2400 Series II CHNS/O Analyzer. Chl a samples, as with POC and PON, were collected on 0.7-µm glass fiber filters from 1 to 3 L of seawater and were assayed fluorometrically using acetone extracts (data available at doi: 10.6073/pasta/012de0cf7d1f00951b7289037a3a4c19; Schofield and Vernet, 2016). Bacterial abundance samples were analyzed by flow cytometry following the protocol of Gasol and Del Giorgio (2000), with SYBR <sup>R</sup> Green I nucleic acid staining (Invitrogen, Carlsbad, CA, USA) on an Accuri C6 flow cytometer (BD Biosciences, San Jose, CA) (data available at doi:10.6073/pasta/012de0cf7d1f00951b7289037a3a4c19; Ducklow et al., 2016). Bacterial production rates were derived from rates of <sup>3</sup>H-leucine incorporation (Ducklow et al., 2012).

Samples for bacterial community composition were collected via gentle vacuum filtration (−0.5 bar vacuum) of approximately 2–6 L of seawater through successive 3.0-µm polycarbonate (EMD Millipore, Billerica, MA, USA) and 0.22 µm polyethersulfone (EMD Millipore, Billerica, MA, USA) filters. Filters were flash-frozen with liquid N<sup>2</sup> and stored at −80◦C until further processing.

#### 16S rRNA Gene Library Generation

Bacteria that passed through 3.0µm pore-size filters and were retained on 0.22 µm pore-size filters were analyzed, and therefore our focus is on the free-living bacterial community composition and not the particle attached component. DNA was extracted from cells captured on the 0.22-µm pore size filters using a DNeasy Plant Mini Kit (Qiagen, Valencia, CA, USA) with an additional bead-beating step. Initially, each filter was cut into small pieces and placed into a tube along with 0.1-mm diameter silica beads. After incubation with AP1 and RNase as in the manufacturer's protocol, samples were vigorously vortexed for 1 min to lyse cells. DNA extraction was then completed according to the manufacturer's protocol.

For each DNA sample, the V6 hypervariable region of the bacterial 16S rRNA gene was amplified in two stages following the protocol described previously in Eren et al. (2013). In the first stage, ∼100 bp of the V6 region was amplified in triplicate 33-µl PCR reactions containing 1.0 unit of Platinum Taq Hi-Fidelity Polymerase (Life Technologies, Carlsbad, CA, USA), 1X Hi-Fidelity buffer, 200 µM dNTP PurePeak DNA polymerase mix (Pierce Nucleic Acid Technologies, Milwaukee, WI, USA), 2.0 mM MgCl2, 0.06% BSA, 0.2 µM forward and reverse nonfusion primers (Eren et al., 2013), and approximately 10 ng template DNA. The PCR reaction consisted of a 3-min initial denaturation step at 94◦C, 25 cycles of 94◦C for 30 s, 60◦C for 45 s, and 72◦C for 60 s, and a final 2-min extension at 72◦C. After amplification, triplicate samples were pooled and purified using a MinElute Reaction Cleanup Kit (Qiagen, Valencia, CA, USA) with DNA finally eluted in 10 µl of Qiagen EB buffer.

After quality checks using a Fragment Analyzer (Advanced Analytics, Ames IA), a second fusion PCR reaction was conducted with a set of custom fusion primers consisting of Illumina adaptors, 12 different inline barcodes (forward primers), 8 dedicated indices (reverse primers), and the same V6 primer sequences used in the first round of PCR. The PCR reaction conditions were similar to those described above except that only a single reaction was run for each sample with approximately 2–3 ng purified PCR product as template and 10 reaction cycles. After the final concentration of PCR products was determined using a Qubit 3.0 fluorometer with PicoGreen (LifeTechnologies, Carlsbad, CA, USA), equimolar amounts of each sample were pooled. The 200–240 bp fraction of the sample pool was selected on 1% agarose using a Pippin Prep (SageScience, Beverly, MA, USA) and the final DNA concentration was measured using qPCR (Kapa Biosystems, Woburn, MA, USA) prior to sequencing on one lane of an Illumina HiSeq 1000 cycle paired-end run.

#### Data Analysis

Low-quality sequences were filtered from the resulting data by discarding reads without 100% consensus between forward and reverse paired-end sequencing reads (Eren et al., 2013). Observed taxonomic units (OTUs) were clustered using Qiime's (v 1.9.1) open reference OTU picking with the default UCLUST method, a minimum cluster size of 2, and a 97% similarity threshold and

<sup>1</sup>http://oceaninformatics.ucsd.edu/datazoo/data/pallter/datasets

were assigned Greengenes taxonomy (version 13\_8) (Caporaso et al., 2010; Edgar, 2010; McDonald et al., 2012). After removing OTUs classified as chloroplasts, rarefied libraries were produced by randomly down-sampling to the smallest library size, 43,308 sequences. Subsequent analyses were based on rarefied libraries unless otherwise indicated.

Alpha diversity metrics included the number of OTUs observed in each library, Shannon's diversity, and Pielou's evenness, as well as non-parametric Chao–Jost richness estimates based on un-rarefied libraries as implemented in the R package iNEXT (Chao and Jost, 2012; Chao et al., 2014). The beta diversity between samples was visualized with nonmetric multidimensional scaling (NMDS) based on Bray-Curtis similarity using the metaMDS function in the vegan R package (Oksanen et al., 2016). NMDS was also used to assess several normalization methods for un-rarefied libraries: converting read numbers to relative abundances, as well as Relative Log Expression (RLE) and Trimmed Mean of M-Values (TMM) normalizations as implemented in edgeR (McMurdie and Holmes, 2013).

To identify OTUs with non-random temporal patterns, we first selected OTUs with a relative abundance ≥0.01 in at least three samples. We assessed the seasonality of this subset through local polynomial regression (LOESS) with serial day as the independent variable and relative abundance as the dependent variable, using r <sup>2</sup> > 0.8 as a threshold to identify OTUs with nonrandom temporal patterns. We used these OTUs to generate a co-occurrence network based on Pearson's correlation (r > 0.5) using the CoNet plugin in Cytoscape (Shannon et al., 2003; Faust et al., 2012). We extracted non-overlapping clusters from this network through k-means clustering using the Cytoscape clusterMaker2 plugin after first selecting a reasonable value for k through the evaluation of a scree plot of within-clusters sum of squares (Supplementary Figure S1; Morris et al., 2011). The resulting network clusters were visualized in Cytoscape. This helped us to identify groups of OTUs with similar temporal trends.

In order to model the relationship between environmental drivers and potentially delayed bacterial responses, we used linear interpolation as implemented in the R package zoo to predict values between sampling dates and create regular time series (Zeileis and Grothendieck, 2005). We then performed stepwise linear regression with forward-backward variable selection as implemented in the R packages dynlm and MASS, minimizing Akaike Information Criterion (Venables and Ripley, 2002; Zeileis, 2016). Regression models were tested for multicollinearity among predictor variables and models with variance inflation factors (VIF) greater than 5 discarded. The variables tested included temperature, inorganic nutrient levels, POC, and PON. To account for delayed bacterial responses, we built distributed lag models, testing 0-, 10-, and 20-day time lags (Jorgenson, 1966). Additional details are given in Supplementary Text S1 and Supplementary Table S1.

The R code used for all data analysis and figure production can be accessed at: https://github.com/cmluria/Palmer-Station-Bacterial-Succession. All of our sequence data are MIMARKScompliant (Yilmaz et al., 2011) and have been deposited in the NCBI SRA under the accession number SRP091049. Associated MIMARKS-compliant metadata appear in Supplementary Table S2.

# RESULTS

Our sampling period encompassed changes in day length from ∼4 h of sunlight when we began sampling in July to ∼22 h of sunlight in December. Water temperatures reached a minimum of −1.1◦C in August and a maximum of 1.6◦C in January (Supplementary Figure S2). Sea ice cover in the Palmer region is variable from year to year; 2013 was a heavy sea ice year with dense sea ice cover during the winter that persisted at significant levels until December (Stammerjohn et al., 2008; Massom et al., 2014). Chl a, a proxy for phytoplankton biomass was low (<0.6 µg l−<sup>1</sup> ) until October when a brief increase in chl a occurred. The primary phytoplankton bloom peaked in summer (early January), based on chl a, and then declined towards pre-bloom levels in March (**Figure 1A**). Dissolved inorganic nutrients (phosphate, silicate, and nitrate) were drawn down during the summer bloom, while particulate carbon and nitrogen increased, reflecting phytoplankton production (Supplementary Figure S2). Silicate drawdown was less than nitrate during the summer bloom, suggesting that other phytoplankton groups in addition to diatoms contributed to the bloom. Bacterial production was very low but still detectable from July to October (**Figure 1B**). By January bacterial production increased substantially, peaking about 3 weeks after the peak in chl a (**Figures 1A,B**). Our bacterial production rates measured from the seawater intake were similar to those measured at a near shore station (Supplementary Figure S3; data available at doi: 10.6073/pasta/814628d18d4e23753d1164f0bd81095e). Stepwise linear regression revealed that chl a with time lags of 0, 10, and 20 days was a good predictor of both bacterial production (r <sup>2</sup> = 0.93) and abundance (r <sup>2</sup> = 0.62; Supplementary Figure S4; Supplementary Table S1). Several other individual factors (phosphate, silicate, nitrate, POC, and PON) were also significantly correlated with bacterial production (Supplementary Table S1).

In total, 68 samples, spread across 24 sampling dates, were sequenced, yielding 15 million short-read V6 16S rRNA gene sequences (∼43,000–550,000 per library) corresponding to the free-living bacterial community, assigned to 28,857 OTUs. Given recent discussion regarding the statistical validity of library resampling (McMurdie and Holmes, 2014), we assessed different methods to characterize bacterial richness and community composition using un-rarefied sequence libraries. Although the method described by Chao et al. (2014) is intended to estimate true richness based on coverage regardless of sequencing depth, both observed (p < 0.001, r <sup>2</sup> = 0.4) and estimated richness (p < 0.001, r <sup>2</sup> = 0.35) significantly correlated with library size (Supplementary Figure S5). This may indicate that the species abundance distributions vary widely among our datasets (Gwinn et al., 2016). Likewise, NMDS based on un-rarefied sequence libraries showed the influence of initial library size even after relative abundance, RLE, and TMM normalizations

(Supplementary Figure S6). For these reasons, all subsequent analyses relied on libraries rarefied to the smallest library size of 43,308 sequences.

Observed richness of the free-living bacterial community varied significantly between midwinter and midsummer (p < 0.0001), ranging from a maximum of approximately 1800 OTUs in July to a minimum of approximately 700 OTUs in January (**Figure 1**). On average, Chao–Jost estimated richness was four times greater than observed richness. Richness was negatively correlated with chl a (p < 0.001, r <sup>2</sup> = 0.16) and bacterial production (p < 0.0001, r <sup>2</sup> = 0.55). A Tukey's HSD comparison across months confirmed that the decline in richness was significant in January and February (p < 0.05). Shannon Diversity and Pielou's evenness were also significantly lower in January (p < 0.0001).

**Figure 2** provides a broad overview of the changes in relative abundance of free-living bacterial taxa that occurred over the course of the season. Oceanospirillales and Pelagibacteraceae had the greatest relative abundance in winter and spring (July– November). SAR406, Rhodospirillales, and Deltaproteobacteria, including Nitrospina and SAR324, were present at greater relative abundances in winter and early spring (July–September) than later in the field season. Changes in the relative abundance of taxonomic groups became most evident in January when Rhodobacteraceae had greater relative abundance than Pelagibacteraceae. Polaribacter also increased significantly in January. We did not account for variation in 16S rRNA copy number and thus may have underestimated the abundance of lower copy number taxa like Pelagibacteraceae and overestimated the abundance of higher copy number taxa like Colwelliaceae. In addition to changes in relative abundance for broad taxonomic groups, co-occurrence network analysis with k-mean clustering revealed five temporal pattern types, with individual OTUs that peaked (i) during the winter or early spring, (ii) just prior to the summer bloom, (iii) very early in the summer bloom, (iv) at summer mid-bloom, or (v) after the summer bloom subsided (**Figures 3** and **4**). Some OTUs that were assigned the same family level classification (e.g., Rhodobacteraceae) had distinct temporal patterns corresponding to different times during the summer bloom.

NMDS of relative abundances of OTUs demonstrated a shift in community composition from July through December before the onset of the summer phytoplankton bloom (**Figure 5**). Community composition continued to change from December to January as the summer phytoplankton bloom developed (**Figure 4**). March samples clustered midway between the July and November and late January-early February samples, perhaps reflecting the beginning of a return to a pre-bloom community structure. Bacterial communities fell into five groups based on hierarchical clustering of Bray–Curtis similarities. These groups corresponded to different time periods and ranges of bacterial production: July to November (winter and spring; <5 pM l−<sup>1</sup> h −1 ), December and March (before and after the summer bloom; 5–21 pM l−<sup>1</sup> h −1 ), early January (beginning of the summer bloom; 26–44 pM l−<sup>1</sup> h −1 ), late January to early February (summer mid-bloom; 50-129 pM l−<sup>1</sup> h −1 ), and late February (late in the summer bloom; 17–43 pM l−<sup>1</sup> h −1 ).

#### DISCUSSION

We observed a significant decline in free-living bacterial OTU richness in summer as bacterial production increased during a phytoplankton bloom. Previous studies of temperate systems have also reported richness maxima in the winter and minima in the summer (Gilbert et al., 2012; Ladau et al., 2013). Similar seasonal trends in bacterial OTU richness have been

reported in Antarctic waters (Murray and Grzymski, 2007; Ghiglione and Murray, 2012; Grzymski et al., 2012; Luria et al., 2014). Gilbert et al. (2012) observed relatively gradual changes, reporting that day length and serial day alone explained over 66% of the observed variance in richness during a 6-year time series study of the English Channel. Likewise, during a 6-year study in the Northwest Atlantic Ocean, El-Swais et al. (2015) observed a much longer period of decreased richness during the summer and spring that accompanied an extended period of increased primary production. In contrast, we observed only a brief period (∼4 weeks) of decreased richness that coincided with increased bacterial production rates. The abbreviated nature of phytoplankton blooms in the WAP along with high-resolution temporal sampling enabled us to observe this pattern, which contrasts with the multiple annual blooms and/or long periods of sustained growth seen in temperate systems. The results of our study demonstrate close temporal linkages between episodic but biologically significant events like phytoplankton blooms and changes in bacterial production and richness.

In addition to changes in alpha diversity, we measured temporal variation in the taxonomic composition of freeliving bacterial communities. The most abundant winter bacterial taxonomic group in our amplicon sequence data was Pelagibacteraceae, a globally abundant clade that correlates negatively with primary production (Morris et al., 2002; Sowell et al., 2009). In addition, several bacterial taxa involved in alternative energetic pathways, e.g., SAR406 (sulfur oxidation; Wright et al., 2013), Nitrospina (nitrite oxidation; Spieck et al., 2014), and SAR324 (sulfur oxidation and carbon fixation; Sheik et al., 2014) dominated early in the winter, in accordance with previous suggestions that chemolithoautotrophy is an important metabolism contributing to cellular production during the winter (Manganelli et al., 2009; Grzymski et al., 2012; Williams et al., 2012). Although we did not sequence archaea in this study, ammonia-oxidizing Thaumarchaeota contribute significantly to chemolithoautotrophy during winter, and overall ammonia oxidation rates are higher in winter than summer (Tolar et al., 2016). The seasonality in Nitrospina may be related to changes in ammonia oxidation rates.

While species richness did not change significantly until the phytoplankton bloom in summer, free-living bacterial community composition began to change as early as October, with declining relative abundances of Pelagibacteraceae, SAR406, SAR324, and Nitrospina. As the season progressed, increasing competition or top down control may have caused declines in the relative abundances of these taxa, while increasing water temperatures and therefore increased water column stratification would have prevented replenishment of taxa from deeper waters. The increase in relative abundance of taxa like Flavobacteriaceae, Polaribacter, Flammeovirgaceae, and Rhodobacteraceae before the summer phytoplankton bloom is somewhat surprising given the low levels of chl a and bacterial production during the time period. A brief increase in chl a in late October coincided with a decline in richness and an increase in bacterial production. While richness increased when chl a subsided, bacterial production remained slightly elevated over the next 2 months. These pre-summer changes in community composition could reflect a direct response by

certain bacterial taxa to increasing sunlight, in addition to lowlevel and ephemeral increases in phytoplankton. For example, some marine Flavobacteria, including Polaribacter, contain proteorhodopsin, a light-driven proton pump that can enhance growth in low-nutrient conditions, while Rhodobactaceae have been shown to carry out aerobic anoxygenic photosynthesis (Gonzalez et al., 2008; Kimura et al., 2011; Voget et al., 2015; Xing et al., 2015). Whether or not pre-summer changes

in bacterial community composition are associated with bacterial feedbacks on phytoplankton growth is not known (Amin et al., 2012, 2015; Prieto et al., 2015; Wang et al., 2016).

As the summer phytoplankton bloom developed, we observed changes in community composition that corresponded to changes in bacterial production. The relative abundance of free-living Polaribacter doubled while that of Rhodobacteraceae tripled during the phytoplankton bloom. We noted that different OTUs within Flavobacteriaceae, including Polaribacter, and Rhodobacteraceae peaked at different points during the phytoplankton bloom. This suggests niche specialization by different taxa within these groups, perhaps corresponding to different stages in the successive degradation of phytoplanktonderived organic compounds (Teeling et al., 2012; Klindworth et al., 2014). Polaribacter occurred in a cluster with only one other OTU (**Figures 3** and **4**), suggesting a strong competitive advantage for Polaribacter early in the bloom, in support of the hypothesized role of this taxon (Williams et al., 2013). OTUs of Rhodobacteraceae that peaked later in the bloom may indicate that they utilize secondary products of decomposing organic matter, such as low molecular weight compounds (Voget et al., 2015). We observed a mid-bloom increase in the relative abundance of Gammaproteobacteria, including Shewanella and Colwelliaceae, taxa previously associated with substrate rich environments (Baelum et al., 2012; Delmont et al., 2014).

Early February peaks in the abundance of OTUs classified as Rhodobacteraceae and Colwelliaceae corresponded to almost complete depletion of nitrate and nitrite, an unusual event in the WAP that might have caused phytoplankton to release carbohydrates (Carlson and Hansell, 2014). It is important to note that our focus on the <3.0 µm fraction of the bacterial community excluded particle-associated bacteria, which may have played an increasingly important role as the phytoplankton bloom developed (Riemann et al., 2000).

We used distributed lag models to statistically test the dynamic relationship between phytoplankton abundance based on chl a and bacterial production and abundance during this period. The inclusion of 10- and 20-day lags in our model allowed us to capture a greater portion of the variability in these relationships. In January, peak bacterial production occurred about 20 days after the peak in chl a, in agreement with previous studies of Antarctic phytoplankton bloom dynamics (Billen and Becquevort, 1991; Ducklow et al., 2001). A secondary chl a peak in mid-January would have presumably also contributed to the peak in bacterial production that occurred about 10 days later. Bacterial reliance on high molecular weight phytoplankton-derived organic matter that must undergo extracellular hydrolysis prior to bacterial uptake is thought to drive such temporal lags (Billen and Becquevort, 1991; Lancelot et al., 1991; Ducklow et al., 2001; Kirchman et al., 2001). Furthermore, some DOM fractions may become available through intermediate trophic processes (e.g., zooplankton sloppy feeding or excretion; Ducklow et al., 2012). The significance of both 10- and 20-day lags may reflect the complex nature of WAP DOM degradation, in which individual bacterial taxa respond within different time frames and specialize in different fractions of the DOM pool (Nikrad et al., 2014; Bowman and Ducklow, 2015; Kim and Ducklow, 2016).

We hypothesize that resource supply from phytoplanktonderived carbon was the primary factor driving the summertime shift in free-living bacterial community composition, although biotic interactions and top–down control by grazing and viral lysis likely modulate the patterns that we observed (Bird and Karl, 1999; Brum et al., 2016). Our measurements did not directly address the underlying mechanisms controlling temporal patterns in bacterial community composition. In particular, we did not directly measure the composition or turnover of organic compounds that we hypothesize contributed to changing bacterial communities. Our study and other similar studies lack a multi-year dimension to be definitive. Nevertheless, the general pattern that we observed is consistent with the view that increased resource supply of a diverse pool of organic compounds that results from phytoplankton blooms plays a strong role in structuring bacterial communities (Teeling et al., 2012; Buchan et al., 2014).

Taken together, our bacterial production, richness, and community composition data indicate strong trophic coupling between bacteria and phytoplankton. Our results agree with previous reports that WAP bacterial density and production are positively correlated with primary production (Church et al., 2003; Ortega-Retuerta et al., 2008), but contrast with earlier reports that the microbial loop is uncoupled from primary producers during spring phytoplankton blooms (Bird and Karl, 1991, 1999; Karl et al., 1991; Duarte et al., 2005). Although community shifts like those we observed suggest strong potential for degradation of labile phytoplankton-derived organic matter (Landa et al., 2014), WAP bacterial standing stocks and productivity are low relative to primary production (∼5% versus 10–20% global average; Ducklow et al., 2012; Kim and Ducklow, 2016). Despite the longstanding debate about the role of temperature in limiting bacterial production in cold water, recent analyses found that temperature is not the principal factor regulating bacterial growth in the Antarctic (Thingstad et al., 1991; Ducklow et al., 2012). However, there is growing evidence, including results from this study, that organic matter availability is a primary factor controlling bacterial growth and community composition in Antarctica (Kirchman et al., 2009; Ducklow et al., 2012; Kim and Ducklow, 2016).

# AUTHOR CONTRIBUTIONS

fmicb-07-01731 November 1, 2016 Time: 17:3 # 10

CL, LA-Z, HD, and JR designed and conducted the study. LA-Z contributed sequence data for the study and input on analyses and manuscript content. CL analyzed data and wrote the manuscript with guidance from JR. LA-Z, HD, and JR provided input and edited the manuscript. All authors approved the final version of the manuscript.

#### FUNDING

CL was partially funded by the Graduate School and the Department of Ecology and Evolutionary Biology at Brown

#### REFERENCES


University and the Brown University-Marine Biological Laboratory Joint Graduate Program. This material is based upon work supported by the National Science Foundation under Grant Nos. ANT-1142114 to LA-Z, OPP-0823101 and PLR-1440435 to HD, and ANT-1141993 to JR.

#### ACKNOWLEDGMENTS

We acknowledge the help of the following people for providing field and laboratory assistance at Palmer Station as part of this project: Jamie Collins, Sharon Grim, Sean O'Neill, Monica Stegman, Sebastian Vivancos, and Madelyne Willis, as well as the staff at Palmer Station. We thank Annaliese Jones, Leslie Murphy, and Naomi Shelton for conducting laboratory analyses. We thank Casey Dunn for his support. We thank the two reviewers for helping improve the manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01731/full#supplementary-material




Antarctica, During Palmer LTER Field Seasons, 1991–2015. New Brunswick, NJ: Rutgers University.



and minimum information about any (x) sequence (MIxS) specifications. Nat. Biotechnol. 29, 415–420. doi: 10.1038/nbt.1823


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Luria, Amaral-Zettler, Ducklow and Rich. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Changes in Marine Prokaryote Composition with Season and Depth Over an Arctic Polar Year

Bryan Wilson<sup>1</sup> \*, Oliver Müller <sup>1</sup> , Eva-Lena Nordmann<sup>1</sup> , Lena Seuthe<sup>2</sup> , Gunnar Bratbak <sup>1</sup> and Lise Øvreås <sup>1</sup>

<sup>1</sup> Marine Microbiology Research Group, University of Bergen, Bergen, Norway, <sup>2</sup> Department of Arctic and Marine Biology, UiT - The Arctic University of Norway, Tromsø, Norway

As the global climate changes, the higher latitudes are seen to be warming significantly faster. It is likely that the Arctic biome will experience considerable shifts in ice melt season length, leading to changes in photoirradiance and in the freshwater inputs to the marine environment. The exchange of nutrients between Arctic surface and deep waters and their cycling throughout the water column is driven by seasonal change. The impacts, however, of the current global climate transition period on the biodiversity of the Arctic Ocean and its activity are not yet known. To determine seasonal variation in the microbial communities in the deep water column, samples were collected from a profile (1-1000 m depth) in the waters around the Svalbard archipelago throughout an annual cycle encompassing both the polar night and day. High-throughput sequencing of 16S rRNA gene amplicons was used to monitor prokaryote diversity. In epipelagic surface waters (<200 m depth), seasonal diversity varied significantly, with light and the corresponding annual phytoplankton bloom pattern being the primary drivers of change during the late spring and summer months. In the permanently dark mesopelagic ocean depths (>200 m), seasonality subsequently had much less effect on community composition. In summer, phytoplankton-associated Gammaproteobacteria and Flavobacteriia dominated surface waters, whilst in low light conditions (surface waters in winter months and deeper waters all year round), the Thaumarchaeota and Chloroflexi-type SAR202 predominated. Alpha-diversity generally increased in epipelagic waters as seasonal light availability decreased; OTU richness also consistently increased down through the water column, with the deepest darkest waters containing the greatest diversity. Beta-diversity analyses confirmed that seasonality and depth also primarily drove community composition. The relative abundance of the eleven predominant taxa showed significant changes in surface waters in summer months and varied with season depending on the phytoplankton bloom stage; corresponding populations in deeper waters however, remained relatively unchanged. Given the significance of the annual phytoplankton bloom pattern on prokaryote diversity in Arctic waters, any changes to bloom dynamics resulting from accelerated global warming will likely have major impacts on surface marine microbial communities, those impacts inevitably trickling down into deeper waters.

Keywords: Arctic, marine microbiology, seasonality, nutrient cycling, depth, climate change, phytoplankton bloom, mesopelagic

Edited by: Connie Lovejoy, Laval University, Canada

#### Reviewed by:

Nico Salmaso, Fondazione Edmund Mach, Italy Dolors Vaque, Consejo Superior de Investigaciones Científicas, Spain

> \*Correspondence: Bryan Wilson bryan.wilson@uib.no

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Marine Science

Received: 01 October 2016 Accepted: 20 March 2017 Published: 13 April 2017

#### Citation:

Wilson B, Müller O, Nordmann E-L, Seuthe L, Bratbak G and Øvreås L (2017) Changes in Marine Prokaryote Composition with Season and Depth Over an Arctic Polar Year. Front. Mar. Sci. 4:95. doi: 10.3389/fmars.2017.00095

# 1. INTRODUCTION

Polar regions are vulnerable and most sensitive to global climate change. Therefore, there is an increasing research focus needed on these high latitude environments. During the annual cycle, the poles undergo some of the most extreme environmental changes on the planet, from that of the subzero permanently dark winter to the relative warmth and perpetual daylight of the summer. The extensive biodiversity of these regions is understandably well-adapted to these periodic shifts but accelerated atmospheric warming is irrefutably altering conditions here and therefore needs to be studied in detail. The Arctic is warming three times faster than the global mean warming rate (Trenberth et al., 2007) and the extent and thickness of sea ice in the polar oceans is steadily decreasing (Chen et al., 2009), at a rate of up to ten percent per year (Comiso et al., 2008). Recent decades have also seen the summer melt season increase in length (Markus et al., 2009) and the percentage of thin first-year ice increase as compared with thicker multiyear ice (MYI) (Comiso, 2012), such that the late summer Arctic Ocean may be ice-free before the end of the twenty-first century (Boe et al., 2009) or sooner still (Kerr, 2012). Some of these striking environmental changes occurring in the Arctic are related to the inflow of Atlantic water to the Arctic Ocean. The West Spitsbergen Current passing through the eastern Fram Strait is the most significant inflow to the Arctic Ocean and it has intensified over the last decades (Schauer et al., 2004). Increased inflow of the warm and highly saline Atlantic Water affects water column stability and is also probably one of the main drivers of the recent sea ice loss north of Svalbard (Onarheim et al., 2014; Randelhoff et al., 2015). The cumulative consequence of these effects has been the exposure of polar seas to increasing levels of solar radiation (Perovich et al., 2007). Furthermore, enhanced permafrost thawing (Romanovsky et al., 2010) in concert with the profound influence of several large river systems (an hydrology peculiar to the Arctic) (Anderson, 2002) and a greater erosion of exposed coastlines (Lantuit et al., 2012), is leading to an increased terrigenous input of carbon to the Arctic Ocean (Frey and McClelland, 2009). All of which has the potential to radically impact the primary production and successive trophic levels of the polar marine environment (Anderson and Macdonald, 2015).

The greater part of high latitude research has been carried out in the Arctic (predominantly due to the relative logistical ease of working in the region when compared with Antarctica) and the Svalbard archipelago in particular has become a key site for Arctic marine studies, particularly with regards to its being the confluence of both the Arctic and Atlantic Oceans (Hop et al., 2002, 2006; Svendsen et al., 2002). The ocean around the Western coastline of Svalbard is a sea iceassociated pelagic ecosystem (Svendsen et al., 2002) and as with other high latitude locations, seasonal variations in light (and thus in primary production) are more pronounced here than elsewhere. The extreme seasonality of these environmental drivers has revealed a number of trends particular to the polar regions. A single major spring bloom along the retreating ice edge accounts for >50% of the annual primary production around Svalbard and in the Northern Barents Sea (Sakshaug, 2004). By late summer, this develops into a successional postbloom stage, comprising different phytoplankton populations (Sherr et al., 2003) and it follows, their different associated successional heterotrophic prokaryote consortia, in particular the Flavobacteriia and Gammaproteobacteria classes (Alonso-Sáez et al., 2008; Teeling et al., 2012). This phenomenon of the polar phytoplankton blooms (Williams et al., 2013) followed by the heterotrophic bacterial populations also seemingly drives the annual disappearance of the chemolithoautotrophic marine Archaea from surface waters (Kalanetra et al., 2009; Alonso-Sáez et al., 2012; Pedneault et al., 2014) and the subsequent seasonal fluctuations in prokaryote diversity (Murray et al., 1998; Ghiglione and Murray, 2012; Grzymski et al., 2012; Ladau et al., 2013).

A significant fraction of the phytoplankton primary production sinks out of these surface waters (Reigstad et al., 2008), contributing a single annual major input of organic carbon and energy to the microbial communities residing in dark mesopelagic and deep waters. This subsurface realm dominates the global ocean biome and whereas the Arctic Ocean is the shallowest of the five major oceanic divisions, still its average depth is >1000 m deep. These aphotic zones are characterized by higher pressures, lower temperatures and higher inorganic nutrient concentrations than the photic surface waters above (Arístegui et al., 2009; Orcutt et al., 2011). Yet it is these physicochemical factors, in addition to their remoteness from the surface wind effects and solar irradiation that affect the upper layers so, that also determines their characteristic stability (Orcutt et al., 2011). The waters at these depths contain the largest pool of microbes in aquatic systems (Whitman et al., 1998) and play a major role in ocean biogeochemistry, comprising extraordinarily high genetic and metabolic diversity (Arístegui et al., 2009). The marine snow (primarily dissolved and particulate organic matter) produced by the spring and summer phytoplankton blooms in the stratified epipelagic zone is transported down during winter into the mesopelagic zone by convective mixing and subduction after cooling of the sea surface (Arístegui et al., 2009; Grzymski et al., 2012). Chemolithoautotrophic processes (such as archaeal ammonia oxidation) then come into play during the dark winter months (Grzymski et al., 2012) and the resulting nitrate buildup fuels the subsequent phytoplankton spring bloom (Connelly et al., 2014). However, should suggested models of a freshening Arctic be correct (Comeau et al., 2011), surface Arctic basin waters in a warming world may become increasingly stratified, such that the vertical flux of nutrients between deeper waters and the epipelagic zone may be reduced; primary productivity would consequently be lessened and this annual biogeochemical cycle, so essential for Arctic Ocean productivity, would inevitably be disrupted (Tremblay et al., 2008).

As the majority of Arctic studies of marine microbial communities have either been carried out in the more amenable spring and summer seasons or in shallow waters, the primary objective of the present study was to expand upon these data. More specifically, we wished to identify the key mediators of the prokaryotic microbial community in the Atlantic water inflow to the Arctic Ocean during the light-driven summers vs. the dark winter night, seasons characterized by the massive variation in availability of fresh photosynthesis-derived carbon. Additionally, we intended to compare the cold, deep and dark mesopelagic ocean with the cold, shallow and dark surface waters above, to gain more insight into the driving mechanisms resulting from such environmental conditions. High-throughput sequencing technologies have previously highlighted the extreme microbial seasonality of the polar regions (Kirchman et al., 2010; Christman et al., 2011; Connelly et al., 2014). In this study we implement the same technologies to sequence reverse-transcribed total RNA (with its significantly shorter life span than DNA) to provide a timely snapshot of the more metabolically-active fraction of the marine microbial community.

### 2. MATERIALS AND EXPERIMENTAL METHODS

#### 2.1. Sampling

Samples were taken from various transects bisecting the West Spitsbergen Current along the coast of Svalbard, a Norwegian archipelago in the Arctic Ocean (**Figure 1**) from the research vessels RV Lance and RV Helmer Hanssen, operating under either the MicroPolar and Carbon Bridge projects (**Table 1**). Samples (25–50 L) representative of the water column profile were collected from a range of water masses (defined in Paulsen et al., 2016) between 1m and 1000 m (**Table 1**) using Niskin bottles mounted on a rosette deployed from the vessels. Water samples were filtered through 0.22µm Sterivex Filter Units (Merck-Millipore, MA, USA) via a peristaltic pump and frozen at −80◦C immediately.

#### 2.2. Chlorophyll a Measurement

The concentration of chlorophyll (chl) a was determined fluorometrically (Parson et al., 1984). Sample water was filtered onto triplicate Whatman GF/F glass-fiber filters (Sigma-Aldrich, MO, USA). The chl a on the filters was immediately extracted in 5 mL methanol (>99.8% v/v) at 4◦C in the dark for 12 h without grinding. The fluorescence of the extracts was measured with a fluorometer (Model 10-AU, Turner Designs, CA, USA), calibrated with pure chl a (Sigma-Aldrich).

#### 2.3. RNA Extraction

RNA were extracted directly from filters with the AllPrep DNA/RNA Mini Kit (Qiagen, CA, USA) using a protocol modified from the manufacturer's instructions. Briefly, filters were thawed on ice and an extraction buffer (comprising 990µL Buffer RLT Plus and 10µL β-mercaptoethanol per filter) prepared. Extraction buffer (1 mL) was added to each filter and filters vortexed vertically for 2 min, inverted and vortexed for a further 2 min. Lysate was removed from the filter using a 10 mL syringe and transferred to a 1.5 mL microfuge tube. Lysate (700µL) was loaded on to an AllPrep DNA spin column and centrifuged for 30 s at 8000 × g, saving the flow-through for subsequent RNA extraction. Centrifugation steps were repeated for any remaining lysate volume, as necessary.

One volume 70% (v/v) ethanol was added to the flow-through, mixed by pipetting and 700µL transferred to an RNeasy spin column in a 2 mL microfuge tube, before centrifugation for 15 s at 8000 × g. Centrifugation was repeated for the remaining liquid volume as necessary, discarding the flow-through each time. Buffer RW1 (700µL) was added to the RNeasy spin column and the tube centrifuged again for 15 s at 8000 × g and the flow-through discarded. Buffer RPE (500µL) was added to the RNeasy spin column and the tube centrifuged for 15 s at 8000 × g, discarding the flow-through. Buffer RPE (500µL) was again added to the RNeasy spin column and the tube centrifuged for 2 min at 8000 × g. The RNeasy spin column was placed in a fresh 2 mL microfuge tube and centrifuged for 1 min at full speed. The RNeasy spin column was placed in a fresh 1.5 mL microfuge tube and 50µL RNase-free water added to the spin column, before centrifugation for 1 min at 8000 × g. RNasefree water (50µL) was added to the spin column again and the centrifugation repeated. The eluate (100µL) was stored at −20◦C.

The quantity and quality of RNA were assessed using a Qubit 3.0 Fluorometer (Thermo Fisher Scientific Inc., MA, USA) and by agarose gel electrophoresis.

#### 2.4. Reverse Transcription and PCR for High-Throughput Sequencing (HTS) Library Construction

RNA (10 ng) was treated with the DNA-free DNA Removal kit (Invitrogen, CA, USA), prior to reverse transcription using the SuperScript III First-Strand Synthesis System for RT-PCR (Invitrogen), as per the manufacturer's instructions. The V4 region of the 16S rRNA gene was amplified from cDNA using a two-step nested PCR approach. In the first step, triplicate samples were amplified using primers 519F (5′ - CAGCMGCCGCGGTAA-3′ ) (Øvreås et al., 1997) and 806R (5 ′ -GGACTACHVGGGTWTCTAAT-3′ ) (Caporaso et al., 2011). The reaction mixture consisted of 10µL HotStarTaq Master Mix (Qiagen), 500 nM of each primer, 10 ng of cDNA and nucleasefree water to bring the total volume to 20µL. Reactions were initially denatured for 15 min at 95◦C, followed by 25 cycles of denaturation at 95◦C for 20 s, primer annealing at 55◦C for 30 s and extension at 72◦C for 30 s, followed by a final extension step of 72◦C for 7 min. Triplicate amplicons were pooled and purified using the DNA Clean & Concentrator-5 kit (Zymo Research Corporation, CA, USA), as per the manufacturer's instructions, and quantified using a Qubit 3.0 Fluorometer.

In the second step, pooled amplicons were amplified by nested PCR using MID-tagged primers 519F and 806R in a reaction mixture comprising 25µL HotStarTaq Master Mix, 500 nM of each primer, 50 ng of pooled DNA, and nucleasefree water to bring the total volume to 50µL. Reactions were initially denatured for 15 min at 95◦C, followed by 15 cycles of denaturation at 95◦C for 20 s, primer annealing at 62◦C for 30 s and extension at 72◦C for 30 s. This was followed by a final extension step of 72◦C for 7 min. The quantity and quality of RNA was assessed by agarose gel electrophoresis. Amplicons were purified using Agencourt AMPure XP Beads (Beckman Coulter Inc., CA, USA) and quantified again using a Qubit 3.0 Fluorometer and by agarose gel electrophoresis. MID-tagged

FIGURE 1 | Map showing sampling locations and sea ice extent around the Svalbard Archipelago. Maps were generated using the Global Self-consistent Hierarchical High-resolution Geography (GSHHG) data from the National Geospatial-Intelligence Agency (NGA), distributed under the GNU Lesser General Public License (LGPL). Ice data were provided by the Norwegian Ice Service (MET Norway) for the following dates; January 10, March 7, May 23, August 12, and November 7, 2014.

amplicons were then pooled in equimolar amounts for library construction.

Libraries were sent to the Norwegian Sequencing Centre (Oslo, Norway) for HTS on a MiSeq platform (Illumina, CA, USA) using the MiSeq Reagent Kit v2 (Illumina).

### 2.5. Bioinformatic Analyses

16S rRNA gene sequences were processed using a custom BioPython (Cock et al., 2009) script incorporating various bioinformatic tools, along with the QIIME pipeline (Version 1.8.0) (Caporaso et al., 2010b). Briefly, FASTQ files were quality end-trimmed at a phred quality score ≥20 using Sickle (Version 1.33) (Joshi and Fass, 2011) and PhiX contaminants and adapters removed using Bowtie 2 (Langmead and Salzberg, 2012) and cutadapt (Martin, 2011), respectively. Paired-end reads were merged using PANDAseq (Masella et al., 2012) and all reads <200 bp removed. The remaining sequence reads were checked for chimeras with the identify chimeric seqs and filter fasta scripts in QIIME, using usearch61 (Edgar, 2010) and the ChimeraSlayer (Haas et al., 2011) reference database (Gold.fa) found in the Broad Microbiome Utilities suite (http://microbiomeutil. sourceforge.net/). The pick de novo otus script in QIIME (using default parameters) was used for de novo OTU picking (using uclust (Edgar, 2010) and a sequence similarity threshold of 97%), taxonomy assignment (using PyNAST, Caporaso et al., 2010a) at 90% sequence similarity against the Greengenes core reference alignment database (Release 13\_8) (DeSantis et al., 2006), and finally, the assembly of a table of OTU abundances with taxonomic identifiers for each OTU. OTUs were grouped by different taxonomic levels using the summarize taxa and plot taxa summary QIIME scripts. Rarefaction curves and Chao1 values were calculated using QIIME's alpha rarefaction script, whilst principal coordinate analysis plots used the beta diversity through plots script. All statistical figures were produced using the R Software Environment (R Core Team, 2013). High-throughput sequencing data were submitted to the European Nucleotide Archive (ENA) under Accession Number PRJEB19605.

# 3. RESULTS

#### 3.1. Environmental Data

The dataset comprised fifty two samples in total, with the numbers of samples taken on each cruise (**Table 1**) varying with prevailing conditions and ship access time. Summer and winter conditions in the Arctic Ocean around Svalbard are very different (**Figure 1**) and much of the archipelago is ice-bound during the year, with the dark polar night persisting for almost three months. The study area (**Table 1**) is hydrographically characterized by Atlantic water masses, either as pure Atlantic water [AW; T (temperature) > 2◦C and S (salinity) > 34.92; (Walczowski, 2013) and references therein], or as modified colder water masses, such as cold Atlantic Water (cAW) with 0 < T < 2◦C (S > 34.9) and Intermediate Water (IW) T < 0◦C (S > 3 4.9) (de Steur et al., 2014). Arctic water (ArW) was found at some stations and depths. Not all water classified here as ArW necessarily originated from the central Arctic Ocean. This may instead have been water that had undergone freshening and cooling processes, and hence had a density <sup>ρ</sup><sup>θ</sup> > 27.7 kg m−<sup>3</sup> and S <34.92 (or 34.9 when cooler than 2◦C). These are also the physical characteristics for ArW in the central Arctic Ocean. Cold surface water (SW; <sup>ρ</sup><sup>θ</sup> > 27.7 kg m−<sup>3</sup> and S < 34.92) was encountered at stations within the marginal ice zone, created by sea ice melt. Sea ice extended furthest North during the winter months, and furthest South during May and August, prohibiting sampling north of Svalbard during the summer months. Consequently, most stations sampled in May and August were covered by ice, while all stations sampled during March and November were situated in open waters.

Concentration of chl a (a proxy for phytoplankton biomass) showed a clear seasonal cycle in phytoplankton, with lower concentrations throughout the water column during the dark winter months and March, higher concentrations in surface waters in May, and intermediate concentrations in August (**Table 1**). Detailed investigations of the phytoplankton showed a major shift from communities dominated by smaller-celled zooplankton during winter and March, to spring bloom communities dominated by large-celled phytoplankton, such as diatoms and Phaeocystis colonies in May (M. Reigstad, Personal Communication). In August, the phytoplankton communities were diverse and dominated again by smaller cells. The community shift from May to August reflects the general shift from nitrate-based phytoplankton communities in May to phytoplankton communities based on regenerated production in August (M. Reigstad, Personal Communication). In the deep mesopelagic waters, there was much less variation in biological and chemical properties than surface waters over the course of the year.

# 3.2. Prokaryote Diversity

After quality trimming and chimera removal, the complete 16S rRNA dataset (targeting the V4 region) comprised 7 902 016 sequence reads from 52 samples, with on average 127 452 reads per sample, totalling 470 567 OTUs (75.7% of these [356 512 OTUs] were singletons). Prokaryote communities were dominated by taxa typical of marine environments, including the bacterial classes Alphaproteobacteria, Gammaproteobacteria and Chloroflexi-type SAR202 and the archaeal class Thaumarchaeota (**Figure 2**). The relative abundances of these major classes were seen to vary significantly with season and depth. The Thaumarchaeota, Chloroflexi-type SAR202, AB16 (Marine Group A, originally SAR406) and Deltaproteobacteria were only observed to predominate in waters where light availability was low (in surface waters during the winter months and in deeper waters the year round). Representation of the Gammaproteobacteria and Flavobacteriia however was greatest when light availability and phytoplankton levels were highest (in epipelagic waters in the summer months). In the darker months (January, March and November), the relative abundances of the Alphaproteobacteria were recorded at similar levels irrespective of sample depth. In May however, levels were substantially higher in deeper waters, whilst August saw an abrupt shift to similarly high levels in surface waters.

Principal coordinate analyses (PCoA) on unweighted UniFrac distances (**Figure 3**) indicated that light availability and depth


TABLE 1 | Chemical and biological parameters for samples taken from the Arctic Ocean around the Svalbard Archipelago in 2014.

<sup>a</sup>Research Cruise, c, Carbon Bridge; m, MicroPolar; <sup>b</sup>SW, Surface Water; ArW, Arctic Water; AW, Atlantic Water; cAW, Cold Atlantic Water; IW, Intermediate Water; NA, Not Available; c Carbon Bridge; m MicroPolar.

(of which light availability is a factor) primarily drove the phylogenetic beta-diversity across prokaryote communities, with three clearly separated clusters; one cluster (**Figure 3A**) containing communities from deep and dark mesopelagic waters, predominantly those samples collected from January, March and November; a second cluster (**Figure 3B**) comprising

communities from samples collected from shallow and light epipelagic waters in May and August only; and a third cluster (**Figure 3C**), an admixture of communities from shallow and dark epipelagic waters in January, March and November and mesopelagic waters from May and August. Prokaryote alphadiversity generally decreased as light availability increased to a maximum in the summer (**Figure 4** and Figure S1). Within each sampling period, Chao-1 estimates of OTU richness increased with increasing depth, with deeper waters containing a greater richness than surface waters, with maximum richness over the year observed in the deepest January samples. In surface waters, richness decreased until a minimum in August, before increasing again as light availability decreased during the late autumn.

The seasonal and depth variation in the relative abundance of eleven major prokaryote taxa (**Figure 5**) confirmed the impact of increased light availability and the resulting phytoplankton blooms on Arctic marine microbial populations in epipelagic waters. Generally, the dominant taxa in surface waters were seen to exhibit either a significant positive or negative population change in the summer months, whilst corresponding populations in deeper waters remained relatively unchanged. The relative abundance of 16S rRNAs derived from chloroplasts reached a maximum in May (**Figure 5**) and this was reflected by maxima in the Oceanospirillaceae, Alteromonadaceae and Flavobacteriaceae and minima in the archaeal family Cenarchaeaceae, Rhodospirillaceae, Nitrospinaceae and OM27, moving from March into May - in all these taxa, relative abundance increased substantially with the transition into the darker winter months. The family AEGEAN185 were unusual in that there was sharp decrease in the May relative abundance in deeper waters, as were the group SAR324 which saw concurrent minima in both surface and deeper waters.

The second bloom in August, characterized by a different phytoplankton community to the earlier May one (E. S. Egge, Personal Communication), saw maxima in both the families Halomonadaceae and Rhodobacteraceae.

# 4. DISCUSSION

Recent advances in high-throughput sequencing technologies and software development for data management and processing have helped to shed light on the composition and seasonal dynamics of marine microbial communities, suggesting that many follow cyclical and predictable patterns (Fuhrman et al., 2006, 2015). Our study collected water samples in the seas around Svalbard from nominal depths ranging from 1 to 1000 m throughout the year, spanning the whole spectrum of environmental light conditions from the total darkness of the polar night to the perpetual illumination of the polar day. Our analysis of prokaryote 16S ribosomal RNA diversity using high-throughput sequencing confirmed that the microbial communities of Arctic waters during the polar winter and summer varied significantly. Additionally, this data suggested that seasonal and depth-related light availability and sea conditions and the associated phytoplankton blooms are the primary drivers for successional changes in community composition in these waters. Phylogenetic diversity increased with decreasing illumination, with regards to both seasonality and water depth, with the greatest richness to be found in the deepest and darkest water samples. The phytoplankton bloom and post-bloom stages dominated surface water communities in May and August, respectively, and saw corresponding increases in the relative abundance of bloom-associated copiotrophic organisms related to the Gammaproteobacteria and Flavobacteriia. The chemolithotrophic Thaumarchaeota

and Chloroflexi-type SAR202 dominated deep aphotic waters all year round but varied significantly in surface waters with varying light levels, proliferating in the dark winter months and diminishing in the well-lit summer months. Whether these taxa were responding directly to the changes in light availability or nutrient composition, or were outcompeted by the resulting phytoplankton blooms and successional prokaryotes is not known. The objectives of this research were to assess these microbial communities in time and space and investigate how these organisms respond down through the water column to the extreme variations in surface environmental conditions during the polar year.

We observed distinct seasonal fluctuations in prokaryote diversity, comprising high richness in the darker autumn and winter months and lower richness in the late spring and summer (Gilbert et al., 2012; El-Swais et al., 2015), similar to other surveys of surface waters in high latitude marine ecosystems (Murray et al., 1998; Murray and Grzymski, 2007; Ghiglione and Murray, 2012; Grzymski et al., 2012; Ladau et al., 2013). Phylotype richness peaked in January and then decreased through the year (**Figure 4**) until the annual minimum in August, coincident with the late summer phytoplankton post-bloom, before increasing again in the late autumn. Cyclic annual patterns of prokaryote community structure have been observed in waters off the Antarctic Peninsula (Murray et al., 1998; Church et al., 2003; Murray and Grzymski, 2007) and had sampling in our study persisted, we predict that phylotype richness would likely have peaked again in January (Fuhrman et al., 2015). Rarefaction curves (Figure S1) also indicated that there was a distinct change in the number of unique phylotypes with season, which contrasts with seasonal pyrosequencing data for the Western Arctic (Kirchman et al., 2010). However, the authors do concede that the depth of their sequencing efforts may not have captured the complete diversity of these communities. Samples in the current study were rarified to 62500 sequences (tenfold higher than that of Kirchman et al., 2010) and this was sufficient to illustrate seasonal differences in community diversity. Whilst rarefaction curves for the May and August surface samples approached an asymptote, showing that coverage by these libraries was high, sequencing efforts for the winter months were typically undersaturated, as evidenced by the continued upwards curve of the January samples even at 170 000 reads (data not shown).

There was a clear trend in the variation in relative proportions of the two prokaryote domains in surface waters over the annual cycle, with the Archaea increasing to maxima in the winter months and decreasing to almost undetectable levels in the summer months (**Figure 5**). This was reflected in reverse by the Bacteria, mirroring seasonal observations of near-surface prokaryote communities in the Western Arctic (Alonso-Sáez et al., 2008). In our study, the Proteobacteria were the most abundant bacterial phylum throughout the year and of these, it was the class Gammaproteobacteria that predominated, particularly in the summer months, along with class Flavobacteriia of the phylum Bacteroidetes (**Figure 2**). The high relative abundance of these taxa in coastal waters around Svalbard (Zeng et al., 2013) has been associated with the release of dissolved organic matter following a phytoplankton bloom (Ghiglione and Murray, 2012; De Corte et al., 2013; Buchan et al., 2014; El-Swais et al., 2015) and they are seen to dominate Arctic MYI communities (Bowman et al., 2012). The Gammaproteobacteria are typically seen to increase toward the summer months (Alonso-Sáez et al., 2008) and of the strong indicator phylotypes determined during the study year, three of the top five (Oceanospirillaceae, Halomondaceae and Alteromonadaceae) were members of this class (**Figure 5**). Members of both the Oceanospirillaceae and Alteromonadaceae are known to be r-strategist, broad substrate generalists and are frequently seen to be closely-associated with phytoplankton blooms (Teeling et al., 2012; Buchan et al., 2014; El-Swais et al., 2015). However, in comparison with the abundance of physiological and genomic studies of bloom-associated Bacteroidetes and Alphaproteobacteria strains, there is a relative paucity of data regarding the Gammaproteobacteria (Buchan et al., 2014).

Recent whole genome analyses of various members of the Bacteroidetes have confirmed long-held assumptions of a preference for and selective advantage of the phylum when growing on complex organic matter (Abell and Bowman, 2005; Teeling et al., 2012; Fernández-Gómez et al., 2013; Williams et al., 2013), such as that typically produced by marine phytoplankton (Passow, 2002). Consequently, Bacteroidetes and phytoplankton are often found in close association in polar waters (Grossart et al., 2005; Piquet et al., 2011; Williams et al., 2013) with the relative abundance of the former significantly correlated with the emergence of the late spring blooms (Alonso-Sáez et al., 2008). During May of 2014, the Flavobacteriaceae were notable by their prolific increase in relative abundance from negligible levels in March, congruent with the spike in chloroplast 16S rRNA abundance (**Figure 5**) and chl a maxima (**Table 1**). Secondary bacterial production is typically correlated with chl a concentration (Buchan et al., 2014). The Flavobacteriaceae are one of the most commonly found groups in polar ecosystems (Abell and Bowman, 2005), particularly in the summer months (Grzymski et al., 2012; Williams et al., 2012), frequently comprising the majority of Bacteroidetes sequences in these environments (Ghiglione and Murray, 2012; Williams et al., 2013) and amongst the first groups to respond to phytoplankton blooms (Williams et al., 2013). Previously, the Flavobacteriia have been found at peak abundance during the decay phase of a bloom (Riemann et al., 2000; Pinhassi et al., 2004) but this was not the case in our post-bloom August samples (**Figure 2**). Proteorhodopsins, which support photoheterotrophic growth, have however been found in flavobacterial isolates (Gómez-Consarnau et al., 2007) previously and this may explain the peak abundance observed during the higher photoirradiance conditions of May. It has been suggested that increased numbers of Bacteroidetes would be found in the water column during spring and summer as a consequence of increased melting sea ice, either by seeding, as persistent members of sea ice biota (Bowman et al., 2012; Lo Giudice et al., 2012) or as a result of growth on organic matter released by the thawing (Piquet et al., 2011). As an abundance of Flavobacteriial proteins involved in oxidative stress protection have also been recovered from Antarctic metaproteomes, this suggests that the group may also exhibit a higher tolerance to the high solar irradiance found in polar spring and summer waters (Williams et al., 2013) and may indeed come to play a more dominant role in the surface layers of an ice-free Arctic.

A notable outcome immediately apparent from the 16S rRNA phylogenetic diversity data was the glaring disparity in the relative abundance of Alphaproteobacteria, when compared with previously published studies which saw them dominate polar waters (Alonso-Sáez et al., 2008; Manganelli et al., 2009; Bowman et al., 2012; Williams et al., 2013; Zeng et al., 2013). Closer examination revealed that the ubiquitous SAR11 were notable by their unusually low representation in the dataset, with their relative abundance no higher than 2.5% during the entirety of the study. The SAR11 typically comprise the greater part of observed Alphaproteobacteria in global marine communities (Morris et al., 2002), and more pertinently in the Arctic (Alonso-Sáez et al., 2008; Kirchman et al., 2010; Bowman et al., 2012). In a previous study in Svalbard coastal waters (De Corte et al., 2013), SAR11 were only detected at the 16S rDNA level, none being detected in the 16S rRNA; in the present investigation however, similarly low levels of SAR11 were discerned in both 16S rDNA and rRNA (data not shown). In silico analysis of the universal prokaryotic primer set 519F-806R (Øvreås et al., 1997; Caporaso et al., 2011) used in this study suggested a low-binding efficiency of the reverse primer with the SAR11 cluster, an inherent flaw of the primers confirmed in recent studies (Apprill et al., 2015; Parada et al., 2016). Preliminary shotgun metagenomic data for this same polar time-series did reveal a significantly higher abundance of SAR11 (up to 28% of total prokaryote abundance; data not shown) and so, we can assume that the SAR11 are likely underrepresented in this 16S tag amplicon data set.

Members of the Alphaproteobacterial families Rhodospirillaceae and Rhodobacteraceae were however highly abundant in Arctic waters over the year, with both taxa appearing to be significantly correlated (negatively and positively, respectively) with different phytoplankton bloom stages (**Figure 5**). The Rhodobacteraceae were the most abundant OTUs recorded in late summer seas off Svalbard (Zeng et al., 2013) and the genus Roseobacter within this family has been seen to peak with chl a maximum in the Western Arctic (Alonso-Sáez et al., 2008), which is assumed to relate to this taxon's common association with phytoplankton blooms (González et al., 2000; Pinhassi et al., 2004; West et al., 2008; Buchan et al., 2014). In the current study however, Roseobacter species were not detected at the Genus level using the Greengenes database (DeSantis et al., 2006) but comparison of unassigned Rhodobacteraceae OTUs with BLAST confirms that they were present, albeit not at the levels seen in similar studies during the polar summer (Grzymski et al., 2012).

The Arctic Ocean has the greatest freshwater input of any ocean (Carmack, 2007) and so as might be expected, the Betaproteobacteria, one of the most prevalent groups in freshwaters, were present in the surface marine samples all year round (**Figure 2**) albeit at low levels; it is the corresponding absence of rivers in Antarctica which supports their recorded low levels in Southern Oceans (Ghiglione et al., 2012). Interestingly, we observed that the relative abundance of Betaproteobacteria in surface water samples decreased as a function of distance either from the coast or from the pack ice. A number of large glacial fjords flow into the Western coastal waters off Svalbard (**Figure 1**) and the Betaproteobacteria have been retrieved from Kongsfjorden there (Zeng et al., 2009, 2013; Piquet et al., 2010). The group have also been seen to predominate in Arctic pack ice summer melt pools (Brinkmeyer et al., 2004), which might further seed ocean waters upon thawing. In more temperate latitudes, the abundance of Betaproteobacteria decreases with increasing salinity (Garneau et al., 2009) but in the colder waters of the Arctic, they may persist further offshore (Pedrós-Alió et al., 2015). With climate change bringing with it increased freshwater inputs to the Arctic Ocean, we may also see the relative abundance of the Betaproteobacteria in marine microbial communities rise.

The lower diversity of surface waters during summer appears to be related to both taxonomic and methodological factors; whilst the spike in carbon and nutrient concentrations following a phytoplankton bloom inevitably leads to the proliferation of a few specific bacterial groups (Buchan et al., 2014), the elevated levels of phytoplankton in surface waters can often overwhelm efforts to sample representatively. Indeed, during this study, filters were frequently blocked by phytoplankton cell debris and consequently, without preventative measures, chloroplast 16S rRNAs would routinely comprise >90% of amplicon tag sequences, thereby reducing the sequencing coverage of targetted prokaryotes. Accordingly, prokaryote richness in surface waters during the winter months (and in deeper waters year round) was considerably higher, potentially also due to the more complex composition of the dissolved organic carbon pool (Alonso-Sáez et al., 2008) in darker waters not dominated by just a few bloomrelated taxa.

Whilst light availability is thought to be the main driving factor in the epipelagic zone (Giovannoni and Stingl, 2005), conditions within deep dark waters are far from homogenous (Hewson et al., 2006; Teira et al., 2006; Galand et al., 2010). The oceans comprise regional water masses, defined by their distinct temperature and salinity properties, and these circulate around the globe at different spatial scales. Our sampling area off the west coast of Svalbard is notable as being the confluence of North Atlantic and Arctic waters and depending on sample season, location and depth, the deep water column may comprise a number of different water masses (**Table 1**). A study of the deep Arctic Ocean suggested that these water masses may act as physical barriers to microbial dispersal and that communities within these water masses may therefore have a distinct biogeography (Galand et al., 2010). Preliminary analyses of our data however did not seem to suggest a relationship between water mass and community composition (data not shown), with the effect of light and depth being much more significant (**Figure 3**).

Microbial life in the mesopelagic zone is quite different from that of the epipelagic zone (Orcutt et al., 2011) and whilst the absence of light is the most obvious difference, the deep oceans are also typically colder and at a higher pressure (increasing by approximately 10 MPa km−<sup>1</sup> depth) than surface waters. Within each seasonal sample in our study, there was a clear difference in prokaryote composition between epipelagic and mesopelagic communities (**Figure 2**) and as might be expected, these differences were much more pronounced during the lighter months of May and August, driven by the massive change in surface communities during the phytoplankton blooms. In the darker months, communities were much more similar with depth and were dominated throughout the water column by the Thaumarchaeota and Chloroflexi-type SAR202. Seasonal changes had little effect on mesopelagic communities, with the moderate variations in prokaryote diversity congruent with the relatively unchanging nature of the deep marine environment. However, we do see a significant difference in phylotype richness in mesopelagic waters between January and the other months and one reason for this might very well be January's placing in the polar year, in the middle of the dark Arctic winter. The phytoplankton bloom results in a single annual pulse of nutrients and in the successional development of mesozooplankton and heterotrophic prokaryotes, the bulk of the simple compounds will be degraded relatively rapidly in the photic surface waters. What is left sinks as marine snow, which continues to be metabolized by the prokaryote community on its way down through the water column. By the time much of this matter reaches the mesopelagic depths, several months after the initial deposition from surface waters, what remains is likely to be diverse relcalcitrant compounds. The primary factor differentiating the deep ocean from surface waters is that metabolic reactions are based only on chemical redox reactions (rather than photosynthetic processes) and much more variably coupled temporally, spatially and functionally (Orcutt et al., 2011). The more diverse chemical processes ongoing in the deep ocean will therefore likely require more functionally (and phylogenetically) diverse organisms than in surface waters, the richness of which seemingly peaks in the winter months.

Unmitigated atmospheric warming will undoubtedly affect gross changes in the marine environment and its phytoplankton populations through increased photoirradiation and nutrients made available by thawing sea ice and permafrost (Li et al., 2009; Doney et al., 2012; El-Swais et al., 2015), respectively. Whilst the phytoplankton blooms are restricted to the well-lit upper layers of the ocean, the effects of their prolific but short-lived activity are seen throughout the deep ocean beneath. As these summer blooms are inextricably linked to the deep and dark activities of the winter blooms of chemolithoautotrophic organisms, a change in one system will inevitably be manifested in the other, potentially via more subtle effects on specific taxonomic or functional prokaryote groups. As major contributors to the marine nitrogen cycle (Nicol and Schleper, 2006) and the dominant chemolithoautotrophs in polar waters, the role of the Thaumarchaeota (previously classified as members of the phylum Crenarchaeota Brochier-Armanet et al., 2008) is wellestablished (Fuhrman et al., 1992; DeLong et al., 1994; Karner et al., 2001). More recently, the cyclical rise (during winter) and decline (in summer) of the group in the surface waters of polar regions has received some attention (Murray et al., 1998, 1999; Church et al., 2003; Alonso-Sáez et al., 2008, 2012; Grzymski et al., 2012; Williams et al., 2012) and indeed, our data confirmed these findings. The almost total disappearance of the Thaumarchaeota from the photic zone during the summer months (**Figure 5**) is therefore suggestive of photoinhibition of ammonia oxidation (Guerrero and Jones, 1996; Murray et al., 1998; Mincer et al., 2007; Schleper and Nicol, 2010; Merbt et al., 2012), or has been more recently posited, a sensitivity to reactive oxygen species produced as a result of photosynthesis (Tolar et al., 2016). The reasons may however be multifactorial and it is thought that the Archaea may also be outcompeted by phytoplankton (Murray et al., 1998; Ward, 2000, 2005; Church et al., 2003; Herfort et al., 2007; Smith et al., 2014) and Bacteria (which are much more active in the uptake of the labile bloomproduced substrates Alonso-Sáez et al., 2008; Kalanetra et al., 2009), or even subjected to selective viral infection (Labonté et al., 2015). The proportional abundance of Thaumarchaeota has been correlated with ammonium concentrations (Herfort et al., 2007; Kirchman et al., 2007) and their peak abundance in winter surface waters has been hypothesized to result from mixing with deep water masses in Antarctic seas (Kalanetra et al., 2009; Grzymski et al., 2012); however, in areas of the Arctic Ocean where the water column tends to remain stratified during the winter (Forest et al., 2011), recent data suggests that the increase is in fact due to growth and proliferation of surface water Thaumarchaeota populations in situ (Alonso-Sáez et al., 2012). As the polar winter precludes photosynthesis and consequently, a source of labile organic matter, the autotrophic Thaumarchaeota are ideally adapted to bloom in these otherwise limiting conditions (Pedrós-Alió et al., 2015). Interestingly, in both Arctic (Alonso-Sáez et al., 2012) and Antarctic (Kalanetra et al., 2009; Grzymski et al., 2012) winter surface waters, a single Thaumarchaeotal OTU was seen to dominate archaeal 16S rRNA and ammonia monooxygenase amoA gene libraries and we saw a similar dominance in Arctic winter surface waters; however, as we continued down through the water profile, this dominant OTU gradually yields to another Thaumarchaeotal OTU, which ultimately dominates deeper waters. These shallow and deep cladal differences have been described in ammonia-oxidizing Archaea (AOA) previously, based primarily in differences in their amoA genes (Francis et al., 2005; Sintes et al., 2013; Pedneault et al., 2014; among others) but also other metabolic genes (Sintes et al., 2013; Luo et al., 2014; Villanueva et al., 2015).

Any change in Thaumarchaeota populations may have significant impacts on certain of the phytoplankton populations. In unlit surface waters, the Thaumarchaeota oxidize ammonium to nitrate, which in turn promotes phytoplankton growth come the lighter months; however, as the larger diatoms are thought to outcompete picophytoplankton for nitrate (Stolte and Riegman, 1995), diatom populations may be proportionally more affected by a loss of Thaumarchaeota (Comeau et al., 2011). Conversely, a warming freshening Arctic will be increasingly stratified and there will be less mixing with the nutrient-rich depths, potentially favoring smaller picophytoplankton (Li et al., 2009).

Taxa showing similar seasonal dynamics to the chemolithotrophic Thaumarchaeota in surface waters were the Deltaproteobacteria-affiliated SAR324, Nitrospinaceae and OM27 (**Figure 5**), all of which saw drops in relative abundance during the summer months. The nitrite-oxidizing genus Nitrospina has been observed in winter-only samples from both Antarctic (Grzymski et al., 2012) and Arctic (Alonso-Sáez et al., 2010) waters, as well as in temperate waters (El-Swais et al., 2015) and has also been correlated with amoA-containing Thaumarchaeota in Monterey Bay (Mincer et al., 2007). Whilst we recorded generally higher levels of all three taxa in winter surface and year-round deeper waters, active populations were still detected in surface waters in summer months, albeit at much lower levels. The Chloroflexi-type SAR202 cluster is also typical and highly abundant in mesopelagic waters (Morris et al., 2004; Varela et al., 2008; Arístegui et al., 2009; Dobal-Amador et al., 2016). Recent studies have shown the Chloroflexi to be well adapted to deeper oligotrophic waters and to efficiently utilize recalcitrant organic compounds uptake, such as those found in the mesopelagic zone (Yilmaz et al., 2016).

# 5. CONCLUSIONS

This high-throughput sequencing study of a polar year in the Arctic Ocean revealed the driving force of light and phytoplankton blooms on the marine prokaryote community, and in particular the Thaumarchaeota, in both surface and deep ocean waters. Investigations such as this one and others discussed in the text are fundamentally important as a historic record of the current microbial state of the polar oceans and an indicator of the ongoing rate of change (Ducklow et al., 2009). Additionally, we believe that it is crucial to consider the deeper water column in these studies, as part of a whole, dynamic and interconnected marine system. As a complement to this study, we are also investigating shotgun metagenomic and metatranscriptomic data for these same seasonal samples, which will greatly improve the resolution of these preliminary results. We suggest that the future progress of the field would benefit greatly from repeated longer-term investigations, in concert with the continued improvement of omics tools and in this manner, we hope that these data may be used to support the overwhelming physical evidence of our changing global climate.

#### AUTHOR CONTRIBUTIONS

BW analyzed data, prepared figures and tables, and wrote the paper. OM, EN, LS, GB, and LØ collected, optimized, and processed samples. OM, LS, GB, and LØ were involved in project description, design of experiment, discussion, and interpreting data.

#### FUNDING

The MicroPolar project (also responsible for the cruises during March and November) was funded by the Research Council of Norway (RCN 225956). We would also like to thank the Carbon Bridge project (RCN 226415) for allowing us to participate during their cruises in January, May, and August. LØ is also affiliated with the University Centre in Svalbard (UNIS) and was funded, along with OM, by "Microorganisms in the Arctic: Major drivers of biogeochemical cycles and climate change" (RCN 227062). LS was funded through the Carbon Bridge project (RCN 226415).

#### ACKNOWLEDGMENTS

We would like to thank crew members of the Norwegian Research Vessels RV Helmer Hanssen and RV Lance for their assistance in sampling expeditions and all colleagues in the UiB Marine Microbiology group and collaborators abroad who contributed to the research effort.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2017.00095/full#supplementary-material

#### REFERENCES


and surface seawater by 454 sequencing of the 16S RNA gene. ISME J. 6, 11–20. doi: 10.1038/ismej.2011.76


Marquis, K. B. Avery, M. Tignor, and H. L. Miller (Cambridge, UK: Cambridge University Press), 235–336.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Wilson, Müller, Nordmann, Seuthe, Bratbak and Øvreås. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Unanticipated Geochemical and Microbial Community Structure under Seasonal Ice Cover in a Dilute, Dimictic Arctic Lake

Ursel M. E. Schütte1,2 \*, Sarah B. Cadieux3,4, Chris Hemmerich<sup>5</sup> , Lisa M. Pratt<sup>3</sup> and Jeffrey R. White1,6

1 Integrated Program in the Environment, Indiana University, Bloomington, IN, USA, <sup>2</sup> Institute of Arctic Biology, University of Alaska Fairbanks, Fairbanks, AK, USA, <sup>3</sup> Department of Geological Sciences, Indiana University, Bloomington, IN, USA, <sup>4</sup> University of Illinois at Chicago, Chicago, IL, USA, <sup>5</sup> Center for Genomics and Bioinformatics, Indiana University, Bloomington, IN, USA, <sup>6</sup> School of Public and Environmental Affairs, Indiana University, Bloomington, IN, USA

#### Edited by:

Ingrid Obernosterer, Observatoire Océanologique de Banyuls sur mer, France

#### Reviewed by:

André M. Comeau, Dalhousie University, Canada Dimitri Kalenitchenko, Université Pierre et Marie Curie, France

> \*Correspondence: Ursel M. E. Schütte uschuette@alaska.edu

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Microbiology

Received: 01 March 2016 Accepted: 20 June 2016 Published: 05 July 2016

#### Citation:

Schütte UME, Cadieux SB, Hemmerich C, Pratt LM and White JR (2016) Unanticipated Geochemical and Microbial Community Structure under Seasonal Ice Cover in a Dilute, Dimictic Arctic Lake. Front. Microbiol. 7:1035. doi: 10.3389/fmicb.2016.01035 Despite most lakes in the Arctic being perennially or seasonally frozen for at least 40% of the year, little is known about microbial communities and nutrient cycling under ice cover. We assessed the vertical microbial community distribution and geochemical composition in early spring under ice in a seasonally ice-covered lake in southwest Greenland using amplicon-based sequencing that targeted 16S rRNA genes and using a combination of field and laboratory aqueous geochemical methods. Microbial communities changed consistently with changes in geochemistry. Composition of the abundant members responded strongly to redox conditions, shifting downward from a predominantly heterotrophic aerobic community in the suboxic waters to a heterotrophic anaerobic community in the anoxic waters. Operational taxonomic units (OTUs) of Sporichthyaceae, Comamonadaceae, and the SAR11 Clade had higher relative abundances above the oxycline and OTUs within the genus Methylobacter, the phylum Lentisphaerae, and purple sulfur bacteria (PSB) below the oxycline. Notably, a 13-fold increase in sulfide at the oxycline was reflected in an increase and change in community composition of potential sulfur oxidizers. Purple non-sulfur bacteria were present above the oxycline and green sulfur bacteria and PSB coexisted below the oxycline, however, PSB were most abundant. For the first time we show the importance of PSB as potential sulfur oxidizers in an Arctic dimictic lake.

Keywords: microbial communities, geochemistry, Arctic, lakes, seasonally ice-covered

# INTRODUCTION

Lakes are important participants in the global carbon cycle (Tranvik et al., 2009), processing carbon from terrestrial and aquatic ecosystems and contributing 6–16% of the total natural methane (CH4) emissions (Bastviken et al., 2008; Borrel et al., 2011). Despite the fact that most lakes in the Arctic are perennially or seasonally frozen (Walsh et al., 1998; Comeau et al., 2012) very little information is available on microbial communities and nutrient cycling under ice cover (Bertilsson et al., 2013). The ice-covered season affects the ecology and metabolic characteristics of microbial communities and their contributions to the food web and biogeochemical cycling

throughout the year. A reduction in ice coverage and related transition of perennially frozen lakes to seasonally frozen lakes due to warming temperatures underscores the importance of understanding microbiomes under ice for improved prediction of the responses of Arctic lakes to rapidly warming climate.

Few studies have documented prokaryotic communities under ice cover. Microbial biomass is lower under ice cover compared to the ice-free period (Personnic et al., 2009; Twiss et al., 2012) due to lower temperatures, reduced nutrient inputs and decreases in the quality of organic substrates because of lower autochthonous production and limited inputs of labile terrestrial organic matter (Tulonen, 1993; Tulonen et al., 1994; Bergström and Jansson, 2000). However, microbial communities can actively grow during winter (Bertilsson et al., 2013). Beall et al. (2015) found diverse microbial communities in Lake Erie under ice cover including among others Verrucomicrobia, Proteobacteria, and Bacteroidetes. Similarly Glatz et al. (2006) described a diverse community in permanently frozen lakes in Antarctica comprising Proteobacteria, Actinobacteria, Bacteroidetes, and Planctomycetes. These diverse microbial communities have been shown to change systematically during the ice-covered season.

Studies on linking microbial communities with geochemistry in ice-covered lakes are limited, with most focusing on perennially frozen lakes. In two Siberian seasonally ice-covered lakes, Rogozin et al. (2009) found high abundances of purple sulfur bacteria (PBS) at ice on and greatly varying abundances throughout the ice-covered season due to variations in light availability and redox conditions. In meromictic perennially frozen lakes green sulfur bacteria (GSB) are commonly found and PSB are not detected (Lauro et al., 2011). Both PSB and GSB are important sulfur oxidizers in these systems that utilize light. The relative abundance between PSB and GSB can change by season based on changes in physicochemical gradients and light availability (Abella et al., 1979; Blankenship et al., 1995). PSB and GSB differ in their pigment content; PSB use bacteriochlorophyll a (BChl a), while GSB use BChl c, d, and e (Guerrero et al., 1985; Stomp et al., 2007; Rogozin et al., 2009). This divergence in pigment composition of phototrophic microorganisms and the associated absorbance characteristics appears to support species coexistence and biodiversity in aquatic environments (Stomp et al., 2004, 2007). Villaescusa et al. (2010) and Bielewicz et al. (2011) attributed changes in microbial community composition in permanently frozen lakes in Antarctica to changes in overall productivity, stratification, and nutrient availability. Comeau et al. (2012) observed that bacterial community composition differed between two time points of ice coverage in a high Arctic lake in Canada, and that microbial communities were stratified by depth with good agreement to chemical gradients.

Here, we assessed the vertical microbial community distribution and geochemical composition under ice in a seasonally ice-covered lake in Southwest Greenland. During preliminary sampling under ice cover in 2013, we detected a diverse community of PSB forming a pinkish layer at a depth of 6 m and identified the most abundant operational taxonomic unit (OTU, 39%) as Lamprocystis. The dense community of PSB corresponded with both a sharp redox gradient and an increase in sulfide (6H2S) and methane (CH4) concentrations. In 2014, using amplicon-based sequencing targeting the 16S rRNA genes and aqueous geochemical approaches, we found a highly diverse microbial community throughout the water column and very good spatial agreement between CH4, sulfur, and nitrogen gradients and changes in the microbial community composition. In addition, we show for the first time PSB to be the most abundant potential sulfur oxidizers in a dimictic, Arctic lake compared to green sulfur bacteria (GSB) and purple non-sulfur bacteria (PNSB) that have been shown to be the more abundant sulfur oxidizers in permanently frozen lakes in the Arctic (Comeau et al., 2012) and Antarctica (Karr et al., 2003).

# MATERIALS AND METHODS

#### Site Description

The region of southwest Greenland between 66◦ and 68◦N is characterized by low continental climate and is strongly influenced by the high-pressure system over the Greenland Ice Sheet (GIS; Bennike, 2000). Mean annual air temperature at the Kangerlussuaq weather station from 1974 to 2012 was −6 ◦C. During the warm season of late May to early September, mean temperatures exceed +10◦C and peak to +20◦C in July. The cold season extends from November to late March, with average daily temperature below −9 ◦C. Precipitation is minimal, with annual precipitation <150 mm yr−<sup>1</sup> , mostly in the summer months. Bedrock in the region is composed of complexly deformed Archean felsic gneisses (Jensen et al., 2002). The terrestrial plant community is primarily dwarf-shrub tundra, dominated by species such as Salix arctica, Salix glauca, Betula nana, Vaccinium uliginosum, Sphagnum spp., and various grasses and sedges.

The area has continuous permafrost <50 cm below the surface extending down to 150–500 m depth (Jørgensen and Andreasen, 2007). Lakes in the region are frozen for ∼10 months of the year, with ice-out occurring by mid-June and re-freezing in mid-September (Anderson and Brodersen, 2001). As a result of minimal precipitation, lakes are supplied with water mainly through contributions from melting of the limited snowpack. Through-going drainage is absent and groundwater seepage into the small lakes is assumed to be limited due to continuous permafrost. Consequently, the small lakes of the region are primarily isolated evaporitic basins at this point in time.

# Potentilla Lake

Potentilla lake (67◦ 040 51.6400N, 50◦ 210 17.1400W) is located ∼6 km from the terminal moraine of the Russell Glacier (Colcord et al., 2015; Webster et al., 2015; Cadieux et al., 2016; Goldman et al., 2016). It is a relatively small lake, with a surface area of 1.6 ha, volume of 56,000 m<sup>3</sup> , and maximum depth of 7.8 m and it is ice-covered for about 9 months of the year. The lake basin is elliptical and asymmetric. Potentilla is a dilute, oligotrophic lake, with mean conductivity of 170 µS cm−<sup>1</sup> and mean dissolved organic carbon of 11 mg L−<sup>1</sup> (Cadieux et al., 2016). Under open-water conditions in 2013, the pH ranged from 7.2 to 9.0. Currently, Potentilla is a hydrologically closed basin, with no direct inflow or outflow channel, although there is evidence of overflow during previous wetter periods. While many lakes in

the Kangerlussuaq region are cold monomictic (Anderson et al., 2001), Potentilla lake is a dimictic lake.

#### Sampling and In Situ Analysis

Sampling for this study took place in April 2014, at the end of winter stratification when the lake was covered with ∼2 m of ice and ∼20 cm of snow. A hole (∼30 cm in diameter) was augered through the ice above maximum depth (Zmax). Profiles of temperature (T, ◦C), dissolved oxygen (DO, %), pH, and oxidation-redox potential (ORP, mV) were measured using a YSI 6093 Data Sonde (Yellow Springs Inc., Yellow Springs, OH, USA). A Li-Cor LI193 Spherical Quantum Sensor (Li-Cor, Lincoln, NE, USA) was used to measure photosynthetically active radiation (PAR, µmol m−<sup>2</sup> s −1 ) attenuation with depth. Water samples were taken with an electronic submersible pump at 0.5 m depth intervals from just below the ice at 2.5 to 7.0 m (nine sample intervals). For chemical analysis, samples were filtered within 24 h of collection using a series of Whatman GF/F glass microfiber filters and Millipore membrane filters and frozen immediately until analysis. For microbial analysis, 1 L of water was collected into sterile HDPE Nalgene bottles and filtered through Millipore Isopore Membrane Filters (0.22 µm) that were frozen immediately and transferred to a −80◦C freezer until analysis.

#### Chemical Analysis

Total dissolved sulfide concentrations (6H2S) in the water column were determined gravimetrically by precipitation of H2S(aq) and HS−(aq) as cadmium sulfide (CdS) precipitate. Ten liters of water were collected in Nalgene carboys, and fixed in the field immediately upon collection by using precharged sample containers with 200 mL of supersaturated cadmium chloride (CdCl2) solution (Szynkiewicz et al., 2009). The precipitated CdS was then extracted using an acid-volatile sulfide (AVS) technique (Szynkiewicz et al., 2009) and the S precipitated as silver sulfide (Ag2S). Sulfide concentrations were then determined by the gravimetric yields of collected Ag2S, and assuming a stoichiometry of H2S/HS−:Ag2S of 1. Minimum detection limits using cadmium precipitation are 0.5 µM based upon laboratory experiments with pure H2S and water. Concentrations of sulfate (SO<sup>4</sup> <sup>2</sup>−), nitrite (NO<sup>2</sup> −), nitrate (NO<sup>3</sup> <sup>−</sup>) and ammonium (NH<sup>4</sup> <sup>+</sup>) were analyzed using a Dionex ICS 2000 Ion Chromatograph using a CS12A analytical cation column, CSRS 300 4 mm suppressor, and 20 mM methanesulfonic acid eluent for cations and AS11-HC analytical anion column, ASRS 4 mm suppressor and 30 mM potassium hydroxide eluent for anions. Minimum detection limits were NH<sup>4</sup> <sup>+</sup> = 0.5, NO<sup>2</sup> <sup>−</sup> = 0.6, SO<sup>4</sup> <sup>2</sup><sup>−</sup> = 1.8, and NO<sup>3</sup> <sup>−</sup> = 0.2 µeq L−<sup>1</sup> .

Water samples for dissolved CH<sup>4</sup> in the water column were immediately stripped in the field using a headspace-equilibrium technique (Westendorp, 1985) to extract CH<sup>4</sup> from water using a 1 L Erlenmeyer flask. Headspace gas in the flask was displaced into a Cali-5-Bond bag using surficial lake water. The concentrations of dissolved aqueous CH<sup>4</sup> were measured using a Los Gatos Research (LGR) Methane Carbon Isotope Analyzer (MCIA; LGR, Mountain View, CA, USA) that was operated at the field lab in Kangerlussuaq (Cadieux et al., 2016). All samples were processed within 24 h of collection. The total concentration of CH<sup>4</sup> in each sample was corrected for dilution and calculated from the sum of the measured headspace partial pressure and the dissolved CH<sup>4</sup> remaining after gas stripping, according to Henry's law using values from Lide and Fredrikse (1995) and Cadieux et al. (2016). Instrumental uncertainty on CH<sup>4</sup> concentrations from the MCIA was ± 0.5 ppmv, which is one standard deviation of the values for gas standards analyzed during sample runs.

### Microbial Community Analysis

Genomic DNA and RNA were co-extracted using the PowerWater RNA isolation kit (MO BIO Laboratories, Carlsbad, CA, USA). As we were interested in describing changes in community composition we only used the genomic DNA. Amplification of the 16S rRNA genes was performed using the forward primer 515f and a barcoded reverse primer 806r (Caporaso et al., 2012<sup>1</sup> ). Each reaction was 25 µL in volume, with 5 µL 5× HF Buffer (New England Biolabs, Ipswich, MA, USA), 1.5 µL of 50 µM MgCl2, 0.5 µL 10 mM dNTPs, 0.5 µL of each 10 µM primer (Integrated DNA Technologies, Coralville, IA, USA) 0.25 µL of 2000 U/mL Phusion DNA Polymerase (New England Biolabs, Ipswich, MA, USA), 10–20 ng of DNA, and 15.75 µL PCR water. Amplification was performed using an initial incubation step at 94◦C for 3 min followed by 35 cycles of 94◦C for 45 s, 50◦C for 1 min, 72◦C for 1.5 min, and a final extensions step at 72◦C for 10 min. PCR amplicons were cleaned using the PCR purification kit (QIAGEN, Valencia, CA, USA). Amplicons were pooled in an equimolar mixture and the pool was cleaned an additional time using the Agencourt AMPure XP (Beckman Coulter, Brea, CA). Sequences were generated on the MiSeq Illumina platform using MiSeq Reagent Kit v2 (Illumina Inc, San Diego, CA, USA) and custom-made sequencing primers<sup>2</sup> . Quality control and analysis of the sequences were performed following the Mothur MiSeq SOP (Kozich et al., 2013).

#### Sequence Analysis

Quality control of the sequences was performed following the Mothur MiSeq SOP (Kozich et al., 2013) with the following settings. The MiSeq sequencer sequences each PCR amplicon from both ends giving a Read 1 and Read 2. For each read pair, a contig was assembled from Read 1 and the reverse complement of Read 2. If a base only existed in one of the reads and had a quality score (Q score) less than 25, it was marked as ambiguous. If the reads disagreed on a base and the Q score disparity was six or more, the higher quality base was used. Otherwise the base was marked as ambiguous. If the contig was longer than 260 base pairs in length, or contained one or more ambiguous bases, it was discarded. The contigs were then aligned to the SILVA 16S reference alignment http://blog.mothur.org/2014/08/08/SI LVA-v119-reference-files/ and reads that did not map between position 13859 and 23447 of the alignment were discarded. We then used the Mothur pre.cluster command to combine

<sup>1</sup>http://www.earthmicrobiome.org

<sup>2</sup>http://www.earthmicrobiome.org/emp-standard-protocols/16s/

sequences within two base pairs of a more abundant sequence to further reduce read error. We then used the chimera.uchime command in Mothur to remove chimeras from each sample with the setting dereplicate = t so that if a sequence was marked as a chimera in some but not all samples, it was only removed from samples in which it was determined to be a chimera. To remove sequences other than Bacteria or Archaea, we classified each contig sequence against the RDP PDS 16S training set using the classify.seqs Mothur command and removed any sequence classified as Chloroplast, Mitochondria, unknown, or Eukaryota. We clustered the remaining sequences using average linkage into OTUs using Mothur with a similarity cutoff of 97% (Schloss et al., 2009). The OTUs obtained through Mothur were classified using both the SILVA database (Quast et al., 2013) and the RDP PDS 16S training set. The taxonomic identification was consistent between both databases, but in a few cases SILVA provided classification with finer phylogenetic resolution. Thus, the data presented here is based on the classification obtained with the SILVA database. Mothur classifies all sequences and reports the most common result for each OTU. Each level of classification is scored by the percentage of reads within the OTU that agree. We truncated all OTU taxonomic identifications to the most specific level with at least 99% consensus. We used the proportions of each OTU within each sample for downstream analysis to account for differences in number of sequences obtained across samples. Sequences were submitted to the NCBI Sequence Read Archive, accession number SRP075219.

We calculated Chao1 richness estimator and inverse Simpson diversity for all samples using the "summary.single" command in Mothur. This command rarefies all samples to the sequence count of the smallest sample. This process is repeated 1000 times and the average values for both Chao1 and inverse Simpson are reported.

We did a classical multidimensional scaling followed by a multiple response permutation procedure (MRPP) using Bray– Curtis distance and 999 permutations to test for significant differences in microbial community composition above and below the oxycline. Methods such as non-metric multidimensional scaling were not applicable due to the large difference in number of samples and variables.

#### RESULTS

#### Overall Microbial Community Composition

The microbial community was diverse, with richness and diversity overall increasing with depth throughout the water column (**Figure 1**). We obtained overall 6,149,349 sequences after quality control and number of sequences obtained across samples ranged from 438,086 to 883,870. Community composition shifted significantly below 5.0 m depth (MRPP, p = 0.006). Richness estimates ranged from 8,557 OTUs at 2.5 m under the ice cover to over 41,308 OTUs at a depth of 6.0 m (Chao1 richness estimator, **Figure 1**). Under ice cover, suboxic conditions were observed decreasing from 45% at 2 m to 21.5% DO saturation at 4.5 m, and anoxic conditions (<2% DO) from 5.5 m down to the

FIGURE 1 | Microbial richness and diversity throughout the water column using the Chao 1 estimator and inverse Simpson index based on the number of OTUs and their relative abundances detected in each sample.

sediment/water interface (**Figure 2**). Herein, we will refer to 5.0 m as the oxycline. ORP directly correlated with DO, with a sharp decrease in ORP from 87 mV at 4.5 m to −123 mV at 5.5 m (**Figure 2**). Community composition of the abundant members appeared to respond strongly to redox conditions, shifting from a predominantly heterotrophic aerobic community in the suboxic waters to a predominantly heterotrophic anaerobic community in the anoxic waters (**Figure 3**). OTUs belonging to Sporichthyaceae, Comamonadaceae, and the LD12 freshwater group within the SAR11 Clade were abundant above the oxycline and higher abundances of OTUs belonging to the family Methylococcaceae, phylum Lentisphaerae, and PSB such

as Thiodictyon were detected below the oxycline. Phototrophs such as cyanobacteria were present directly under the ice, where PAR was >10 µmol m−<sup>2</sup> s −1 , but had low relative abundance of below 0.001%. We detected Melainabacteria in particular at 6.5 and 7.0 m with abundances of 0.2%, a phylum related to Cyanobacteria but obtaining their energy by fermentation not photosynthesis. Above the oxycline more OTUs were present that had a relative abundance of over 5–6% and that below the oxycline fewer OTUs dominated the microbial community, but a larger number of rare OTUs resulted in higher overall richness.

#### Methanogenesis and Methane Oxidation

The distribution of methanogens and methanotrophs was consistent with CH<sup>4</sup> concentrations throughout the water column (**Figure 4**). The concentrations of CH<sup>4</sup> ranged from 0.1 µM below ice cover to 225 µM at 7.0 m (**Figure 4C**). Above the oxycline, CH<sup>4</sup> concentrations were <10 µM. Methanogens were detected throughout the water column but were most abundant below the oxycline with the highest relative abundance of 0.0034% at 7.0 m (**Figure 4A**). We detected a total of 48 OTUs all belonging to the Euryarchaeota. Most OTUs were classified within the order Methanomicrobiales (Methanospirillum, Methanoregula) and only a few of the OTUs belonged to the genus Methanosaeta. Overall, the relative abundance of potential methanogens was very low compared to other OTUs suggesting that methanogenesis under ice cover was restricted primarily to the sediments.

We detected a large diversity of potential methanotrophs throughout the water column (2,243 OTUs), representing one of the most abundant functional groups within the microbial community (>20%; **Figure 4B**). All of the methanotroph OTUs classified within the family Methylococcaceae (type I methanotrophs). The relative abundance of potential methanotrophs ranged from 5.34 to 13.08% with a significant increase at 6.5 m that can be attributed to the increase in relative abundance of one OTU within the genus Methylobacter (**Figure 4B**). Two OTUs made up 77% and four OTUs made up 86% of all OTUs detected, which suggests that at the time point sampled only a few OTUs accounted for most of the methanotrophs detected. Methanotrophs were detected throughout the water column but were in high relative abundance below the oxycline where DO saturation was below 2% suggesting that aerobic and anaerobic CH<sup>4</sup> oxidation may be occurring in Potentilla lake.

#### Sulfur Oxidation and Reduction

Changes in relative abundance of sulfur oxidizers agreed with the concentrations of 6H2S throughout the water column. The community composition of sulfur oxidizers can be divided into two zones, above and below the oxycline. Above the oxycline we observed an overall lower relative abundance of potential sulfur oxidizers (<1%), which is consistent with 6H2S concentrations being below the detection limit (5 µM; **Figure 5**). The OTUs detected in the suboxic upper water column were identified as GSB within the order Chlorobiales and the PNSB family Rhodobacteraceae. Below the oxycline the relative abundance of PSB increased, with the community of sulfur oxidizers being dominated by Thiodictyon and Lamprocystis (**Figure 5**). We identified a total of 1,196 OTUs as PSB and all were classified as Chromatiaceae. The most abundant OTU was classified as Thiodictyon and had a relative abundance of 4% below the oxycline. This increase in the relative abundance of sulfur oxidizers below the oxycline and the shift in its community composition were consistent with chemical profiles. Increased 6H2S concentrations below 5.5 m could support sulfur oxidation and a higher relative abundance of sulfur oxidizers. In addition, the increase in 6H2S corresponded with a decrease in DO, as well as a decrease in PAR. The combination of chemical and physical changes below 5.5 m was coincident with a shift in community abundance to sulfur oxidizers dominated by PSB.

Potential sulfur reducers were only present at very low abundances in the suboxic water column (0.05–0.07%) and also had a significant increase in relative abundance below the oxycline (5.5–8.7%; **Figure 5**). With 1,999 OTUs, we detected a large diversity of potential sulfur reducers. The more abundant OTUs of potential sulfate reducers fall within the genus of Gram-negative metal and sulfur-reducing Geobacter, the sulfatereducing Desulfobulbus, and sulfate-, sulfite-, and sulfur-reducing Desulfobacteraceae. Sulfate was present throughout the water column, decreasing down the water column from 49 µM below the ice to 11 µM above the sediment. The decrease in SO<sup>4</sup> 2− concentrations corresponded with both DO and ORP reductions with depth (**Figures 2** and **5**). Changes in relative abundance of potential sulfate reducers corresponded with a decrease in SO<sup>4</sup> <sup>2</sup><sup>−</sup> concentrations observed below the oxycline. Below the oxycline, SO<sup>4</sup> <sup>2</sup><sup>−</sup> and 6H2S concentrations displayed opposite trends suggesting that sulfate reduction was producing H2S.

### Denitrification, Ammonia Oxidation, and Nitrogen Fixation

Potential denitrifiers composed a large part of the overall microbial community (7–16%; **Figure 6**). The distribution of potential denitrifiers followed the concentration of NO<sup>2</sup> <sup>−</sup> and NO<sup>3</sup> <sup>−</sup> through the water column. Concentrations of NO<sup>2</sup> − (median 90 µM) were approximately eight times greater than NO<sup>3</sup> <sup>−</sup> (median 10.5 µM). Relatively uniform concentrations of NO<sup>2</sup> <sup>−</sup> and NO<sup>3</sup> <sup>−</sup> were observed down the water column except for increases at 2.5 m and 4.0 m. These sudden increases in both NO<sup>2</sup> <sup>−</sup> (20%) and NO<sup>3</sup> <sup>−</sup> (>50%) corresponded with a slight increase in the relative abundance of two potential denitrifiers. One OTU within the family Comamonadaceae in the order Burkholderiales decreased by ∼12% between 2.5 and 3.5 m and then increased by ∼12% to 4.0 m. Similarly, Polynucleobacter within the order Burkholderiales decreased by ∼37% between 2.5 and 3.0 m and increased by ∼15% between 3.5 and 4.0 m and then decreased by a 40% below 5.0 m. Except for a few OTUs changing in relative abundance (Comamonadaceae within the order Burkholderiales, Rhodocyclaceae within Betaproteobacteria), the relative abundance of most OTUs identified as potential denitrifiers stayed consistent throughout the water column (**Figure 6**). This result differs from the other functional groups described above where we observed a clear shift in community composition below the oxycline.

Potential ammonia oxidizers made up a low percentage (0.1–0.34%, **Figure 6**) of the overall microbial community composition. Of the more abundant potential ammonia oxidizers, all except one classified within the order Nitrosomonadales. One OTU classified within Nitrospirales, an order that contains several genera, the largest being Nitrospira, which is known for converting NO<sup>2</sup> <sup>−</sup> to NO<sup>3</sup> <sup>−</sup> (Maixner et al., 2006). Concentrations of NH<sup>4</sup> <sup>+</sup> increased with depth, ranging between 11 and 55 µM. The relative abundance of potential ammonia oxidizers followed this increase in NH<sup>4</sup> <sup>+</sup> with depth suggesting that the increase in NH<sup>4</sup> <sup>+</sup> may result in a higher activity of ammonia oxidation, but can be attributed to the increase in two OTUs classified within the Gallionellaceae. Thus, similar to potential denitrifiers, the composition of ammonia oxidizers stayed stable throughout the water column and changes in community composition can be attributed to changes in the relative abundances of only a few OTUs.

Similar to potential ammonia oxidizers, the abundance of potential nitrogen fixers was low and ranged between 0.15 and 0.69%, which fits the general pattern that phototrophic bacteria were only present in low abundance even directly under the ice where PAR > 10 µmol m−<sup>2</sup> s −1 . The community composition of potential nitrogen fixers differed markedly above (Rhizobiales dominates) and below (Chlorobium dominates) the oxycline and one OTU classified as Cyanobacteria decreased in its relative abundance 100-fold between 2.5 (0.02%) and 6.5 m (0.0002%; **Figure 6**). Interestingly we detected 135 OTUs classified as Melainabacteria. Their relative abundance increased with depth to about 0.44%. This sibling phylum to cyanobacteria has been recently characterized as non-photosynthetic, anaerobic, and obligate fermentative (Di Rienzi et al., 2013).

#### DISCUSSION

#### Sulfur Oxidation and Reduction

We detected significant abundances of PSB around the oxycline. All OTUs identified as PSB were classified within the Chromatiaceae, which are known to be able to deposit elemental sulfur within their cells (Imhoff, 2006). Overall PSB

line shows the oxycline at 5.0 m.

were the most abundant potential sulfur oxidizer compared to both GSB and PNSB, particularly below the oxycline. The abundance and diversity of PSB had similarities (Tonolla et al., 1999; Rogozin et al., 2009, 2012; Peduzzi et al., 2011) and differences (Karr et al., 2003; Koizumi et al., 2003; Taipale et al., 2011; Comeau et al., 2012) with other studies under ice cover. Rogozin et al. (2009, 2012) found high densities of PSB, predominantly Lamprocystis, in two meromictic, seasonally icecovered lakes in Siberia and low abundance of GSB or no GSB. In one of the lakes, the density of PSB was sufficiently high in some years to induce self-shading. In perennially icecovered Lake Fryxell in Antarctica, Karr et al. (2003) detected a large diversity of phototrophic purple bacteria, but the authors only detected sequences similar to those of known PNSB. In a perennially ice-covered, meromictic lake in the Canadian Arctic, 40% of sequences belonged to the phylum Chlorobi, a group of GSB and a few to PNSB just below the chemocline, but PSB were not detected (Comeau et al., 2012). In contrast, we detected all three groups of sulfur oxidizing bacteria and found both coexistence and zonation throughout the water column. PNSB such as Rhodobacteraceae were more abundant above the oxycline and PSB and GSB coexisted below the oxycline.

Adaptive divergence of pigments, with differentially tuned absorbance spectra, has been suggested as an important factor

with the number given after the taxonomic name. If no number is given, the group only contains one OTU. Taxonomic groups with abundances too low to be visible on the graph are labeled with an '<sup>∗</sup> ' and marked with a gray box instead of a colored box. Dashed gray horizontal line shows the oxycline at 5.0 m.

in the diversity of phototrophic microorganisms in aquatic environments (Stomp et al., 2004, 2007). Development of mutualistic vertical structures in microbial communities of perennially frozen meromictic lakes in Antarctica may be partly driven by availability of light, key nutrients, and phage-mediated predation (Lauro et al., 2011). These are likely factors in the vertical structure of plankton in seasonally ice-covered lakes of the Arctic as well. For example, PSB and GSB differ in primary pigments and in the wavelengths of peak absorbance. PSB peak absorbances are between 440–486 nm and 780–860 nm, while peaks for GSB are 400–460 nm and 680–720 nm (Stomp et al., 2007). Light attenuation through snow, ice, and water is greater for longer wavelengths. Thus, we would expect the relative abundance of PSB to be greater at somewhat shallower depths. However, both PSB and GSB carry out anoxygenic photosynthesis using reduced sulfur and are likely inhibited by elevated DO present above the oxycline, and are adapted to low light conditions (Vila and Abella, 2001). GSB have been

reported to grow under near-infrared light at photon fluxes <10 µmol m−<sup>2</sup> s −1 (Saikin et al., 2014).

Major shifts in dominant species from season to season are expected in highly oligotrophic ice-covered systems (Lauro et al., 2011; Wilkins et al., 2013). In 2013, Potentilla lake was ice-covered, but had no snow cover. Our microbial community analysis showed that 30% of the overall microbial community was attributed to a single OTU of the PSB Lamprocystis at a depth of 5 m. In 2014, Potentilla lake had a significant snow cover in addition to the ice cover and PSB only contributed about 8% of the overall microbial community at 5 m. The additional snow cover changed light penetration and that likely resulted in a reduction in PSB. However, lower light levels may also have driven changes in the relative abundances of sulfur oxidizers. Further, PSB can oxidize various reduced sulfur species including sulfide, thiosulfate (Holkenbrink et al., 2011), as well as oxidize elemental sulfur (Dahl, 2008; Frigaard and Dahl, 2009). Similarly, GSB typically oxidize sulfide and thiosulfate to sulfate but lost essential genes necessary for elemental sulfur oxidation (Gregersen et al., 2011). Hydrogen sulfide concentrations dropped below the oxycline and if elemental sulfur was indeed present below the oxycline it could have resulted in PSB outcompeting GSB. However, more likely reasons seem to be differences in methodology used to determine the presence of different photosynthetic sulfur oxidizers and differences in the physical structure of the lakes sampled. Most studies use differences in BChl a versus BChl d to determine the amount of PSB and GSB present, respectively. In comparison, we used a next generation sequencing technique, which may have a higher resolution in detecting rare sulfur oxidizers. Further,

dimictic lakes such as the one sampled in this study mix twice a year and while Potentilla lake has ice cover for about 10 months a year it is not ice-covered in the summer. This suggests that seasonally dimictic Arctic lakes may be less stable systems compared to meromictic perennially frozen lakes. Thus, we postulate that PSB are more competitive than GSB in these less stable lake systems while GSB are more competitive in stable lake systems such as perennially frozen meromictic lakes in Antarctica (Lauro et al., 2011) and that PSB potentially are an important component in sulfur oxidation in Arctic dimictic lakes.

Below the oxycline we observed an inverse relationship in H2S and SO<sup>4</sup> <sup>2</sup>−, with H2S increasing down the water column corresponding with the continuous decrease in SO<sup>4</sup> 2− concentration, consistent with a significant increase in the presence of potential sulfur reducers. Overall, the groups of potential sulfate reducers identified were similar to those previously identified in perennially frozen, Antarctic Lake Fryxell (Karr et al., 2005). Previous studies have shown strong competition between sulfate reducers and methanogens in Arctic lake sediments (Lovely et al., 1982; Purdy et al., 2003) and found that when SO<sup>4</sup> <sup>2</sup><sup>−</sup> is not limiting, sulfate reducers inhibit methanogenesis by decreasing the hydrogen partial pressure below levels methanogens can utilize (Lovely et al., 1982). In Potentilla lake, potential sulfur and sulfate reducers were significantly more abundant than methanogens in the lower water column. Even at low SO<sup>4</sup> <sup>2</sup><sup>−</sup> concentrations (<50 µM), a large relative abundance of potential sulfur and sulfate reducers (5–9%) were supported that may outcompete methanogens. In other Arctic lakes, increasing temperatures and reduced ice cover have resulted in an increase of SO<sup>4</sup> <sup>2</sup><sup>−</sup> and sulfur accumulation in the sediments (Drevnick et al., 2010). Increasing SO<sup>4</sup> 2− under a warming climate may result in greater sulfur reduction rates, suppressing methanogenesis, and thereby decreasing the concentration of CH<sup>4</sup> in Arctic lakes.

# Methane Oxidation and Methanogenesis

Potential methanotrophs composed a large proportion of the microbial community. Below 6.5 m concentrations of CH<sup>4</sup> were comparatively high (160–220 µM) and DO saturation was <2%, suggesting that methane oxidation was occurring either aerobically or anaerobically. Several studies have shown anaerobic methane oxidation (AOM) coupled to sulfate (Boetius et al., 2000; Deutzmann and Schink, 2011; Milucka et al., 2012), iron, and manganese reduction (Beal et al., 2009) in marine systems, and coupled to nitrate and nitrite reduction in a bioreactor system (Haroon et al., 2013) and an enrichment culture from a peatland (Zhu et al., 2012). The known anaerobic methanotrophic Archaea (ANME) are related to the genus of methanogens Methanococcoides (Knittel et al., 2005; Schleper et al., 2005; Cui et al., 2015) and distantly to Methanosarcinales and Methanomicrobiales. We detected two OTUs of very low abundance within the Methanosarcinales, but several OTUs within the order of Methanomicrobiales. The ANMEs distantly related to the Methanomicrobiales were shown to carry out sulfate dependent anaerobic methane oxidation (Knittel et al., 2005). However, compared to the relative abundance of the aerobic methanotrophs we detected, the relative abundance of ANMEs was low. This suggests AOM may be occurring, but at low rates. More likely, AOM was occurring in the sediments and DO levels at that depth were high enough for methane oxidation to be carried out by microaerophilic methanotrophs (Ren et al., 1997; van Bodegom et al., 2001).

The two most abundant aerobic methanotrophic OTUs detected were classified as Methylobacter and within the same family Methylococcaceae. Both of these aerobic methanotrophs were highly abundant below the oxycline (>20%) and made up the majority of the methanotrophic community in that part of the water column. The high relative abundances of aerobic methanotrophs and that DO was <2% below the oxycline suggest that these methanotrophs are microaerophilic (Ren et al., 1997). The OTU classified within Methylobacter was also abundant in the upper water column above the oxycline suggesting that CH<sup>4</sup> was being oxidized efficiently throughout the water column under ice cover. In other boreal and polar lakes, the flux of CH<sup>4</sup> during ice-melt associated with spring overturn has been shown to be a significant input of CH<sup>4</sup> into the atmosphere in the spring (Phelps et al., 1998; Juutinen et al., 2009; López Bellido et al., 2009; Song et al., 2012). The decrease in CH<sup>4</sup> concentration up the water column in Potentilla lake may be attributed to consumption by both AOM and microaerophilic methanotrophs, thereby mitigating the potential emission of CH<sup>4</sup> to the atmosphere during ice-melt and spring overturn.

Potential methanogens were most abundant below the oxycline but their overall relative abundance was very low throughout the water column. Quality and quantity of nutrients could potentially be low at the end of the ice-covered season because ice cover decreases input of nutrients from the terrestrial surroundings and atmosphere (Bertilsson et al., 2013). More likely, methanogenesis occurs in lake sediments where more favorable electron acceptors are absent (Bertilsson et al., 2013). This would suggest that the source of CH<sup>4</sup> observed in the water column originated in the sediments, and that the CH<sup>4</sup> in the lower water column was oxidized once oxygen levels were sufficient for methane oxidation or by AOM. The types of methanogens identified, such as Methanosaeta and Methanomicrobiales, are similar to methanogens detected in sediments of a seasonally ice-covered, dimictic, oligotrophic lake in Germany (Conrad et al., 2007) but differ from the ones in perennially frozen, meromictic, Antarctic Lake Fryxell (Karr et al., 2006). Lake Fryxell is a highly sulfidic and stratified lake compared to Potentilla lake, suggesting that methanogenic communities in Potentilla lake are determined by a set of factors including duration and amount of snow and ice cover, amount of nutrients available in addition to polar climate.

# Ammonia Oxidation and Denitrification

The relative abundance of potential ammonia oxidizers was stable throughout the water column, but potential nitrite oxidizers increased in abundance below the oxycline. Ammonium is an important nitrogen source for other microorganisms such as methane and sulfur oxidizers, and the low relative abundance of potential ammonia oxidizers compared to methanotrophs and PSB suggests that ammonia oxidizers may not be competitive for ammonium. Pouliot et al. (2009) found strong vertical

differentiation of archaeal ammonia oxidizers in an Arctic marine-influenced meromictic lake, with the largest abundances around the oxycline. The authors suggest that Archaea may be important contributors to ammonia oxidation, which is consistent with studies in other systems that showed the importance of Thaumarchaeota as ammonia oxidizers (Church et al., 2003; Ochsenreiter et al., 2003; Zhang et al., 2008, 2010; Lehtovirta et al., 2009; Dodsworth et al., 2011; Stahl and de la Torre, 2012). However, in our study OTUs classified within the Thaumarchaeota were very rare (∼0.0001%). Nitrite oxidation should be supported throughout the water column as nitrite concentrations were higher compared to other icecovered lakes (Auguet and Casamayor, 2013). Gallionella, was the most abundant potential nitrite oxidizer below the oxycline consistent with Alawi et al. (2007) who cultivated a nitrite oxidizer most closely related to the genus Gallionella. Potential nitrite oxidizers were particularly low in abundance above the oxycline suggesting that aerobic nitrite oxidizers could not compete for oxygen with aerobic ammonia oxidizers above the oxycline. However, below the oxycline, potential nitrite oxidizers increased, suggesting they were competitive for NO<sup>2</sup> − (Nakano and Zuber, 2004). The combination of high NO<sup>2</sup> − levels, increased NH<sup>4</sup> <sup>+</sup> concentrations and the presence of OTUs classified as Anammoximicrobium (0.01–0.03%, data not shown) below the oxycline suggest the potential for anaerobic ammonia oxidation. Anaerobic ammonia oxidation converts NO<sup>2</sup> <sup>−</sup> and NH<sup>4</sup> <sup>+</sup> into nitrogen gas and is responsible for up to 50% of nitrogen losses in marine environments (Kuenen, 2008). Concentrations of NH<sup>4</sup> <sup>+</sup> and relative abundances of potential ammonia and nitrite oxidizers may be currently low in Potentilla lake because of limited surface runoff. However, the amount and type of surface runoff may change as climate change alters precipitation, snowmelt, and permafrost thaw (Anderson et al., 2002; Vincent et al., 2012).

The potential denitrifiers we detected are known for complete and incomplete denitrification. For example, representatives of the group Comamonadaceae have been shown to carry out complete denitrification in activated sludge systems (Etchebehere et al., 2001; Ginige et al., 2005). This suggests that while Comamonadaceae, one of the most abundant OTUs of potential denitrifiers we found, may be using aerobic respiration in the upper water column above the oxycline, they could also be important for denitrification below the oxycline. Incomplete denitrification is known to produce nitrous oxide, a potent greenhouse gas (Fernandes et al., 2010; Abed et al., 2013). However, we detected OTUs that were classified within Alcaligenaceae, a group of denitrifiers that have been shown to contain the nosZ gene that encodes for the nitrous oxide reductase (Velusamy and Krishnani, 2013). This enzyme reduces nitrous oxide to dinitrogen (Henry et al., 2006), thus potentially resulting in complete denitrification of nitrate and reducing the buildup of nitrous oxide in Potentilla lake. The presence of NO<sup>3</sup> − inhibits NO<sup>2</sup> <sup>−</sup> reduction (Neubauer and Götz, 1996) resulting in a buildup of NO<sup>2</sup> <sup>−</sup>, which may account for NO<sup>2</sup> <sup>−</sup> levels that exceeded NO<sup>3</sup> <sup>−</sup> levels by a factor of eight. The decrease in NO<sup>3</sup> <sup>−</sup> levels below 4.0 m is reflected in the decrease in the relative abundance of potential denitrifiers. Voytek et al. (1998) found that NO<sup>3</sup> <sup>−</sup> levels were always below detection limit in a permanently frozen lake in Antarctica. Compared to this study, NO<sup>3</sup> <sup>−</sup> levels in Potentilla lake were high, which may be an explanation for the high relative abundances of potential denitrifiers. The large relative abundances of potential denitrifiers indicate that denitrifiers may be carrying out an important part of nitrogen cycling. Understanding denitrification dynamics in Arctic lakes may be particularly important as terrestrial runoff due to warming temperatures is predicted to rise in Arctic landscapes (Vincent et al., 2012).

# Microbial Response to Environmental Change

We also sampled Potentilla lake under ice cover in 2013 and found similar geochemical conditions but differences in the microbial populations. In late winter 2013, PSB composed a large proportion of the overall microbial community with one OTU identified as Lamprocystis, composing >30% of the total abundance at 6.0 m depth. In the following winter 2014, the overall abundance of PSB was lower and, while Lamprocystis was present, the most abundant PSB OTU was classified as Thiodictyon, which suggests that microbial community members may cycle between abundant and rare. The dynamic yearto-year changes in abundance are consistent with field and modeling studies of Ace Lake in Antarctica (Lauro et al., 2011).

It should be noted that for most functional groups detected in 2014, most OTUs were rare (<1%). For example, four out of 1778 OTUs identified as potential methanotrophs had a relative abundance >1%. This does not indicate that the rare OTUs were not active. Dormant and rare members in the communities can be disproportionally active (Jones and Lennon, 2010). Additionally, a large proportion of the active bacterial community can cycle between abundant and rare (Campbell et al., 2011), and rare members can become active spontaneously (Epstein, 2009) or due to environmental changes. The differences in microbial community composition observed between 2013 and 2014 may be occurring regularly in response to environmental factors such as the extent of snow pack development. In 2013, the lake had no snow cover, whereas in 2014, it was covered by ∼20 cm of snow. The associated differences in light penetration may account for the changes in PSB community composition between 2013 and 2014, illustrating the importance of light as a microbial driver for phototrophic growth in oligotrophic polar lakes (Stomp et al., 2007; Wilkins et al., 2013). Under accelerated warming in the Arctic, ice-cover thickness and duration are likely to decrease and snow-cover regimes will change. We suggest that these changes will likely drive significant variations in microbial ecologies related to differences in light penetration and thermal structure of Arctic lakes.

# CONCLUSION

We showed that microbial communities changed consistently with geochemical properties throughout the water column under

ice in Potentilla lake. The microbial community was diverse and changed markedly in composition above and below the oxycline with a surprisingly low relative abundance of phototrophs. For the first time, we showed the importance of PSB as potential sulfur oxidizers in Arctic dilute dimictic lakes under seasonal ice cover. Concentrations of CH<sup>4</sup> decreased up the water column, which we attribute to consumption by both AOM and microaerophilic methanotrophs, thereby mitigating the potential emission of CH<sup>4</sup> to the atmosphere during spring overturn and ice-melt. The high concentrations of NO<sup>3</sup> <sup>−</sup> and NO<sup>2</sup> − indicate that denitrifiers may be carrying out an important part of nitrogen cycling. Together with previous studies, we show that the microbial community dynamics under ice-cover are complex. Broad-scale surveys, involving metatranscriptomics and metaproteomics in combination with detailed geochemical analysis, are needed to identify local and regional processes and their interactions that control the microbial dynamics of these types of lakes and their response to climate change (Reuss et al., 2013).

#### AUTHOR CONTRIBUTIONS

US was involved in the experimental design of this study, advised and trained the undergraduates to determine microbial

#### REFERENCES


community composition, was heavily involved in the data analysis, and was the main writer of this manuscript. SC designed the experiment, carried out the field work, did all the geochemical analysis, and was heavily involved in data analysis and writing of the manuscript. CH did all the bioinformatic analysis of the MiSeq sequences, was involved in data analysis and writing of the manuscript. LP was involved in the experimental design, field work, data analysis, and writing of the manuscript. JW was involved in the experimental design, field work, geochemical analysis, data analysis, and writing of the manuscript.

# FUNDING

This work was supported by NASA ASTEP (NNX11AJ01G).

#### ACKNOWLEDGMENTS

We thank Kathryn Fledderman and Lauren Banina for their help with the sequencing analysis of the microbial communities, and Amy Goldman and Seth Young for assistance in the field with sample collection.


sedimentary records from several small lakes in southwestern Greenland. Org. Geochem. 82, 33–41. doi: 10.1016/j.orggeochem.2015.02.005


Nitrospira-like bacteria. Environ. Microbiol. 8, 1487–1495. doi: 10.1111/j.1462- 2920.2006.01033.x



Zhu, B., van Dijk, G., Fritz, C., Smolders, A. J., Pol, A., Jetten, M. S., et al. (2012). Anaerobic oxidization of methane in a minerotrophic peatland: enrichment of nitrite-dependent methane-oxidizing bacteria. Appl. Environ. Microbiol. 78, 8657–8665. doi: 10.1128/AEM.02102-12

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Schütte, Cadieux, Hemmerich, Pratt and White. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Microbial Community Structure and Interannual Change in the Last Epishelf Lake Ecosystem in the North Polar Region

Mary Thaler 1, 2, 3 \*, Warwick F. Vincent <sup>1</sup> , Marie Lionard1, 2, Andrew K. Hamilton<sup>4</sup> and Connie Lovejoy 1, 2, 3

<sup>1</sup> Département de Biologie, Centre d'Études Nordiques and Takuvik Joint International Laboratory, Université Laval, Québec, QC, Canada, <sup>2</sup> Institut de Biologie Intégrative et des Systèmes, Université Laval, Québec, QC, Canada, <sup>3</sup> Québec-Océan, Université Laval, Québec, QC, Canada, <sup>4</sup> Faculty of Applied Science, Civil Engineering, University of British Columbia, Vancouver, BC, Canada

#### Edited by:

Marcelino T. Suzuki, Université Pierre-et-Marie-Curie and CNRS, France

#### Reviewed by:

Sara Beier, Leibniz Institute for Baltic Sea Research (LG), Germany Antonio Quesada, Autonomous University of Madrid, Spain

> \*Correspondence: Mary Thaler mary.thaler.1@ulaval.ca

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Marine Science

Received: 05 July 2016 Accepted: 12 December 2016 Published: 10 January 2017

#### Citation:

Thaler M, Vincent WF, Lionard M, Hamilton AK and Lovejoy C (2017) Microbial Community Structure and Interannual Change in the Last Epishelf Lake Ecosystem in the North Polar Region. Front. Mar. Sci. 3:275. doi: 10.3389/fmars.2016.00275 Climate warming is proceeding rapidly in the polar regions, posing a threat to ice-dependent ecosystems. Among the most vulnerable are microbial-dominated epishelf lakes, in which surface ice-damming of an embayment causes a freshwater layer to overlie the sea, creating an interface between distinct habitats. We characterized the physicochemical and biotic environment of Milne Fiord epishelf lake (82◦ N, Canada) in three successive summers (2010–2012), and on one date of profiling (5 July 2011) we collected samples for high through-put amplicon sequencing of variable regions of small subunit rRNA to characterize the microbial community (Eukarya, Bacteria and Archaea). Potentially active water column communities were investigated using reverse-transcribed rRNA, and phytoplankton were further characterized by accessory pigment analysis. Cluster analysis of pigment data showed a demarcation between freshwater and marine communities, which was also evident in the sequence data. The halocline community of Eukarya was more similar to the deeper marine sample than to the freshwater surface community, while the Archaea and Bacteria communities at this interface clustered more with surface communities. In 2012, conductivity-depth profiles indicated shallowing of the freshwater layer and mixing across the halocline, accompanied by lower picocyanobacteria and higher picoeukaryote concentrations. Picocyanobacteria cells were more evenly distributed throughout the water column in 2012, implying partial deep mixing. Several mixotrophic taxa of Eukarya were more abundant in the freshwater layer, where low nutrient concentrations may favor this lifestyle. Unusual features of Milne Fiord microbial communities included benthic taxa not previously reported in marine water columns (notably, the archaeon Halobacteriales), and dominance by taxa that are typically present in sparse concentrations elsewhere: for example, the Chlorophyte group Radicarteria and the betaproteobacterium Rhodoferax. Milne Fiord epishelf lake is the last known lake of this kind remaining in the Arctic, and the fate of this distinct microbial ecosystem may ultimately depend on the stability of the Milne Fiord ice shelf, which has experienced a negative mass balance over the past half century.

Keywords: Arctic, Radicarteria, climate change, cyanobacteria, Halobacteriales, microbial biodiversity, picoeukaryotes, Rhodoferax

# INTRODUCTION

Epishelf lakes form in high latitude environments when an iceshelf blocks the mouth of a fiord or other embayment, allowing freshwater from precipitation and glacier run-off to accumulate atop the denser seawater (Veillette et al., 2008; Laybourn-Parry and Wadham, 2014). Numerous epishelf lakes occur in Antarctica (Gibson and Andersen, 2002; Laybourn-Parry et al., 2006), and many were once located in the Arctic, forming behind ice shelves along the northern coast of Ellesmere Island (Vincent et al., 2001). However, the break-up of these ice-shelves over the past century has caused the drainage of most of these lakes in the northern hemisphere. Most recently, the Disraeli Fiord epishelf lake disappeared in the early 2000s (Mueller et al., 2003), and the Petersen Bay epishelf lake was lost in 2005 (White et al., 2015).

The Milne Fiord epishelf lake, located behind the Milne Ice Shelf, is the last of its kind in the High Arctic. The present form of the lake was created between 1959 and 1984 by deterioration of the landward edge of the ice shelf, while the lake is bounded on its landward side by the floating tongue of Milne Glacier (Mortimer et al., 2012). There has been an ongoing warming and attrition of ice along northern Ellesmere Island over the last decade. For example, Ward Hunt Lake lost its thick perennial ice cover for possibly the first time in more than 50 years in 2011, and in 2012 was ice-free for almost 1 month (Paquette et al., 2015). These observations imply that the ice-bound epishelf lake ecosystem at Milne Fiord may also have experienced change over the same period.

The microbial diversity of Milne Fiord epishelf lake was studied in 2007 using clone libraries (Veillette et al., 2011); however, the sampling was restricted to dark incubated enrichment cultures and did not include the halocline. The study showed large differences in bacterial species, as well as in phytoplankton pigment composition between samples from the surface freshwater and deep saltwater. The zooplankton community in Milne Fiord was found to contain an unusual mix of freshwater and marine zooplankton species, as observed in the former Disraeli Fiord epishelf lake ecosystem (Van Hove et al., 2001).

Our objectives in the current study were twofold. First, we aimed to determine the interannual variability in the summer properties of the Milne Fiord lake ecosystem, including its primary producers, by way of a suite of limnological and biological measurements over 3 years. The sampling covered a period of warming along the northern Ellesmere coastline, 2010–2012, and we hypothesized that the physical, chemical and biological properties of this ecosystem would show evidence of change. Secondly, we aimed to undertake an in-depth characterization of all three domains of the microbial communities using high through-put SSU rRNA amplicon sequencing. We addressed this second objective by way of a more detailed water column sampling and molecular analysis in July 2011. We focused on RNA to provide an indication of the taxa more likely to be involved in protein synthesis (Blazewicz et al., 2013), and size-fractionated this "active community" by sequential filtration. We hypothesized that the microbial community structure of Milne Fiord epishelf lake would reflect its unique highly stratified water column structure.

# MATERIALS AND METHODS

#### Study Site and Sampling

Milne Fiord (lat. 82◦ 35.540′N; long. 80◦ 35.761′W) is located on the northern coastline of Ellesmere Island, Nunavut, Canada (**Figure 1**). It has a maximum recorded depth of 436 m, with the depth of the surface freshwater layer thought to correspond to the minimum thickness of the ice shelf. In 2009, the thickness of this epishelf freshwater layer was 14.3 m, slightly less the 16.6 m recorded 2 years previously (Veillette et al., 2011). The lake had an area of 64.4 km<sup>2</sup> in May 2012 and is typically covered by perennial freshwater ice. However, satellite imagery showed that this ice partially broke up in August 2012, at the end of the warm 2012 season. The Milne Ice Shelf itself has been in a state of negative mass balance since at least 1953 (Mortimer et al., 2012).

The epishelf lake was sampled over 3 years on 9 July in 2010, 5 July 2011, and 5 July 2012, as part of the program "Northern Ellesmere Island in the Global Environment" (NEIGE). Access was by helicopter from our field laboratory at Ward Hunt Island, located 115 km east of the fiord. Samples were collected in Neige Bay (unofficial name), a shallow inlet on the eastern shore of the fiord, which has a depth of 84 m. At the time of sampling, the average ice thickness on the epishelf lake was 84 ± 16 cm over the 3 years.

# Water Column Profiles

Physico-chemical profiles of the water column were made using a conductivity-temperature-depth profiler (XR-420 CTD-RBR profiler; RBR Ltd, Ottawa, Canada), and a Hydrolab DS5X profiler (Loveland, CO, USA). These data are archived in NEIGE (2016a). Owing to a problem with the profiler in 2012, oxygen data was instead measured using a YSI Pro Plus multiparameter handheld meter on discrete water samples collected in the main fiord, rather than in Neige Bay. These oxygen samples were collected the day before the other sampling activities occurred. Samples for chemical and biological analyses were taken using a Kemmerer bottle (Wildlife Supply Company,

Yulee, FL, USA) from various depths (1.5–30 m). Twelve liters of water were transferred into acid- and fiord water rinsed 20-l propylene Cubitainers, and kept cold and dark until transport to the field laboratory on Ward Hunt Island. The water was subsampled for nutrients, flow cytometry and high-performance liquid chromatography (HPLC) pigment analyses within 4 h.

# Chemical Analysis

Water samples for chemical analyses were collected from 1.5, 16, and 30 m depths in all years, and an additional sample at 10 m was obtained in 2010 and 2012. Aliquots of 40 ml were placed in brown bottles for total organic carbon analyses. For total phosphorus (TP) and total nitrogen (TN), 0.833 ml of H2SO<sup>4</sup> (30%) was added to 125 ml of sample water and stored in the dark at 4◦C until later analysis at the Institut National de Recherche Scientifique–Eau, Terre & Environnement, Québec, Canada (INRS-ETE). TN and dissolved organic carbon (DOC) were assessed using a Shimadzu VCPH Lachat with detection limits of 0.02 mg l−<sup>1</sup> and 0.05 mg l−<sup>1</sup> respectively. UV-VIS spectrophotometry was used for TP quantification with detection limits of 0.5 and 0.2µg l−<sup>1</sup> , respectively. Dissolved inorganic carbon (DIC) was analyzed using a CombiPal 3800 Varian with a detection limit of 0.02 mM.

# Cell Enumeration by Flow Cytometry (FCM)

In 2010, the water column was sampled at 10 depths: 1.5, 6, 8, 10, 12, 14, 16, 18, 22, and 30 m. In 2011 and 2012, an eleventh sample depth was added at 3 m. Aliquots of 4 ml were placed in Cryovials and fixed with 160 µl of glutaraldehyde 25%, Grade I, for a final concentration of 1%. While in the field, these samples were stored in a dry-shipper previously cooled with liquid nitrogen, and upon returning to the lab at −80◦C. Cell analyses were performed using a BD FACSCalibur flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA) following the method of Casamayor et al. (2007). In the present study, two fluorescence channels were used: FL3 (670 nm) for picoeukaryotes, and a combination of FL3 and FL2 (585 nm) for picocyanobacteria. FL3 and FL2 detected the natural fluorescence of chlorophyll a (Chl a) and phycobilins respectively. Beads of 1-µm diameter (Fluoresbrite, Calibration Grade; Polysciences, Inc., Warrington, PA, USA) were added to samples to assess picoeukaryote and picocyanobacteria size. Trucount beads (BD Biosciences, Franklin Lakes, NJ, USA) were used to calibrate the 1-µm bead solutions and to assess population size. To ensure quality of cell detection, a minimum of 15,000 events were measured per analysis, with less than 1000 events s−<sup>1</sup> . Each sample was analyzed in triplicate.

# Pigment Analysis

In 2010, samples for pigment analysis by HPLC were collected from the same four depths as for nutrient analyses, while in 2011 and 2012 they were collected from the same 11 depths as for FCM analyses. We filtered 0.3–2 l of water through 25 mm diameter Whatman GF/F glass fiber filters. These were then wrapped in aluminum foil and immediately placed into a dry shipper (previously cooled with liquid nitrogen) in the field, then transferred to −80◦C for storage until pigment analysis at Laval University. Within 1 month of collection, the filter samples were extracted in the dark by sonication treatment for 30 s at 17 W in 3 ml of 95% methanol (by volume). The supernatant was recovered after vortexing and centrifugation (4150 rpm for 15 min at 4◦C), filtered through a 0.2-µm pore size Polytetrafluoroethylene (PTFE) Acrodisc filter (PALL Corporation, Ann Arbor, MI, USA) and placed in a 2-ml amber vial with an argon gas atmosphere for immediate analysis by HPLC.

HPLC analyses in 2010 were performed using a ProStar HPLC system (Varian, Palo Alto, CA, USA) with a Symmetry C8 column (3.5-µm pore size, 4.6 × 150 mm; Waters Corporation, Milford, MA, USA) at 25◦C, and a C8 guard column (5-µm pore size, 3.9 × 20 mm; Waters Corporation). In 2011 and 2012, HPLC analyses were performed using a Thermo Scientific system (Thermo Scientific, West Palm Beach, FL, USA) with a Hypersil Gold C8 HPLC column (3.0-µm pore size, 4.6 x 150 mm, Thermo Scientific) at 25◦C, with a C8 guard column. Carotenoids were quantified by their absorbance at 450 nm in a Diode Array Detector (350–750 nm) and chlorophylls were detected by fluorescence (excitation at 400 nm and emission at 650 nm). The HPLC separation method followed Zapata et al. (2000). Pigments were identified based on retention time and spectral comparisons with standards from DHI (Water & Environments, Horsholm, Denmark). The HPLC data are archived in NEIGE (2016b). Clustering analysis was performed using Bray-Curtis distance, as implemented in the R package vegan (Oksanen et al., 2010). Pigments were interpreted as indicators of specific phytoplankton phyla following Wright and Jeffrey (2006).

#### RNA Collection and Extraction

Water samples for RNA extractions were collected from Milne Fiord epishelf lake (82◦ 35.479 N; 80◦ 35.824 W) on 5 July 2011 using a Kemmerer bottle at three depths: 1.5, 16, and 30 m. Samples were transported in 20-l polyethylene containers in darkness, and processed within 4 h. Four liters of water were filtered sequentially through a 47-mm diameter 3-µm pore size polycarbonate filter and a 0.22-µm Sterivex filter cartridge (Millipore, Etobicoke, ON, Canada). For the 1.5-m sample, the 3-µm filters were replaced as they became clogged with material, for a total of four filters, which were stored in a single 1.5-ml tube. The Sterivex cartridge for this sample was replaced after 2 l, and the two cartridges were preserved separately. Samples were preserved by adding 1 ml RLT buffer (Qiagen, Germantown, MD, USA), and frozen in liquid nitrogen for transport. Samples were stored at −80◦C.

RNA was extracted from filters and Sterivex cartridges using a Qiagen AllPrep DNA/RNA Mini Kit. For the 1.5-m sample, all four 3-µm filters were used for the large size fraction, while only the first Sterivex cartridge was used for the small size fraction. The RLT buffer was removed before extraction and replaced with RLT-plus from the extraction kit. Cells were lysed by an initial incubation with 3 µg ml−<sup>1</sup> lysozyme (Sigma-Aldrich, St. Louis, MO, USA) at 37◦C for 45 min, and 0.78 mg ml−<sup>1</sup> proteinase K (Bio Basic, Markham, ON, Canada) with 1.11% SDS at 65◦C for 15 min. The lysate was then homogenized using the QiaShredder columns included with the extraction kit. An incubation with DNAse I (0.23 units µl −1 ) for 15 min at room temperature in RDD buffer (Qiagen, Hilden, Germany), was added between wash steps. To increase the final concentration of RNA, 35 µl of elution buffer was passed twice through the RNeasy column at the last step. The absence of DNA was immediately confirmed by a PCR of the 16S rRNA gene on the extracted RNA using the primers 8F and 1492R (Edwards et al., 1989; Stackebrandt and Liesack, 1993) in a 25-µl reaction volume containing High-Fidelity buffer 1X, 0.2µM of each primer, 0.2 µM dNTPs, 0.2 mg ml−<sup>1</sup> bovine serum albumin (BSA), and 1 U of Phusion High-Fidelity DNA polymerase (all reagents from New England Biolabs, Ipswich, MA, USA), using a 16S amplicon from a bacterial clone library as a positive control. The thermal cycling protocol was an initial denaturation at 98◦C for 30 s, followed by 30 cycles of denaturation at 98◦C for 10 s, annealing at 55◦C for 30 s, and extension at 72◦C for 30 s, with a final extension of 72◦C for 4.5 min. PCR products were checked on a 1% agarose gel stained with 4% v/v ethidium bromide. No band was observed on the gel. Extracted RNA was converted to cDNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA).

#### High Throughput Multiplex Tag Sequencing

The V4 region of the eukaryotic 18S rRNA gene, and the V6- V8 and V3-V5 regions of the respective bacterial and archaeal 16S rRNA genes were amplified using the primers described in Comeau et al. (2011). PCR was carried out as above, but in a total volume of 20µl and with BSA omitted. Three reactions were performed for each sample, using 0.5, 1, and 2µl of template cDNA. For eukaryotes, the thermal cycling protocol was as described above. For Bacteria, the protocol was as described above but with 25 cycles instead of 30 and a final extension step of 7 min. For Archaea, the protocol was also as described above but with an annealing temperature of 52◦C, 35 cycles, and a final extension step of 5 min. PCR products from the three different concentrations of template cDNA were pooled and purified by use of magnetic beads (Axygen, Corning, New York, NY, USA). PCR product bound to magnetic beads was washed twice with 80% ethanol and eluted in 10 mM Tris buffer (pH 8.13). A second PCR used primers provided by Illumina (San Diego, CA, USA) to add barcodes to the PCR products. Each sample received a unique combination of forward and reverse primers. The purified DNA template was diluted 10X, and concentrations of reagents and thermocycler protocol were the same as in the first PCR, except that the thermocycler was run for only 15 cycles. Two reaction volumes of 25µl each were run for each sample and then pooled before PCR products were again purified using magnetic beads. The DNA concentration was measured spectrophotometrically using a Nanodrop 1000 (Nanodrop, Wilmington, DE, USA). Samples were mixed equimolarly, along with barcoded samples from other projects, and the amplicons were then paired-end sequenced on an Illumina MiSeq platform at the IBIS/Laval University Plate-forme d'Analyses Génomiques (Québec, QC, Canada). Raw reads have been deposited in the NCBI database under accession number SRP076684. An initial run was truncated because of an error in reagent volumes. The sequences from this run were analyzed together with the sequences from a completed run, resulting in 16% more total reads.

#### Sequence Processing

Sequences were analyzed using the UPARSE pipeline (Edgar, 2013) as implemented in QIIME (Caporaso et al., 2010). Reads with phred quality scores <20 were discarded. Chimeras were detected both de novo and by reference to the SILVA database (Pruesse et al., 2007) and the intersection of good sequences from these two methods were retained, before clustering into Operational Taxonomic Units (OTUs) at 97% similarity (98% for eukaryotes). Singleton OTUs were discarded. OTUs were assigned a taxonomic identity using the Mothur classifier (Schloss et al., 2009) with a 0.8 confidence threshold using an in-house curated taxonomic database (Lovejoy et al., 2016). Very abundant OTUs were manually checked for chimeras by performing separate BLAST searches on the front and back halves of the sequence, and a few abundant OTUs that could not be identified with Mothur were successfully identified using a BLAST search.

The sequences amplified with Archaea-specific primers were found to contain a number of sequences from the other two domains even after processing. These were detected and removed using the Ribosomal Database Project Classifier (Wang et al., 2007). An Archaea reference tree was constructed using an alignment of 73 16S rRNA sequences obtained from GenBank and aligned using MUSCLE (Edgar, 2004). The alignment had 1496 characters. A maximum-likelihood tree was constructed using RAxML v7.2.7 (Stamatakis, 2006), with 100 bootstraps, and rooted using Korarchaeum cryptofilum. Short archaeal reads were mapped onto the reference tree using the Evolutionary Placement Algorithm (EPA) of RAxML (Berger et al., 2011). Only read placements with likelihood weight of >0.5 were retained.

The phylogenetic distance among communities from different depths and size fractions was analyzed for each of the three domains of life using the weighted Unifrac beta diversity metric (Lozupone and Knight, 2005) as implemented in Mothur.

# RESULTS

#### Air Temperature and Water Column Structure

Meteorological records from nearby Ward Hunt Island beginning in 1996 have shown that regional summer air temperatures were warmer during 2010–2012 in comparison to most of the preceding decade (Paquette et al., 2015; CEN, 2016). Air temperature records from an automated weather station in Milne Fiord are shorter, beginning in 2009, but reveal that 2011 and 2012 were particularly warm summers. For the period 2009–2015, the mean (standard deviation) number of melting degree days was 203 (76), compared to 185, 278, and 253 in 2010, 2011, and 2012, respectively. The maximum hourly air temperature for 2009–2015 was 19.9◦C, occurring in July 2012.

Water column profiles showed a several meter thick relatively warm (up to 1.5◦C) freshwater layer above colder (<−1.5◦C) seawater in all years (**Figure 2**). Salinity increased from about 0.1 at the surface to >30 below 25 m depth, with a sharp salinity gradient at the base of the freshwater layer. The depth of the halocline, defined as the level of the maximum salinity gradient, varied from approximately 15 m in 2010, to 14 m in 2011, and 11 m in 2012. In each year, the depth of the thermocline and oxycline coincided with the halocline, with the warm freshwater layer relatively well-oxygenated compared to that of seawater below. The nutricline also appeared to coincide with the halocline, with the surface freshwater depleted in both nitrogen and phosphorus each year (**Figure 3**).

Water properties above 23 m depth in Neige Bay were similar to measured water properties in the main fiord (not shown), while the properties of deep water in Neige Bay were distinct, and nearly homogenous with depth to the bottom. These

FIGURE 2 | Physical profiles of Milne Fiord epishelf lake over 3 years (2010–2012) showing salinity, temperature, oxygen concentration, and chlorophyll a concentration. Note the fewer depths for chlorophyll a in 2010 (dashed line). Oxygen data in 2012 are for the main fiord; all other profiles are from Neige Bay.

observations suggested the existence of a 23 m-deep sill at the mouth of Neige Bay. The 2012 oxygen profile was collected in the main fiord and showed higher oxygen concentrations below 25 m compared to the other years when profiles were collected in Neige Bay. This suggests that the deep water in Neige Bay was slightly depleted in oxygen relative to the main fiord, but oxygen concentrations were >7 mg l−<sup>1</sup> at all depths suggesting that the water was likely renewed annually or even more frequently.

#### Pigment Composition

Maximum Chl a (**Figure 2**) was observed in the surface freshwater layer and reached 0.72µg l−<sup>1</sup> at 10 m in 2011, and 2.53µg l−<sup>1</sup> at 6 m in 2012. In 2010, a peak was not observed in the upper layer, perhaps because only one sample was taken in this layer, at 1.5 m. In all years, zeaxanthin, likely associated with picocyanobacteria, was observed in the upper water column down to 22 m (**Figure 4**). Violaxanthin, chlorophyll b, alloxanthin and lutein were found as important pigments in the freshwater surface layer, indicating the presence of chlorophytes and cryptophytes. In 2011, the presence of chlorophyll c2, peridinin and diadinoxanthin just beneath the ice suggested the presence of dinoflagellates. In the deeper seawater layer, fucoxanthin and chlorophyll c<sup>3</sup> were observed, indicating the presence of either diatoms or prymnesiophytes.

#### Flow Cytometry

Picocyanobacteria and Chl a-containing picoeukaryotes were mainly detected in the surface freshwater layer (**Figure 5**). Maximal abundance of picocyanobacteria was observed at different depths across years of sampling: at 8 m in 2010 with 2.1 × 10<sup>4</sup> cell ml−<sup>1</sup> , and at 14 m in 2011 with 2.8 × 10<sup>3</sup> cells ml−<sup>1</sup> . In 2012, no maximum was observed and the abundance was in general lower, ranging from 0.3 to 1 × 10<sup>3</sup> cells ml−<sup>1</sup> down the water column. The picoeukaryotes were more evenly distributed through the water column, with minor peaks at 18 (2010), 16 (2011), and 14 m (2012). The average abundance in the freshwater layer was 3.4 (±0.8) × 10<sup>3</sup> cells ml−<sup>1</sup> in 2010 and 2.7 (±0.5) × 10<sup>3</sup> cells ml−<sup>1</sup> in 2011, rising to 6.6 (±1.9) × 10<sup>3</sup> cells ml−<sup>1</sup> in 2012.

# High Throughput Sequencing (HTS)

Approximately 925,000 reads were retained after processing for quality control (**Table 1**), consisting of a total of 3575 microbial OTUs. In all domains, OTU richness was lowest in freshwater sample immediately beneath the ice (1.5 m). This richness was similar in both size fractions, with the exception of 30 m, where Bacteria had a greater OTU richness in the >3µm size fraction, pointing to a greater diversity of particle-attached bacteria at this depth. Archaeal sequences could not be identified below phylum level using sequence similarity, but with EPA approximately 70% of sequences could be placed on a reference tree at class level or below (**Figure 6**; Figure S1B). Most of the sequences whose placement fell below the Maximum Likelihood cut-off of 0.5

FIGURE 4 | Mass ratios of photosynthetic pigments relative to chlorophyll a in the water column of Milne Fiord epishelf lake over 3 years (2010–2012), determined by HPLC analysis. Note the fewer sample depths in 2010 (dashed line). Fuco, fucoxanthin; Viol, violaxanthin; Lut, lutein; Zea, zeaxanthin; Chl, chlorophyll; Allo, alloxanthin; Diadin, diadinoxanthin; Perid, peridinin.

TABLE 1 | Number of sequences and Operational Taxonomic Units (OTUs) for samples from Milne Fiord epishelf lake in July 2011, according to depth and size fraction, after quality screening.


appeared related to Marine Benthic Group B (at 30 m) and Group I.1a (at the surface), both members of Thaumarchaeota.

Samples clustered based on their phylogenetic distance by depth rather than size fraction for all three domains of life, and surface samples were the most distinct from other depths (**Figure 7**). The surface community of Eukarya separated from the 16 and 30-m communities, while for the Bacteria and Archaea the 30-m sample separated from the cluster of surface and 16-m communities.

With the exception of the 30-m sample, the bacterial community was dominated by the phylum Proteobacteria (**Figure 8**). A few taxa, such as Flavobacteria and the Roseobacter-group were more abundant in the >3 µm (particle-attached) fraction, while Sphingobacteria and Polaromonas were more abundant in the <3µm (freeliving) fraction, but the majority of taxa showed no distinction by size. Some bacterial taxa showed no clear pattern with depth when considered at a high taxonomic level (for example Gammaproteobacteria, which constituted 5–15% of reads in all samples), but evidence of vertical structure was observed when Unifrac distances were applied at the OTU level. The bacterial community at the surface was strongly dominated by a single betaproteobacterium OTU (45%) which was 99.5% identical to the corresponding V4 region of Rhodoferax saidenbachensis. At the halocline, there were noticeable peaks of the betaproteobacterium genus Polaromonas (17%) and the alphaproteobacterial Roseobacter-group (33% in the >3 µm size fraction), while several non-Proteobacteria taxa appeared restricted to 30-m samples, including Acidobacteria, Chloroflexi, Planctomycetes, and the environmental clade PAUC34f.

The dominant microbial eukaryotes were the Chlorophyte group Radicarteria, dinoflagellates, and, in the freshwater layer, a species of cryptophyte (**Figure 9**). Fungi were not detectable in our sequences. Although the surface had the lowest OTU richness, it had the greatest evenness of high-level eukaryote groups, including the cryptophyte Teleaulax gracilis (23%), chrysophytes (3.5% of reads in surface samples), and the rhizarian Protaspis (4.7%) (**Figure 9A**). Ciliates at the surface included a higher proportion of Litostomatea (3.4%), and dinoflagellates included the photosynthetic genus Woloszynskia (4.0%). The community composition of photosynthetic taxa agreed well with pigment data from HPLC. All of these

surface-associated taxa diminished markedly in relative abundance at greater depth, or disappeared entirely. In contrast, Radicarteria comprised a high proportion of reads at all depths, though it reached its greatest dominance at the halocline (57%; **Figure 9A**, Figure S1A), while a Unifrac analysis of this taxon did not separate the 16 and 30-m depths (analysis not shown). The distribution of several alveolate taxa showed vertical structure (**Figure 9B**), with the chloroplast-bearing dinoflagellate Prorocentrales becoming dominant at 16 and 30 m, while the genus Katodinium increased to 2.9% in the <3 µm fraction at 30 m. Marine Alveolates, mainly the parasitic MALV I, appeared restricted to the 30-m sample (4.1%).

Although there was a large number of unclassified reads in Archaea, some structuring by depth was apparent. The

fractions, Large (>3 µm) and Small (<3 µm).

Archaea environmental clade Pendant-33 was only detected in the freshwater layer (3.8% in the small size fraction), while Halobacteriales was restricted to 30 m. At all depths, the most dominant taxon by far was Group I.1a Thaumarchaeota (Figure S1A).

# DISCUSSION General Features of the Milne Fiord Epishelf Lake Ecosystem

Previous studies on Milne Fiord epishelf lake (Veillette et al., 2008, 2011) have drawn attention to the highly stratified nature of this ecosystem, with freshwater overlying saltwater and a vertical separation of freshwater and marine biota, as found in Antarctic epishelf lakes (Laybourn-Parry and Wadham, 2014). Our observations confirmed this vertical structure, with a strong halocline separating the upper layer, derived from snow and glacial meltwater, from the water below which was derived from the Arctic Ocean. Similarly, the HPLC results showed a separation between pigmented organisms above and below the halocline (**Figure 10**), although this transition was less sharply defined in the FCM data (**Figure 5**). The RNA sequences showed some taxa present at all depths, but in general there was strong separation between the upper and lower communities (**Figures 7**–**9**). Particularly for the eukaryotes, this separation was less evident at higher taxonomic levels (**Figure 8A**) than with OTU-level analysis (**Figure 7**), highlighting the importance of analyzing communities at the appropriate taxonomic scale.

# Microbial Community Composition

Identifying the taxonomic affiliation of short reads generated by HTS requires comprehensive, curated taxonomic databases. While many excellent resources exist, such as SILVA and the Ribosomal Database Project (Pruesse et al., 2007; Cole et al., 2014), these large databases cannot keep pace with the current rapid rate of phylogenetic revision. To improve the taxonomic resolution of our reads, we developed and applied a curated database for Eukarya and Bacteria (Lovejoy et al., 2016), following the approach in Comeau et al. (2011) and incorporating the published taxonomic revisions for abundant groups in our northern lake and ocean datasets, including Milne Fiord epishelf lake. This tailored database allowed a high degree

of taxonomic resolution for some groups, as far as specieslevel. On the other hand, the lack of reference sequences for Archaea made taxonomic identification by sequence similarity ineffective. We therefore used EPA as described in Thaler and Lovejoy (2015) to identify short reads by mapping them onto a reference phylogenetic tree (**Figure 6**). Since 16S rRNA gene analysis is known to be poor at resolving high-level taxonomy in Archaea, particularly in distinguishing Crenarchaeota and Thaumarchaeota (Brochier-Armanet et al., 2008), future environmental studies may choose to target other marker genes.

The microbial community in the Milne Fiord epishelf lake identified by HTS in 2011 was in broad agreement with FCM and HPLC results. In the same year, phytoplankton in the layer beneath the lake ice were dominated by a combination of the microplanktonic chlorophytes in the Radicarteria group that have a reported cell diameter of 30µm (Balzano et al., 2012) and the nanoplanktonic cryptophyte Teleaulax gracilis, that has a reported cell diameter of 7–12µm (Laza-Martínez et al., 2012; **Figure 9**). The dominance of these larger cells may account for the lower concentrations of pigmented picoeukaryotes detected by FCM in 2011 relative to other years (**Figure 5**). In 2009, a high proportion of prasinophytes was detected at the halocline from HPLC (Veillette et al., 2011), likely Micromonas sp. (Lovejoy et al., 2007); in contrast, this group was among the least abundant phytoplankton taxa in 2011. HTS did not indicate any particular taxon that could have accounted for the fucoxanthin and chlorophyll c<sup>3</sup> detected at 30 m (**Figure 4**), although prymnesiophytes (2.7% of reads) seem more likely than diatoms (<1% of reads). One discrepancy between the approaches was the Cyanobacteria, which appeared as high as 3000 cells ml−<sup>1</sup> in FCM, but were always <1% of sequences. This was likely because heterotrophic bacteria were present at a much higher concentration than cyanobacteria, competing for PCR primers. Overall, FCM, HPLC, and HTS provided complementary views of the communities.

Clustering analysis showed that pigment composition grouped primarily according to salinity of the sample, and to a lesser extent by sampling year (**Figure 10**), suggesting that different photosynthetic communities occurred above and below the halocline. However, at least one abundant taxon, the eukaryote Radicarteria, was abundant at all depths. Samples were clustered by depths based on the abundance and phylogenetic distances between different rRNA sequences, in contrast to other studies which found that community composition of particleattached and free-living bacteria correlated with different environmental variables (Dupont et al., 2014). This may be because, in the Milne Fiord epishelf lake, most physico-chemical variables are so strongly correlated with depth (**Figure 2**) that their differing effects cannot be disentangled.

For Bacteria, the overall dominance of Betaproteobacteria is indicative of a freshwater influence, as found elsewhere in the Arctic (Galand et al., 2008b). We found Betaproteobacteria at a lower proportion in deeper, more saline waters (**Figure 8**). Finer taxonomic identification indicated the presence of both marine and freshwater groups in the halocline, with peaks in the relative abundance of the typically-marine alphaproteobacterial Roseobacter-group (Pujalte et al., 2014) and the sometimes freshwater or ice-associated betaproteobacterium Polaromonas (Darcy et al., 2011). This latter taxon has also been reported from the water column beneath Antarctic sea-ice (Irgens et al., 1996), and it dominated the 5-m enrichment sample from Milne Fiord epishelf lake in 2007 (Veillette et al., 2011). In addition to marine and freshwater influences, the lake also receives glacial inputs. To date, there have been few HTS studies of glacier microbial communities. One study of cryoconite holes in Svalbard (Edwards et al., 2014) found similar groups at a high taxonomic levels, but a detailed metaanalysis would be needed to determine whether glacial meltwater may have had an influence on the Milne Fiord epishelf lake community.

Two previous HTS studies of freshwater environments on northern Ellesmere Island have reported Fungi, although at relative abundances of <1% (Comeau et al., 2016). Their absence from our sequencing data indicates that their concentration in Milne Fiord Epishelf Lake must have been quite low. The MALV I found in the deep layer (**Figure 9B**) belonged to the Radiolarian-associated clade RAS1 identified by Bråte et al. (2012). Radiolarians, though found in many mesopelagic water columns, were very rare in our 30 m samples (<0.001%). However, these authors found that transitions between hosts may be common in the evolutionary history of this group, so these MALV could conceivably be infecting other protist groups in the seawater layer below the lake.

Mixotrophy may be a key survival strategy for many microbial flagellates that are at the base of food webs in nutrientpoor Arctic lakes (Charvet et al., 2014), and many of the taxa found in the freshwater layer of the lake were potential phagotrophic mixotrophs. For example, one species of the chloroplast-bearing woloszynskioid dinoflagellates is known to feed on nanoplanktonic algal prey (Kang et al., 2011), and has previously been reported to form a bloom in the freshwater that accumulates beneath sea-ice in the Baltic Sea (Spilling, 2007). Unrein et al. (2014) detected bacterivory in cryptophytes in a mixed natural community that included the genus Teleaulax. Finally, one of the ciliate groups associated with the surface, Litostomatea, includes species which can acquire phototrophy through a cryptophyte endosymbiont or through kleptoplasty of cryptophyte plastids (Stoecker et al., 2009). Given the relative abundance of available cryptophyte prey immediately below the ice, it may be speculated that the litostomate ciliates in the surface of the lake might use this strategy; however, microscopy combined with fluorescent in situ hybridization would be need to confirm this conjecture. Dinoflagellate mixotrophs, in the class Prorocentrales (Stoecker, 1999), also comprised a high proportion of eukaryote reads at 16 and 30 m (**Figure 9B**).

Many of the dominant taxa in Milne Fiord epishelf lake have previously been reported from Arctic water columns, and some are known to be ice-associated, such as the eukaryotes Teleaulax (Piwosz et al., 2013) and Protaspis (Werner et al., 2007), the bacterium Polaromonas (Irgens et al., 1996), and the archaeon Halobacteriales (Collins et al., 2010). Teleaulax gracilis has only been described recently, with accompanying revision of the genus (Laza-Martínez et al., 2012), and therefore there are no historic records for this species. However, the related species T. acuta, has been reported to form a bloom in an Atlantic-influenced fiord in Svalbard (Kubiszyn et al., 2014). Similarly, among the Archaea, Thaumarchaeota are generally thought of as marine in Arctic estuarine systems (Galand et al., 2008a), and Group I.1a has been reported as the dominant archaeon in first-year sea-ice (Collins et al., 2010). However, Group I.1a is also reported from diverse freshwater environments worldwide (Bomberg et al., 2008), and Galand et al. (2008b) identified a freshwater clade within the group. The increase in relative abundance of this group at 16-m depth (Figure S1B) suggests they were primarily marine, though able to tolerate euryhaline conditions. Since Veillette et al. (2011) found that Group I.1a also dominated archaea in the enrichments from Milne Fiord epishelf lake in 2007, they may be a consistent feature of this community.

Other taxa identified in Milne Fiord epishelf lake have either not been previously reported from the Arctic, or have never been observed as dominant in a fiord environment. For example, the Radicarteria clade, which includes Carteria radiosa and Carteria obtusa from the polyphyletic chlorophyte genus Carteria (Nakada et al., 2008), dominated the eukaryote community at all depths and size fractions (**Figure 9A**). Historically, these two species have mainly been reported by researchers in Russia and the Ukraine, perhaps because of a greater familiarity with their taxonomic descriptions in those countries. Nearly all microscopic records are from terrestrial, mainly freshwater systems (Compère, 1966; Khondker et al., 2007; Tsarenko, 2008; Demchenko, 2011; Hindáková and Hindák, 2012; Gorokhov, 2013; Tarasova et al., 2013), but also one hypersaline lake (Gheorghievici et al., 2015). However, two marine strains have been brought into culture, both from the Arctic: one from the ice algae community in Baffin Bay (Bigelow National Center for Marine Algae and Microbiota, CCMP1189), and one from the deep chlorophyll maximum of the Beaufort Sea (Balzano et al., 2012). The identities of both these cultures were confirmed by phylogenetic analysis relating them to cultured freshwater strains of C. radiosa and C. obtusa. This suggests that members of this clade are euryhaline, or have diversified to fill different ecological niches, similar to the dinoflagellate Polarella (Rengefors et al., 2015) or the Peridinium-Scrippsiella complex, which is found in both seawater and cold water lakes (Logares et al., 2009). The dominance of this eukaryote phytoplankton in Milne Fiord epishelf lake is in distinct contrast to other aquatic Arctic communities studied by HTS. For example, in meromictic lakes located near Milne Fiord, the dominant phytoplankton taxa were chrysophytes (Charvet et al., 2014), while coastal marine waters in the Arctic are dominated by haptophytes, diatoms, and picochlorophytes, depending on the season (Pedrós-Alió et al., 2015).

Protaspis, the most abundant genus among OTUs of the Phylum Rhizaria, has been detected in Arctic and Baltic seaice (Werner et al., 2007; Majaneva et al., 2012), but its presence here in the water column may be unusual. The only planktonic record to date is a parasitoid of diatoms, Protaspis longipes (=Cryothecomonas longipes; Hoppenrath and Leander, 2006); however, the low frequency of diatom sequences in this layer (<1%) suggests that these Protaspis were either bacterivorous or parasitic on non-diatom taxa.

The betaproteobacterium OTU that dominated the surface water of the lake (**Figure 8B**) had closest sequence similarity to R. saidenbachensis, a non-photosynthetic, strictly aerobic, psychrotolerant species which has been described from the sediments of a temperate lake (Kaden et al., 2014). Based on related environmental sequences, these researchers speculated that it has an affinity for low-nutrient environments, which would be the case for Milne Fiord epishelf lake. The genus has also been recovered in a clone library from sediment in the perennial ice of an Antarctic lake (Gordon et al., 2000). However, our phylogenetic analysis of the whole 16S rRNA gene using reference sequences obtained from NCBI GenBank suggested there is a high level of undescribed diversity in the genus Rhodoferax (Figure S2), which may conceal a large amount of phenotypic variability. Zwart et al. (2002) describes an environmental clade they call "Rhodoferax sp. Bal47," which they report as widespread in freshwater, estuarine and coastal habitats worldwide, and which was also detected in the Mackenzie River outflow in the Arctic (Galand et al., 2008b). However, our analysis shows that their clade was misidentified and actually corresponds to the wellstudied freshwater genus Limnohabitans (Figure S2), which occurred at very low relative abundance in our samples (<0.001% in the >3µm fraction only). The present study may therefore be the first to demonstrate that Rhodoferax can be a dominant component of a bacterial community in a freshwater environment.

The Archaea order Halobacteriales was originally believed to be obligately halophilic, but they have since been detected in tidal sediments of low to moderate salinity (Purdy et al., 2004). Members of this taxonomic order have also been detected by T-RFLP fingerprinting in first-year Arctic sea ice, under conditions of extreme salinity many times that of seawater (Collins et al., 2010). To our knowledge, ours is the first study that shows them to be a major fraction of the archaeal community in a marine water column.

#### Ecosystem Effects of Deteriorating Milne Ice Shelf

The multi-annual time series presented in this study is for only 3 years of study, reflecting the great difficulty of accessing this remote site in the extreme high Arctic. However, in the context of other Arctic epishelf lakes that have experienced abrupt, non-linear changes right before total break-up (Mueller et al., 2003; White et al., 2015), the shorter time-scale may nevertheless be sufficient to capture regime-changing events. The freshwater layer of the Milne Fiord epishelf lake had a recorded thickness of 17.5 m in 1983, and similar values were measured in 2004 (Veillette et al., 2008). However, in 2007 the halocline was at 16.6 m, rising further to 14.3 m by 2009 (Veillette et al., 2011). Our observations in the present study indicated a similar halocline depth in 2010 and 2011, shoaling to 14 m (**Figure 2**), but the chemical data imply some degree of mixing between these 2 years, with a pronounced increase in TP and DIC at 16 m, accompanied by a drop in DOC (**Figure 3**). The presence of the most abundant taxa in the RNA sequence data at all sampling depths in 2011 is also consistent with halotolerant species in a partially mixed layer.

Much greater physico-chemical changes were observed in 2012, by which time the halocline had risen to 11 m. The salinity profile was particularly striking in that the halocline had lost the extreme sharpness of transition from freshwater to saltwater observed in previous years, and there was a pronounced increase in TP and decrease in DOC at 10 m (**Figure 3**), possibly as a result of this mixing. In part these changes may relate to the long term thinning of the ice shelf that is evident over timescales of decades. However, short term variations have also been observed, including strong seasonal changes in the thickness of the lake (by several meters), and episodic mixing events in the fiord that can change the depth of the lake, including in the winter of 2011–2012 (Hamilton, A.K. et al., unpublished data).

The biological observations from 2012 showed marked differences relative to the previous years. The picocyanobacteria declined by an order of magnitude, and were distributed throughout the water column, consistent with mixing, and the picoeukaryotes showed a more than twofold increase (**Figure 5**). Chl a concentrations were higher in 2012 than 2011, while oxygen concentrations also indicate potentially higher net photosynthetic rates in 2012 than in either of the two previous years (**Figure 2**). There was a rise in the ratio of some accessory pigments per unit Chl a, notably Chl b (**Figure 4**), implying an increased proportional abundance of green algae. In part these changes may be the result of year-to-year variability, but the trend of decreasing freshwater layer thickness over the period 2004–2012 and the weakening of the density gradient at the halocline implies that this ice-dammed ecosystem is in a period of transition associated with the meltdown of its ice shelf. If the Milne Ice Shelf thins further and fractures, the freshwater layer would completely drain from the lake, as observed in Disraeli Fiord (Mueller et al., 2003), and the microbial community would shift to an exclusively marine composition. Although the ultimate ecosystem of Milne Fiord remains uncertain, it may come to resemble the stratified, flagellate-dominated fiords observed in Svalbard by Carrier (2016).

#### CONCLUSIONS

The epishelf lake in Milne Fiord has many features that appear unique among Arctic aquatic ecosystems, including a combination of marine and freshwater taxa, the predominance of the eukaryotic phytoplankton Radicarteria, and a number of other taxa rarely reported from marine water columns. Our multi-annual observations cover a limited period, but they imply that 2010–2012 corresponded to a period of transition in terms of community structure as inferred from pigments, as well as in the physical and chemical environment. If the ice cover of the lake continues to break up seasonally, increased light and the mixing of nutrients and salt into the nutrient-depleted surface freshwaters will likely have further effects on microbial community structure.

# AUTHOR CONTRIBUTIONS

WV and ML designed the study; ML, WV, and AH undertook the field sampling and profiling; ML and MT performed laboratory analyses; MT and CL analyzed and interpreted sequence data;

# REFERENCES


MT and ML led the drafting of the manuscript, and all authors contributed to and have approved the final version.

# FUNDING

This research was funded by the Natural Sciences and Engineering Research Council of Canada (NSERC); the Networks of Centres of Excellence program ArcticNet; the Canada Research Chair program; the Northern Scientific Training Program; the Canadian Foundation for Innovation; the Canadian Northern Studies Trust; Centre d'études nordiques (CEN); Fonds de Recherche du Québec-Nature et Technologie (FRQNT); and the Association of Canadian Universities for Northern Studies (ACUNS).

#### ACKNOWLEDGMENTS

We thank Sébastian Bourget, Sophie Charvet, and Denis Sarrazin for assistance in the field and Pierre Galand for taxonomic advice. Logistical support was provided by the Polar Continental Shelf Program (PCSP) and we are also grateful to Parks Canada for the use of their facilities at Ward Hunt Island.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00275/full#supplementary-material


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Thaler, Vincent, Lionard, Hamilton and Lovejoy. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Changes of the Bacterial Abundance and Communities in Shallow Ice Cores from Dunde and Muztagata Glaciers, Western China

Yong Chen<sup>1</sup> , Xiang-Kai Li<sup>1</sup> , Jing Si<sup>2</sup> , Guang-Jian Wu3,4, Li-De Tian3,4 and Shu-Rong Xiang1,3,4 \*

<sup>1</sup> School of Life Science, Lanzhou University, Lanzhou, China, <sup>2</sup> Institute of Modern Physics, Chinese Academy of Sciences, Lanzhou, China, <sup>3</sup> Key Laboratory of Tibetan Environment Changes and Land Surface Processes, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, China, <sup>4</sup> Laboratory of Ice Core and Cold Regions Environment, Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Science, Lanzhou, China

#### Edited by:

Julie Dinasquet, University of California, San Diego, USA

#### Reviewed by:

Angelina Lo Giudice, National Research Council, Italy Elizabeth Bagshaw, Cardiff University, UK

> \*Correspondence: Shu-Rong Xiang srxiang@ns.lzb.ac.cn

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Microbiology

Received: 14 July 2016 Accepted: 13 October 2016 Published: 01 November 2016

#### Citation:

Chen Y, Li X-K, Si J, Wu G-J, Tian L-D and Xiang S-R (2016) Changes of the Bacterial Abundance and Communities in Shallow Ice Cores from Dunde and Muztagata Glaciers, Western China. Front. Microbiol. 7:1716. doi: 10.3389/fmicb.2016.01716 In this study, six bacterial community structures were analyzed from the Dunde ice core (9.5-m-long) using 16S rRNA gene cloning library technology. Compared to the Muztagata mountain ice core (37-m-long), the Dunde ice core has different dominant community structures, with five genus-related groups Blastococcus sp./Propionibacterium, Cryobacterium-related., Flavobacterium sp., Pedobacter sp., and Polaromas sp. that are frequently found in the six tested ice layers from 1990 to 2000. Live and total microbial density patterns were examined and related to the dynamics of physical-chemical parameters, mineral particle concentrations, and stable isotopic ratios in the precipitations collected from both Muztagata and Dunde ice cores. The Muztagata ice core revealed seasonal response patterns for both live and total cell density, with high cell density occurring in the warming spring and summer months indicated by the proxy value of the stable isotopic ratios. Seasonal analysis of live cell density for the Dunde ice core was not successful due to the limitations of sampling resolution. Both ice cores showed that the cell density peaks were frequently associated with high concentrations of particles. A comparison of microbial communities in the Dunde and Muztagata glaciers showed that similar taxonomic members exist in the related ice cores, but the composition of the prevalent genus-related groups is largely different between the two geographically different glaciers. This indicates that the microbiogeography associated with geographic differences was mainly influenced by a few dominant taxonomic groups.

Keywords: live cell density, taxonomic group, micro-biogeography, glacier, Tibet Plateau

# INTRODUCTION

A variety of microorganisms including bacteria, archaea, fungi, protozoa, algae, and viruses, and even invertebrates, have been found in glaciers and ice sheets in the Arctic, Antarctic, Greenland, and in other mountains across the world (Skidmore et al., 2005; Nkem et al., 2006; Miteva et al., 2009; Zhang et al., 2009; Branda et al., 2010; Anesio and Laybourn-Parry, 2012; Price and Bay, 2012;

Møller et al., 2013; Stibal et al., 2015; Zawierucha et al., 2015; Kaczmarek et al., 2016). Microorganisms can travel long distances and successfully colonize in cryoconite and snow, and then eventually become buried in ice (Prospero et al., 2005; Takeuchi et al., 2006; Miteva et al., 2009; Anesio and Laybourn-Parry, 2012; Yallop et al., 2012; Boetius et al., 2015; Bagshaw et al., 2016). Bacteria are the most dominant life forms in extremely cold, oligotrophic, and frozen water environments. Some of the glacier bacteria have been found to be phylogenetically distinct from those found in temperate environments, demonstrating the biogeography of individual microorganisms in the glacier ice (Christner et al., 2003; Xiang et al., 2010; Anesio and Laybourn-Parry, 2012; Franzetti et al., 2013; Knowlton et al., 2013). Previous studies have also shown apparent geographic patterns of microbial communities across the snow slope surfaces of mountain glaciers Kuytun 51, Qiangyong, and Rongbuk and among the mountain ice cores Dunde (140-m-long, drilled in 1987), Malan (102 m-long, drilled in 1999), Muztagata (37-m-long, drilled in 2003), and Puruogangri (89-m-long, drilled in 2000), and deep ice cores Greenland GISP2D and Antarctic Vostok 5G and Byrd, which illustrates the various microbial responses to climatic and environmental changes of glaciers and ice sheets (Xiang et al., 2009, 2010; An et al., 2010; Knowlton et al., 2013). The micro-biogeography of whole communities may be influenced by the dynamics of taxonomic groups. However, it is still not clear why specific microorganisms live in certain geographical glaciers, namely the geographic difference of the microbial taxonomical groups, which may behave as ecologically coherent units and environmental predictors in glacier systems.

Only a few taxonomic groups are able to colonize and dominate in the snow, although numerous microorganisms are trapped in the surface snow (Zhang et al., 2008, 2009; Xiang et al., 2009, 2010; An et al., 2010). Previous limited data of glacier surface snow have shown that the bacteria Comamonadaceae and Flavisolibacter sp. are common in both the Kuytun 51 and Qiangyong glaciers but only Rhodoferax (Betaproteobacteria) is dominant in the Kuytun 51 glacier (Xiang et al., 2010). The changes of the dominant bacteria in glaciers are mainly influenced by processes such as wind deposition (airborne or aerosol-associated microorganisms by prevailing winds and dust-associated microorganisms by dust storm events), precipitation deposition (microbial deposition with snow, wet-deposition), and post-deposition by microbial growth in the warming seasons on the glacier surface snow (Xiang et al., 2009; Price and Bay, 2012; Bottos et al., 2014; Peter et al., 2014; Meola et al., 2015; Miteva et al., 2015; Pearce et al., 2016). Among these processes, post-deposition has an important role in the transition of microbial communities in glaciers. Recent studies have shown influences of postdeposition on the transition of communities from the lightsensitive cyanobacteria dominated in the surface snow to the non-light-sensitive bacteria buried in the subsurface snow (Xiang et al., 2009). The geographic differences in microbial communities across the mountain glaciers could be attributed to the mountain barriers, which might control the microbial deposition by changing the prevailing wind directions and moisture sources; while the geographic patterns of the dominant microbial colonizers in glaciers might be also influenced by the local climatic and environmental conditions (Nkem et al., 2006; Xiang et al., 2009, 2010; Demetras et al., 2010; Meola et al., 2015).

The primary goal of this study was to evaluate how the geographic difference of bacterial communities at a taxonomic group level was controlled by the prevailing wind patterns across the mountain glaciers in western China. We investigated two different glaciers, the Muztagata glacier (38◦ 170N, 75◦ 040E) and the Dunde ice cap (38◦ 060N, 96◦ 240E). Six structures of bacterial communities were established from the Dunde ice core columns (at field depth 0.8–5.3 m) using bacterial 16S rRNA gene clone library technology. Additionally, live bacteria were examined and related to the physicalchemical parameters from the Muztagata and Dunde ice cores.

#### STUDY AREA, DATA COLLECTION, AND METHODOLOGY

In this study, data were collected from the Muztagata Glacier (38◦ 170N, 75◦ 040E), the Dunde ice cap (38◦ 060N, 96◦ 240E), and the Puruogangri ice cap (33◦ 540N, 89◦ 100E) where precipitation patterns were mainly controlled by two different circulations- westerly and monsoon (as indicated by the highlighted arrows in the **Figure 1**; **Table 1**). The Muztagata Glacier is located in the most western margin of the Tibetan Plateau, where precipitation is mainly controlled by westerly circulation originating in the arid and semiarid regions, including the deserts Sary-Ishykotrau, Muyun Kum, Kyzyl Kum, Kara Kum, Taklimakan, and Gurbantunggut (Wake et al., 1990). The Dunde ice cap is located in the northern margin of the Qaidam Basin and in the Qilian mountain region on the northeastern Tibetan Plateau, where the winter precipitation results from the incursion of westerly depressions along the southern slopes of the Himalaya (Murakami, 1987; Davis et al., 2005), while the summer precipitation is derived from the monsoon circulation from the Bay of Bengal to central Himalaya, and further to the Qaidam Basin and large depressions in Takalimakan Desert and Daidam Basin (Chen and Bowler, 1986; Davis et al., 2005). The Puruogangri ice caps are located in the center of the Tibetan Plateau, where precipitation is derived from a westerly direction during winter and Indian monsoons in the summer (Wake et al., 1993; Shi and Liu, 2000).

The ice core Muztagata (37-m-long) was extracted at 7010 m ASL (above sea level) from the Muztagata Glacier in the summer of 2003 (Tian et al., 2006). The Dunde ice core (9.5-m-long) was extracted at 5325 m ASL from the Dunde ice cap summit in October 2002 (Wu et al., 2009). The visible stratigraphic features were recorded immediately after ice core drilling. All ice cores were returned frozen to the freezer room (air temperature

between −18◦C to −24◦C) at the Key Laboratory of the Ice Core and Cold Regions Environment of the Chinese Academy of Sciences. The ice core sections were split lengthwise into four portions and stored in a refrigerated room with a temperature of −18◦C to −24◦C.

A 10 ml aliquot of melt-water from the Muztagata and Dunde ice cores was used for the analysis of the mineral particles. Total micro-particle concentrations were measured by using a Coulter counter Multisizer 3 (Beckman). A total of 44 ice samples were analyzed from the Muztagata ice core taken at a depth of 2.5–12.5 m, and 74 samples were analyzed from the Dunde ice core taken at a depth of 0.50–9.8 m.

A 10 ml aliquot of melt-water from the Dunde ice core was used for analysis of the stable isotopic ratios, <sup>18</sup>O/16O (δ <sup>18</sup>O) in the precipitation. A Finnegan MAT-252 mass-spectrometer was used to determine δ <sup>18</sup>O values within ± 0.05%. The Dunde ice core was dated by using seasonal δ <sup>18</sup>O variations and annual visible dust layers and confirmed by the previous


data (Takeuchi et al., 2009). The Muztagata ice core dating and δ <sup>18</sup>O data were previously described by Tian et al. (2006).

The ice core sections were cut into small ice columns in intervals of 12–30 cm using a band saw within the walk-in freezers (−18◦C to −24◦C). Microbial analyses were carried out on 156 and 37 samples from Muztagata and Dunde, respectively. The ice samples were cut between the visible dust layers, and ice layers were collected separately. The improved procedures were used for the decontamination of the outer surfaces of ice core samples. The snow and firn-ice columns (length approximately 15 cm, diameter 5 cm) were decontaminated by cutting away the 10-mm annulus with an autoclaved sterile sawtooth knife. The knife was sterilized over an alcohol flame following each ice slice cut. A total of three sterile sawtooth knives were used for each ice sample. The decontaminated samples were then completely melted in clean and sterile glass beakers at 4◦C. These handling procedures were undertaken at temperatures below 20◦C within a sterile, positive pressure laminar flow hood as described before (Yao et al., 2006). The freshly melted water (10 ml) from the Muztagata and Dunde ice cores was 10-fold diluted with sterile filtered water. A total of 100 µl of diluted sample was added to the known concentration of fluorescentdyed bead solution Trucount (Becton Dickinson) mixture with the cell sorting markers carboxyfluorescein diacetate (cFDA) and propidium iodide (PI). Three groups of bacteria could be identified based on the difference of the bound probes: cFDA-stained, cFDA/PI-double-stained, and PI-stained group, indicating viable, injured, and dead cells, respectively (Xiang et al., 2009). The cFDA and PI staining were separately prepared by following the method of Amor et al. (2002), except for the cell suspensions that were incubated for 15 min in the dark at room temperature (25◦C) for cell staining. The 100 µl sterile filtered water served as a reagent blank. The live and total cell numbers in the melt-water were determined with a precision ± 0.05% by using a FACSCalibur flow cytometer (Becton Dickinson Immunocytometry Systems, San Jose, CA, USA) and following the manufacturer's instruction.

For DNA analysis, six clone libraries of the bacterial 16S rRNA genes were collected from the Dunde ice cap. Approximately 400 ml of ice core melt-water was used for the DNA extraction. DNA extraction and further clone library establishment procedures were conducted by following the same protocols as previously used in a microbial analysis of the Kuytun 51 Glacier samples (Xiang et al., 2009). All reagent transfers for DNA analysis were performed within a sterile, positive pressure laminar flow hood. All reaction tubes and micropipette tips were autoclaved, and all solutions except for the Taq DNA (2.5 U, TakaRa) polymerase were passed through sterile 0.2 µm filters (Xiang et al., 2004). The 16S rRNA gene amplicons used for the establishment of clone libraries from the Dunde ice core were generated by PCR amplification with the bacterial universal primer pair 8f (5<sup>0</sup> -AGAGTTTGATCATGGCTCAG) and 1492R (5'-CGGTTACCTTGTTACGACTT; Lane, 1991; Weisenburg et al., 1991). To avoid possible bias, the three PCR products were pooled and used to establish a clone library from each ice column. A total of 137 clones were selected for sequencing by HaeIII-based ARDRA (amplified rRNA restriction analysis) out of the 406 clones from the Dunde ice core. Each sequence was named using the initial of Dunde ice cap (DD1, noted for one out of the five ice cores drilled in October 2002, Wu et al., 2009), along with the ice depth (D84, D107, D238, D324, D386, and D466: 84, 107, 238, 324, 386, and 466 cm below the snow surface) followed by the clone number (1–163). For example, clones DD1D84-9, DD1D107-55, and DD1D466-123 were the clone representatives of the ice core DD1 taken at the depth 84, 107, and 466 cm below the snow surface. The GenBank accession numbers of the cloned sequences obtained from the Dunde ice core are KU060881– KU061017.

All 137 sequences from the Dunde ice cap were checked by DECIPHER (Wright et al., 2012, sequence chimera check tool<sup>1</sup> ) and aligned with the Blast references (Altschul et al., 1990) by using ClustalX (Thompson et al., 1997). A Neighbor-Joining phylogeny for the aligned sequences was constructed using MEGA 6.0<sup>2</sup> (Tamura et al., 2013) pairwise deletion mode for gaps (with bootstrap analysis, 100 replicates) and subroutines Maximum Composite Likelihood (MCL) for substitutions. The archaeal 16S rDNA sequences from Methanosaeta harundinacea strain 8Ac (accession no. AY817738) and Methanosaeta concilii strain GP6 (accession no. NR102903) were used as outgroup references on all trees. All the obtained sequences from the glaciers were identified by the recognized species and were related to the ecological clusters (e.g., Variovorax sp. and Herbaspirillum sp. in the Betaproteobacteria subphyla). Sequences obtained displaying similarities of >97% with known species were identified as the reported species. Most of the obtained clones were related to known cultivated genera or genus clones (e.g., Ketogulonicigenium sp., Cyanobacterium sp., and Sphingobacterium sp.). A few clones had <97% similarity with reported species, and thus were designated separately.

#### RESULTS

#### Seasonal Changes in Physical-Chemical and Biological Parameters in the Muztagata Ice Core

There was an obvious seasonal effect on temperature and biological parameters along the ice core extracted at 7010 m ASL of the Muztagata Glacier (**Figure 2**). An apparent seasonal temperature change was indicated by the proxy value of the stable isotopic ratios, <sup>18</sup>O/16O (δ <sup>18</sup>O), with a low value in winter and a high value in summer (**Figure 2B**). The live cell density was greatly variable and ranged from 6.5 × 10<sup>2</sup> to 2.1 × 10<sup>4</sup> cells/ml between 1964 and 2000 (**Figure 2A**). The total cell density varied from 4.4 × 10<sup>4</sup> to 8.7 × 10<sup>5</sup> cells/ml (**Figure 2C**). Several live cell density peaks were formed during the summer seasons in 1969, 1970, 1973, 1979, 1982, 1983, 1988, 1990, and 1993 for a total of nine events, a1 to a9 (open triangles in **Figure 2A**), while cell density peaks were found in spring (filled triangles in

<sup>1</sup>http://decipher.cee.wisc.edu/FindChimerasOutputs.html <sup>2</sup>http://www.megasoftware.net/

**Figure 2A**). This ice core also had an increased density of the total number of microorganisms in the summer of 1978, 1988, and 1993 (open triangles c1, c2, and c3 in **Figure 2C**), and in the spring of 1995 and 2000 (c4 and c5 in **Figure 2C**), which was consistent with the live cell density patterns (**Figure 2A**). The microbial cell density correlated with the concentrations of mineral particles and possessed a high R 2 value of 0.68 only from 1994 to 2000 (from ice core depth 2.5 to 9.3 m, **Figures 3A,B**), but did not correlate with mineral particle concentrations from 1990 to 1993 (from ice core depth 9.3 to 12.5 m, with R <sup>2</sup> < 0.1, **Figure 3B**).

#### Changes in Physical-Chemical and Biological Parameters in the Dunde Ice Core

Seasonal analysis of the Dunde ice core was not successful due to the limitations of sample resolution (**Figure 4**). Oxygen isotope ratios of the melt-water samples from the Dunde ice core showed a change range from –10.78h to –8.24h (temperature proxy <sup>18</sup>O/16O, **Figure 4D**), while microbial cell density varied from 1.2 × 10<sup>3</sup> to 9.1 × 10<sup>4</sup> cells/ml (**Figure 4B**) and 1.3 × 10<sup>5</sup> to 1.9 × 10<sup>6</sup> cells/ml (**Figure 4C**) for live and total cell density, respectively. Three peaks c2, c3, and c4 of the total cell density were found in the spring of 1988–1989, 1992, and 2000, only one peak, c1, was found in the summer of 1985 (**Figure 4C**). The live cell density response pattern was consistent with the total cell density tendency (the dash lines in **Figures 4B,C**). An abundance of microbial cells frequently occurred at the dirty ice layers (Cell density peaks c1, c3, and c4 at the dust layers labeled as a1, a3, and a4 at the dash lines in **Figures 4A,C**), but were rarely found at the clean ice layer (small density peak c2 at the a1 ice layer in **Figure 4**).

#### Phylogenetic Analysis of Bacterial 16S rRNA Gene Amplified from the Dunde Ice Core

The dominant bacteria in six ice layers of the Dunde ice core were investigated using 16S rRNA gene clone library, sequencing techniques, BLAST and phylogenetic tools. A total of 24 bacterial genera were identified in the Dunde ice core. They belonged to

genera Polaromonas sp., Rhodoferax sp., Variovorax sp., Burkholderiales, Herbaspirillum sp., Xanthomonadaceae, Ketogulonicigenium sp., Devosia sp., Bacteriovorax sp., Hymenobacter sp., Pedobacter sp., Flavobacterium sp., Flectobacillus sp., Cytophagales, Sphingobacteriaceae, Cryobacteriumrelated, Propionibacterium/Blastococcus sp., Salinibacterium/ Frigoribacterium sp., Knoellia sp., Cyanobacteria, Luteolibacter sp., Paenibacillus sp., Anoxybacillus sp., and TM7 candidates (**Figures 5–7**). Three genus groups Cryobacteriumrelated, Salinibacterium/Frigoribacterium sp., and Propionibacterium/Blastococcus sp. were clustered with 65–76% similarity to the known species but grouped with genus Knoellia sp. with 95% similarity in the family members of Actinobacteria (**Figure 6**). Only one clone DD1D107-100 was 100% similar to the uncultured Bacteroidetes clone AKYG1686 (**Figure 7**). All tested bacterial clones in the ice fell into members of bacteria phyla Alpha, Beta, Gamma, and Deltaproteobacteria, Actinobacteria, Bacteroidetes, Firmicutes, Verrucomicrobia, and TM7 candidates.

# Changes in Proportion of the Main Bacterial Genera along the Dunde Ice Core Profile

There was a large difference in the proportion of the main phylogenetic groups along the Dunde glacier depth profile, which indicated the seasonal changes of microbial communities in the glacier (**Figures 8A1–A6**). The bacterial clones were comprised of five dominant genus groups, Polaromonas sp., Pedobacter sp., Flavobacterium sp., Propionibacterium/Blastococcus sp., and Cryobacteriumrelated, which accounted for more than 55% of the total 406 clones and frequently appeared in the six tested ice layers from 1990 to 2000 (dashed lines in **Figures 8A1–A6**). Nine genus groups such as Rhodoferax sp., Variovorax sp., Burkholderiales, Flectobacillus sp., Cytophagales, Sphingobacteriaceae, Knoellia sp., Cyanobacteria rarely occurred in the ice. Other opportunistic bacterial clones occasionally appeared in the ice.

#### FIGURE 5 | Phylogenetic analysis of the 16S rRNA genes for Alphaproteobacteria, Betaproteobacteria, Gammaproteobacteria, and Deltaproteobacteria clones from the Dunde ice core and the closest relatives. The tree was generated by the Neighbor-Joining method after sequence alignment, and rooted with two Methanosaeta strains (accession no. AY817738 and NR102903). Bootstrap values (100 replications) were specified for each Node. Cut-off value for the condensed tree was 55%. Numbers of the obtained snow-ice clones (had the same ARDRA pattern to the sequenced representatives listed on

the tree) and relative sequence affiliations corresponding to GenBank accession number were provided in parentheses. The sequences discussed in this study were noted bold. See a detailed description for the assigned sequence references and numbers in "Study Area, Data Collection, and Methodology."

Dunde ice core and the closest relatives. The tree was constructed by following the protocol as described in Figure 5.

# DISCUSSION

fmicb-07-01716 October 28, 2016 Time: 13:59 # 12

Previous studies have shown the prevalence of specific bacteria in certain local glaciers (Zhang et al., 2008; Xiang et al., 2009; An et al., 2010; Franzetti et al., 2013; Miteva et al., 2015). However, our findings demonstrate that the members of bacterial genus-related groups are highly similar in the related ice cores at a historical scale, whereas the composition of the prevalent genus-related groups is largely different across the geographically different glaciers. This indicates that the micro-biogeography associated with geographic differences was mainly influenced by a few dominant taxonomic groups.

### Methodological Considerations

Contamination of the DNA samples (from the inner core columns) used in this study is unlikely because the outer surfaces of ice core and reagents for DNA analysis were cautiously decontaminated, and all the procedures were performed within a sterile, positive pressure laminar flow hood. Only small DNA fragments (<100 bp) were detected from the ice column control of (autoclaved sterile water), which were not considered for further sequence analysis in this study. It should be noted that Herbaspirillum sequences, also found in this study, have previously been identified as potential contaminants in glacier debris and ice samples (Cameron et al., 2016). However, the experimental procedures used by Cameron were completely different from ours in the present study. Their protocols and procedures were used for glacier cryoconite debris and surface ice samples. Herbaspirillum sp. found in this study, are well-known plant root-associated nitrogen-fixing (Baldani et al., 1986) and non-nitrogen-fixing environmental species (Ding and Yokota, 2004; Dobritsa et al., 2010). They were also reported in the Alaska Gulkana glacier, an Antarctica glacier forefield and the Antarctica Lake Vida brine (Segawa et al., 2011; Bajerski et al., 2013; Kuhn et al., 2014).

Various molecular techniques, CFM with cell stains cFDA, PI, and SYTOX have been used to investigate viable bacteria (Amor et al., 2002; Schumann et al., 2003; Xiang et al., 2009). These tools helped us to examine the abundance of live cells and the potential metabolic activities of microorganisms in an environment. However, the CFM approach has certain limitations because of interference from dust particles or spurious abiotic autofluorescence and underestimation of the accurate cell counts under the typical CFM parameters (Stibal et al., 2015). Despite the limitations, the background noise can be counterweighed by data series from the ice core profiles. In this study, the apparent seasonal tendency suggests that our analyses were based on a substantial fraction of bacteria.

For the phylogenetic analysis of bacteria, more than 600 clones were picked and sequenced. A total of 406 valid bacterial clones were obtained from the Dunde ice core after vector, and chimeric checking. The rarefaction curves of six clone libraries from the ice core were approaching asymptotes (dada not shown). Data also showed the prevalence of a few dominant genus-related groups at the different ice core depths (**Figures 8A1–A6**). This indicated that the identified clones were based on the dominant bacterial taxa.

# Dust Deposition and Microbial Distribution along the Glacial Depth Profiles

The present data sets from the Muztagata glacier at 7010 m ASL (38◦ 170N, 75◦ 040E) revealed a high correlation between dust and microbial abundance from 1994 to 2000, which indicated a strong influence of aeolian activities on the microbial deposition in the glacier snow (**Figures 3A,B**). This was also consistent with another independent microbial investigation on the Muztagata glacier at 6300 m above sea level (Liu et al., 2013). The Dunde ice core also presented a frequent association of microbial cell density peaks with high concentrations of mineral particles (a1, a3, and a4, verse c1, c3, and c4 in **Figures 4A,C**). The strong association of microorganisms with dust was also found in previous data from the Antarctic glacier (Abyzov et al., 1998; Priscu et al., 2008), the Malan glacier (Yao et al., 2006), and the Guoqu glacier on the Tibetan Plateau (Yao et al., 2008). The analyses of trace and rare earth elements extracted from the same series of Dunde ice core sections showed that the fine fractions in the Dunde dust were more similar to those in the western Qaidam Basin and Tarim Taklimakan Desert than those in the Badain Juran and Tengger Desert (Wu et al., 2009). The Nd-Sr isotopic composition of mineral particles in the Dunde ice core is also similar to that in desert sand from Qaidam and Tarim Taklimakan (Wu et al., 2010). All results revealed that the Qaidam Basin and the Tarim Taklimakan Desert was the main source of dust in the Dunder glacier, implying the transportation of dust-borne microorganisms from the western desert Tarim Taklimakan and adjacent Qaidam to the Dunde glacier.

However, the Muztagata ice core data showed independence of microbial load with dust deposition from ice core depth 9.3 to 12.5 m (**Figure 3B**). The Dunde ice core data also showed one small cell density peak c2 appearing at the clean ice layer a2 in **Figure 4**. These results indicate that microbial deposition in the glacier snow does not always associate with the dust deposits or "dirty" wind and may in fact be transported by "clean" wind or snow, which implies influences of the processes such as aerosol and precipitation deposition, along with other factors (Bottos et al., 2014; Pearce et al., 2016).

#### Seasonal Fluctuation of Bacterial Density at Variable Temperatures

The present data sets from the Muztagata glacier at 7010 m ASL (38◦ 170N, 75◦ 040E), revealed clear seasonal patterns with high microbial cell density occurring in the warming summer months (open triangles **Figure 2**), which indicated positive temperature effects on the microbial density patterns. This was consistent with another independent microbial investigation on the Muztagata glacier at 6300 m ASL (38◦ 170N, 75◦ 060E, Liu et al., 2013). The high repeatability of both ice cores from the Muztagata glacier confirmed the reliability of the data sets discussed here. Evidence for a positive temperature effect includes the algae growth of red Chlamydomonas at the surface snow in New Zealand and on the Alaska Harding icefield and on Greenland and Iceland's glaciers in late spring and summer (Thomas and Broady, 1997; Takeuchi et al., 2006; Yallop et al., 2012; Lutz et al., 2015).

Further temperature effects on bacterial growth, colonization and community transition were reported on Kuytun 51 Glacier, where bacterial Cyanobacteria were dominant across the surface snow slope in the warming spring-summer, but rarely in the subsurface, winter-snow-layers (Xiang et al., 2009). As expected, the live cell density during the summer was high as a result of microbial growth in the surface snow. Other groups, Uetake et al. (2006), Yao et al. (2008), and Price and Bay (2012) also found that high microbial abundance was present in the warming spring-summer seasons in the Sofiyskiy glacier in the south Chuyskiy range of the Russian Altai, the Guoqu glacier in the Geladaindong mountain regions, and the deep Arctic and Antarctic ice cores, respectively. The obvious seasonal patterns of bacterial populations with a high cell density in summer strengthen the post-deposition effect on the microbial populations in glaciers.

In addition to those cell density peaks during the summer (open triangles in **Figures 2A,C**), there were also many density peaks in the spring from 1963 to 2000 (filled triangles in **Figures 2A,C**). The seasonal pattern of bacterial density was generally consistent with the dynamic mineral particle deposition with frequently dust outbreaks in spring and summer (**Figures 3** and **4** in this study; Wu et al., 2008; Liu et al., 2015). This indicated an important influence of dust deposition on the microbial communities in glaciers. All of the results suggest the fundamental contribution of dust-microbe deposition to the basic population pool size and the intensifying effect of post-deposition by microbial growth in the warming seasons.

# Geographic Difference of Microorganisms in the Glacier Ice

The present data showed that the Polaromonas sp. from the Dunde ice core clustered together more closely than those from other environments (**Figure 5**). The phenomena of Polaromonas sp. from the same location readily grouping together was also found in the Muztagata and Puruogangri glaciers (An et al., 2010; Xiang et al., 2010). Although Polaromonas sp. were widely distributed across the geographically different glaciers, statistical analyses demonstrated a large genetic distance among 43 unique glacier Polaromonas sequences, which positively associated with geographic distance (Franzetti et al., 2013). Similar geographic phenomenon of individual microorganisms was also found in the deep ice core. Bacteria Alternaria sp. were common in the deep ice cores of Greenland GISP2D and Antarctic Vostok 5G and Byrd, but their DNA sequences were phylogenetically different between the two polar regions (Knowlton et al., 2013). The geographic differences of Polaromonas sp. and Alternaria sp. across the isolated glaciers suggests that the mountain "barriers" to the microbial transportation can be surmounted by suitable adaptations, which leads to the geographic patterns of individual microorganisms.

Geographic differences are not only evident for Polaromonas sp. and Alternaria sp. but also for the taxonomic groups. It is obvious that there is a geographic distinction of taxonomic groups at the cryoconite habitats on three High-Arctic glaciers from the associated moraines and adjacent tundra on the Brøggerhalvøya peninsula, Svalbard (Edwards et al., 2013, 2014). Significant differences in the composition of dominant taxonomic groups are also found between alpine and Arctic cryoconite habitats (Edwards et al., 2014). The present data from the Dunde ice core showed that similar taxonomic groups frequently appeared along the ice core profiles as historical events (**Figures 8A1–A6**). Bacterial genus groups Cryobacteriumrelated, Flavobacterium sp., Pedobacter sp., Polaromonas sp., and Propionibacterium/Blastococcus sp. were frequently found at the six tested ice layers of Dunde glacier from 1990 to 2000 (**Figures 8A1–A6**). Another example of similar group members sharing the related ice core layers can be found in the recently reported Dunde ice. Genera Polaromonas sp. and Flavobacterium sp. commonly found between 1990 to 2000 were also identified from the Dunde ice column AD 1780– 1830 (Zhang et al., 2009). Although the dominant genus-related groups are similar in the related ice cores, the composition of the main genus-related groups is largely different across the geographically different glaciers. The bacteria Cryobacteriumrelated was more abundant in the Dunde ice cap than in the Muztagata glacier, while Enterobacter sp. appeared throughout the four tested ice layers of the Muztagata glacier but rarely in the Dunde ice cap (**Figures 8A1–A6,B1–B4**). Seven genus groups Polaromonas sp., Enterobacter sp., Acinetobacter sp., Flexibacter sp., Thermus sp., Propionibacteria/Luteococcus sp., and Flavisolibacter sp. were frequently identified in the four tested ice layers of Muztagata glacier from 1970–1988 (labeled as the dashed lines in **Figures 8B1–B4**), while Polaromonas sp. and Flexibacter sp. were found at all three tested ice columns of Puruogangri glacier from 1600 to 1920 (Zhang et al., 2009; An et al., 2010). All results clearly show that a few genusrelated groups are dominant in the mountain ice cores and constitute the main taxonomic groups endemic to the local glacier regions. The difference of taxonomic group members across the geographically different glaciers suggests intermingling of the bacterial taxonomic groups to the point of geographic separation. More data of microorganisms in the deep ice are necessary for our better understanding of the biogeography of microorganisms in glaciers.

The geographic pattern of bacterial taxonomic groups could be attributed to the influence of the moisture and dust source area, which might vary across the mountain glaciers on the Tibetan Plateau (**Figure 1**; **Table 1**). Precipitation over the Muztagata glacier is mostly influenced by the westerly depressions, while precipitation over the Dunde ice cap and Puruogangri ice cap is mainly driven by the westerly depressions in winter and Indian monsoon in summer (Murakami, 1987; Wake et al., 1990; Davis et al., 2005). Dust in the mountain glacier Muztagata is mainly derived from deserts including Sary-Ishykotrau, Muyun Kum, Kyzyl Kum and Kara Kum, Taklimakan, and Gurbantunggut (**Figure 1**; Wake et al., 1993), while the Dunde ice cap is very close to the Gobi Desert, Qaidam Basin (**Figure 1**) and thus its dust components are more likely strongly affected by local dust storms and dominated by mineral particles from the two deserts Qaidam Basin and Tarim Taklimakan (Wu et al., 2009, 2010). The dramatic changes of the moisture sources and dust pathways across the mountainous

glaciers may lead to differences in the microbial communities deposited in the glacier snow. Moreover, the heterogeneity of local conditions such as temperature, light intensity, melt-water availability, and nutrient concentrations in the snow may drive the spatial patterning of the microbial community by influencing the colonization of the dominant endemic species in the snow. Concerns on how surface communities are incorporated into the cores, how much they change after burial, and how the post processes contribute the geographic differences of microbial communities are still open questions. More data on the microbiological, meteorological, and physical and chemical characteristics of the glacier surface and subsurface snow and ice cores will be helpful for a better understanding of the biogeography of microorganisms in glaciers.

### CONCLUSION

The members of bacterial genus-related groups were found to be similar in the related ice cores at a historical scale but largely different between the two glaciers Muztagata and Dunde, even if microbial communities fluctuated along the two ice core depth profiles. Compared to the Muztagata glaciers, the Dunde ice core presented distinct members of the taxonomic groups. The five bacterial genus groups Polaromonas, Pedobacter sp., Flavobacterium sp., Propionibacterium/Blastococcus sp., and Cryobacterium-related frequently appeared at the six tested ice layers, constituting the dominant species endemic to the Dunde ice cap, while seven genus groups Polaromonas sp., Enterobacter sp., Acinetobacter sp., Flexibacter sp., Thermus sp., Propionibacteria/Luteococcus sp., and Flavisolibacter sp. were frequently found at the four tested ice depths of Muztagata glacier. The results demonstrate that the spatial differences in microbial communities between the two ice cores are more significant than the temporal differences. This study also showed

#### REFERENCES


a seasonal pattern of microbial cell density with high cell density occurring in the warming spring-summer.

#### AUTHOR CONTRIBUTIONS

YC: Design of the laboratory experiment outline, data collection, analysis, and interpretation, and draft of the manuscript. X-KL: Sequence data analysis, and interpretation. JS: Sequence data collection, analysis, and interpretation. G-JW: Mineral particle concentration examination of ice core, data analysis, and interpretation. L-DT: Oxgen isotope ratio analysis, and interpretation. S-RX: Design of the research outline, data analysis and interpretation, and revision of the manuscript.

#### FUNDING

This work was supported by the NSF project of China (Grant 31400430, 40471025, and 40871046).

#### ACKNOWLEDGMENTS

We would like to thank all the members of the Muztagata glacier and Dunde glacier expedition for their help with the field sample collection. We would like to thank the anonymous viewers for their constructive suggestions and comments on the manuscript. We also would like to thank Ms. Amanda Biernacka-Larocque and American Journal Expert team for their help on the improvement of English use in this paper (this paper was completely rewritten and resubmitted to Frontiers in Microbiology after the earlier version BG-2015-637 was published in Discussion but later declined by the journal Biogeosciences).




**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Chen, Li, Si, Wu, Tian and Xiang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Uptake of Leucine, Chitin, and Iron by Prokaryotic Groups during Spring Phytoplankton Blooms Induced by Natural Iron Fertilization off Kerguelen Island (Southern Ocean)

Marion Fourquez 1, 2 \*, Sara Beier 2, <sup>3</sup> , Elanor Jongmans <sup>2</sup> , Robert Hunter <sup>2</sup> and Ingrid Obernosterer <sup>2</sup>

<sup>1</sup> Antarctic Climate & Ecosystems Cooperative Research Centre, University of Tasmania, Hobart, TAS, Australia, <sup>2</sup> CNRS, Sorbonne Universités, UPMC Univ Paris 06, Laboratoire d'Océanographie Microbienne (LOMIC), Observatoire Océanologique, Banyuls/mer, France, <sup>3</sup> Department of Biological Oceanography, Leibniz Institute for Baltic Sea Research Warnemünde (IOW), Rostock, Germany

#### Edited by:

Anton F. Post, University of Rhode Island, USA

#### Reviewed by:

Hugh Ducklow, Columbia University, USA Pia Moisander, University of Massachusetts Dartmouth, USA

\*Correspondence: Marion Fourquez marion.fourquez@gmail.com

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Marine Science

Received: 30 May 2016 Accepted: 24 November 2016 Published: 15 December 2016

#### Citation:

Fourquez M, Beier S, Jongmans E, Hunter R and Obernosterer I (2016) Uptake of Leucine, Chitin, and Iron by Prokaryotic Groups during Spring Phytoplankton Blooms Induced by Natural Iron Fertilization off Kerguelen Island (Southern Ocean). Front. Mar. Sci. 3:256. doi: 10.3389/fmars.2016.00256 Iron and carbon are essential for microbial heterotrophic activity, but the bioavailability of these elements is low in surface waters of the Southern Ocean. Whether the access to iron and carbon differs among phylogenetic groups of prokaryotes is barely known. To address this question we used iron (55FeCl3), and the carbon compounds chitin (3H-Diacetylchitobiose) and leucine (3H-leucine) as model substrates in combination with MICRO-CARD-FISH during spring phytoplankton blooms induced by natural iron fertilization off Kerguelen Island (KErguelen Ocean and Plateau compared Study 2—KEOPS2; October-November 2011). The application of probes at broad phylogenetic levels indicated an overall similar community composition in surface waters at the 8 investigated sites. The relative contributions of the prokaryotic groups to abundance revealed a strong positive relationship with their respective contributions to the leucine-active community (p < 0.0001; r = 0.93). This relationship was much weaker for chitin (p < 0.001; r = 0.51) and absent for iron (p > 0.05; r = 0.26). These results suggest preferential uptake of iron and chitin by some prokaryotic groups. SAR11 and Cytophaga-Flavobacterium-Bacteroides (CFB) were the dominant contributors to the leucine-active community, while CFB and Archaea had the highest contributions to the chitin-active community. By contrast, Gammaproteobacteria, including SAR86, and CFB revealed the highest contributions to the iron-active community. We found several correlations between the group-specific fractions of active cells for a given substrate and most of them included CFB, pointing to the potential importance of microbial interactions for iron and carbon cycling in the Southern Ocean.

Keywords: iron, chitin, leucine, prokaryotic community composition, MICRO-CARD-FISH, Southern Ocean

# INTRODUCTION

The Southern Ocean is the largest High Nutrient Low Chlorophyll (HNLC) area where major nutrients are perennially present at high concentrations yet phytoplankton biomass remains low. Surface depletion in iron (Fe) was demonstrated to be the cause of these paradoxical HNLC conditions (Martin et al., 1990). The increase in phytoplankton biomass in response to Fe input and the consequent enhancement of the CO<sup>2</sup> uptake were reported in several mesoscale Fe fertilization studies (reviewed in Boyd et al., 2007) and from naturally fertilized regions (Blain et al., 2007; Pollard et al., 2009). These previous studies have reached the same conclusion as to the importance of Fe for the biological pump of carbon in the Southern Ocean.

Despite the numerous studies conducted up to date, it is not well understood how Fe affects heterotrophic prokaryotic metabolism. Fe and carbon are tightly coupled in a suite of metabolic processes crucial for growth. Fe plays a pivotal role in the carbon metabolism because pathways such as glycolysis, the citric acid cycle and processes related to respiration rely on multiple Fe-containing enzymes. A deficiency in Fe ultimately results in a reduction in the metabolic activity with consequences on the energetic status of the organism and its ability to proliferate (Kirchman et al., 2000; Fourquez et al., 2014). Experimental studies using bacterial strains revealed a positive effect of Fe on bacterial growth (Tortell et al., 1996; Fourquez et al., 2014). The response of natural bacterial communities to Fe addition varies among studies (see Obernosterer et al., 2015 for an overview). A common observation is the positive effect of Fe addition on heterotrophic bacteria when incubated in the light and in the presence of phytoplankton. Such effect has been discussed to be due to labile organic matter released by phytoplankton and renders a firm conclusion on Fe or carbon limitation difficult. The picture is further complicated by the potential competition between heterotrophic bacteria and the small representatives of phytoplankton (Fourquez et al., 2015). The close interrelatedness of Fe and carbon renders it challenging to decipher the cycling of these elements within Southern Ocean microbial communities.

In this context, an interesting feature that characterizes surface waters of the Southern Ocean is the low concentration of dissolved organic carbon (DOC, roughly 50µM; Hansell et al., 2009). This results from the low primary production in Felimited surface waters and the large-scale upwelling of low DOC circumpolar Antarctic deep waters. As a consequence, surface water dissolved organic matter (DOM) is characterized by overall low bioavailability. Fe and carbon represent therefore both potentially limiting elements for heterotrophic prokaryotes in Southern Ocean surface waters.

The objective of the present study was to provide a novel perspective on this issue, by investigating the incorporation of Fe and two distinct sources of carbon, leucine and chitin, by major prokaryotic groups in the Southern Ocean using the single-cell approach MICRO-CARD-FISH. In contrast to leucine, a monomeric amino acid, the utilization of the polymer chitin is more complex involving several regulated enzymes as for instance chitinase or β-N-acetyl-hexosaminidases: chitinases cleave the polymer extracellularly into dimers or monomers which are then transported into the cell and further processed (Beier and Bertilsson, 2013). We performed our study during the KEOPS2 project (KErguelen Ocean and Plateau compared Study 2) from 8 October to 30 November 2011 over and downstream of the Kerguelen Plateau in the Southern Ocean. KEOPS2 was performed during the early stage of the phytoplankton bloom that forms annually in the vicinity of the Kerguelen Plateau in response to natural Fe fertilization of the surface waters. During this study we had access to a patchwork of spring phytoplankton blooms induced by large-scale natural Fe fertilization of the Southern Ocean and HNLC-waters, reflecting a range of conditions with respect to the availability of Fe and organic carbon.

# MATERIALS AND METHODS

#### Sample Sites

Samples were collected in surface waters (20 m depth) at 8 sites in the naturally Fe-fertilized region east of Kerguelen Island and at one site in HNLC waters during Austral spring (**Figure 1**). The hydrodynamic properties of the study region are presented in detail in Park et al. (2014) and basic biogeochemical parameters are presented in **Table 1**. The region around Kerguelen Island is characterized by the passage of the Antarctic Circumpolar Current (ACC) that splits into two branches: a main branch circulates to the south of Kerguelen Islands to further join a branch of the Fawn Trough Current (FTC), and a branch that circulates north of Kerguelen Island (Park et al., 2008). A narrow jet of ACC water also flows across the Kerguelen Plateau. This feature corresponds with the northernmost branch of the Polar Front (Park et al., 2014). These hydrographic features result in mesoscale activity that generate contrasted environmental conditions with respect to Fe availability. Station R-2 located south west of Kerguelen Island served as a HNLC station reference. Station F-L represents a bloom site located north of the Polar Front and Station A3-2 is located above the Plateau. Low horizontal advection characterized Station A3- 2, Station R-2 and the E-sites (Park et al., 2014), as discussed below. Stations E (E-1, E-3, E-4, and E-5) were located in a recirculation feature south of the Polar Front. Multiple visits were conducted in this feature. The E stations were sampled in a quasi Lagrangian manner (D'Ovidio et al., 2015) and can be considered as a time-series. Briefly, the quasi Lagrangian strategy was based on the deployment of two drifters equipped with GPS and iridium transmission at Station E-1. Their positions were monitored in real time and the subsequent E stations were placed along the trajectory of these drifters. A broader context of the horizontal circulation in the region of interest was provided by two other approaches. Firstly, the trajectories of water parcels were determined by geostrophic currents derived from satellite altimetry (D'Ovidio et al., 2015). This approach computes backward trajectories thereby providing an estimate for the rate at which water parcels are transported by horizontal stirring. Secondly, this altimetry-based approach was validated by the trajectories of 200 World Ocean Experiment-Surface Velocity Program drifters, 48 drifters deployed in an adaptive strategy

FIGURE 1 | Location of the study area, with a zoom on the Kerguelen Plateau in the Southern Ocean. The positions of the stations located inside the blooms are shown in black dots and the reference station (R-2) outside the bloom is shown as a white dot. The base map represents the bathymetry in meters.



For all parameters mean values ± SD for the mixed layer (ML) are given. n.a., not available; Temp, temperature; Chla, chlorophyll a; DOC, dissolved organic carbon; ML, mixed layer depth; DFe, dissolved iron; µ, prokaryotic growth rate; ¶, only one data point available for the ML.

<sup>a</sup>From Blain et al. (2015).

<sup>b</sup>From Quéroué et al. (2015).

<sup>c</sup>From Tremblay et al. (2015) and Obernosterer et al. (2015).

<sup>d</sup>From Lasbleiz et al. (2014).

<sup>e</sup>From Christaki et al. (2014).

during the KEOPS2 cruise and 24 drifters during a cruise in January 2014 in the study region (D'Ovidio et al., 2015). Further, 120 trajectories measured between 2011 and 2014 in the study region were collected from the Global Drifter Program. Taken together, this information suggests that the water within the recirculation feature south of the Polar Front is retained over a relatively long time period (roughly 60 days, D'Ovidio et al., 2015).

#### Incubations with Radiolabeled Substrates for Microautoradiography

To determine the single-cell activities of leucine and chitin, raw seawater samples (10 mL) were incubated with either [4,5-3H] leucine (Perkin Elmer, specific activity 144 Ci mmol−<sup>1</sup> ; 10 nM final concentration) or [3H]-Diacetylchitobiose (Amersham, specific activity 14 Ci mmol−<sup>1</sup> ; 10 nM final concentration) in the dark at in situ temperature for 6–8 h. Controls were fixed with paraformaldehyde (PFA, 2% final concentration) 10 min prior to incubation with the radiolabeled substrates. The incubations were terminated by adding PFA to the samples (2% final concentration). Finally, samples were filtered onto 0.2 µm PC filters (25 mm filter diameter, Nuclepore) and subsequently rinsed with 0.2 µm filtered Milli-Q water. Filters were stored at −20◦C until treated.

Incubations with <sup>55</sup>Fe are described in detail in Fourquez et al. (2015). Briefly, samples were collected using trace metal clean 10-L modified Niskin bottles set up on an autonomous model 1018 trace metal rosette especially adapted for trace metal work (General Oceanics Inc., USA; Bowie et al., 2015). The seawater collected was further processed in a clean container under a laminar flow hood (ISO class 5). Seawater was filtered at low pressure (<200 mm Hg) through 0.8 µm acid-washed PC filters (47 mm; Millipore) and incubations were performed on the <0.8 µm fraction. Subsamples (100 mL) of filtered seawater were spiked with <sup>55</sup>Fe (as <sup>55</sup>FeCl3, specific activity 1.83 × 10<sup>3</sup> Ci mol−<sup>1</sup> , Perkin Elmer) at final concentrations of 0.2, 1, 5, or 15 nM of <sup>55</sup>Fe and they were incubated in the dark at in situ temperature for 1 and up to 7 days. Subsamples (10 mL) were collected on 0.2-µm PC filters (25 mm diameter; Millipore) and rinsed with the washing solution Ti-citrate-EDTA for 2 min. This step allows eliminating <sup>55</sup>Fe non-incorporated by cells from the filters. This was followed by 10 rinses of 0.2 µm-filtered seawater of 1 min each (Fourquez et al., 2012). Control treatments of the seawater samples were killed with formaldehyde (2% final concentration) and kept for 1 h at 4◦C before the addition of <sup>55</sup>Fe. All the filters were dried and kept at −20◦C until processed.

The samples for microautoradiography were chosen to (1) limit isotope dilution as determined by saturation curves and (2) reach a quota in <sup>55</sup>Fe per cell sufficient for silver grain production during microautoradiography. Fourquez et al. (2012) showed previously that the fraction of cells associated with silver grains was linked to the duration of the <sup>55</sup>Fe incubation. For the present data set, we chose the samples at the time point when these two conditions were met. We found that 5 nM (St. A3-2 and E-4E) and 15 nM (St. E-4W and E-5) of <sup>55</sup>Fe final concentration and a minimal incubation time of 3 days were necessary to meet these requirements.

#### Catalyzed Reporter Deposition–Fluorescence In situ Hybridization and Microautoradiography (MICRO-CARD-FISH)

We used catalyzed reporter deposition–fluorescence in situ hybridization (CARD-FISH) on filter sections from seawater samples following the incubation with the radiolabeled compounds. The relative abundances of the bacterial groups and Archaea were determined using the CARD-FISH protocols described in Sekar et al. (2003) and Alonso-Sáez et al. (2007), respectively. We followed the protocol of Cottrell and Kirchman (2003) for the microautoradiographic development of leucineand chitin-active cells and the protocol of Fourquez et al. (2012) for the development of Fe-active cells. Briefly, for CARD-FISH, filters were embedded in low-melting-point agarose (0.2% final concentration), dried and dehydrated (96% Ethanol, 1 min). To target bacterial groups, the filters were treated with lysozyme (Fluka, 10 mg mL−<sup>1</sup> , 100 mmol L−<sup>1</sup> Tris [pH 8], 50 mM EDTA) for 1 h at 37◦C to allow cell wall permeabilization. To target Archaea, or more specifically Crenarchaeota, cell wall permeabilization was performed as described previously (Tischer et al., 2012). The filters were subsequently washed in Milli-Q water and dehydrated in ethanol (96%) for 1 min. Probe working solution was added at a final concentration of 2.5 ng µL −1 and the hybridization was performed at 35◦C for 2 h. Bacterial probes and archaeal probes were labeled with FITC and Alexa Fluor <sup>R</sup> dyes (dyes 488 or 546, Thermo Fisher Scientific, Waltham, MA, USA), respectively.

The probes Eub338-I, -II, and -III were used for the identification of Bacteria and the Eub338-I antisense probe Non338 was used to determine non-specific binding (Glöckner et al., 1999). The probes SAR11-152R, SAR11-441R, SAR11- 542R, SAR11-732R were used to target the SAR11 cluster (Morris et al., 2002) and the probe Ros537 to target members of the Roseobacter-Sulfitobacter-Silicibacter group (Eilers et al., 2000a). To target Gammaproteobacteria, we used the probe Gam42a (Manz et al., 1992) and the probes SAR86 (Eilers et al., 2000b) and SAR92 (Obernosterer et al., 2011) were used to target the SAR86 and SAR92 clusters, respectively. The probe CF319a (Manz et al., 1996) was used to identify Bacteroidetes. To target Archaea, we used the probe Arc915 (Auguet and Casamayor, 2008) and to target Crenarchaeota Marine Group 1 we used the probe Cren554 (Alonso-Sáez et al., 2007). Earlier experiments with probe Arc915 in freshwater samples revealed that this probe hybridized in some cases also with bacterial cells (data not shown). Therefore, in order to test the specificity of probe Arc915 in the current sample set, a few filters were hybridized consecutively with the probes EUBI/II/III (Alexa Fluor <sup>R</sup> 488 dye) and Arch915 (Alexa Fluor <sup>R</sup> 546 dye). Inactivation of the horseradish peroxidase between the hybridization steps was performed as described elsewhere (Pernthaler and Pernthaler, 2007). We could not detect any unspecific hybridisation via multi-color CARD-FISH with the probes Arc915 and EubI/II/III.

Following the CARD-FISH step, filters were treated for microautoradiography as detailed in Cottrell and Kirchman (2003). Briefly, after hybridization the filters were deposited with cells facing down onto a glass slide previously dipped into a molten (43◦C) solution of NTB2 photographic emulsion (Kodak, diluted 50:50 with milli-Q water). The photographic emulsion was left to solidify on ice for a minimum of 10 min before the slides were stored for autoradiographic exposure times that ranged from 2 to 3 days for leucine and chitin to 7 weeks (Fe). All manipulations were conducted in the darkroom under a safe light. Slides were developed for 2 min in Kodak developer as detailed in Cottrell and Kirchman (2003).

All cells were stained before observation by adding one drop of 4′ ,6-diamidino-2-phenylindole (DAPI, 2 µg mL−<sup>1</sup> final concentration) directly prepared in an antifade mountant (1:4 mixture of Vectashield <sup>R</sup> and Citifluor <sup>R</sup> ) and subsequently detected in images captured with epifluorescent microscopy. The total number of cells (DAPI-cells), cells affiliated with a specific prokaryotic group and cells that assimilated the radiolabeled compound (with silver grains) were counted using a semiautomated microscope (Olympus BX61) and image analysis software (Microbe Counter software) as described previously (Cottrell and Kirchman, 2003). Data were collected from 10 fields of view with at least 500 cells per sample. The proportions of DAPI-cells that were substrate-active were determined as the fraction of total DAPI-cells with silver grains. The composition of the substrate active community was determined as the fraction of cells with silver grains that are probe-positive. The proportion of a given prokaryotic group that was substrate-active was determined as the fraction of probe-positive cells with silver grains.

For a given sample, multiple hybridizations were done with all probes: CARD-FISH only, and CARD-FISH on samples previously incubated with leucine, chitin or iron. The number of replicates varied between 3 and 10 (all probes included) and the variability (coefficient of variation) was on average 24 ± 14% (n = 44) for "high abundant probes" (>5% FISH) and was on average 57 ± 30% (n = 15) for "low abundant probes" (≤5% FISH) such as SAR92 and Cren554.

Bacterial and archaeal group relative abundance was calculated as percentage of probe positive cells vs. total DAPIcells counts. The relative abundances of bacterial groups (EUB probes) accounted for 81 ± 4% DAPI-cells (mean ± SD, n = 8) and nearly reached a total cover at the sites where Archaea were investigated 98 ± 6% DAPI-cells (mean ± SD, n = 4). The sum of the relative contribution of SAR86 and SAR92 accounted for 101 ± 26% (mean ± SD, n = 8) of the cells identified as Gammaproteobacteria.

#### Uptake Experiments

Leucine uptake rates were determined as detailed in Christaki et al. (2014). Briefly, 20 ml triplicate samples plus one trichloroacetic acid (TCA)-killed control were incubated with <sup>3</sup>H-leucine for 3–8 h at in situ temperature. Incubations were terminated by adding 5% ice-cold TCA and cells were subsequently collected by filtration on 0.22 µm cellulose acetate filters (Millipore) and washed with ice-cold 5% TCA. Filters were dissolved in 1 mL of ethyl acetate and 10 mL of Ultima Gold scintillation cocktail (Packard) before being counted by liquid scintillation.

Chitin uptake rates were estimated at 4 stations (A3-2, F-L, E-3, and E-4W) using a concentration series bioassay (Wright and Hobbie, 1966; Zubkov and Tarran, 2005). Raw seawater samples were incubated with [3H]-Diacetylchitobiose (Amersham, specific activity 14 Ci mmol−<sup>1</sup> ) at final concentrations of 0.5, 1, 2, 3, and 6 nM. Ten mL of each addition were incubated at in situ temperature in the dark for 5, 7, and 10 h. Controls were fixed with PFA (2% final concentration) 10 min prior to the addition of chitin. Incubations were terminated by the addition of PFA (2% final concentration). Samples were filtered onto 0.2 µm nitrocellulose membranes (25 mm diameter; Nuclepore), and the filters rinsed with 0.2 µm filtered Milli-Q water (Beier and Bertilsson, 2011) before determining the radioactivity by liquid scintillation. Chitin uptake rates were estimated from the linear regression of radioactivity against incubation time. The calculated chitin turnover times were plotted against the added concentrations of chitin. The slope of the linear regression provides an estimate of the chitin turnover rate (r <sup>2</sup> = 0.918 for E-3; r <sup>2</sup> = 0.944 for E-4W; r <sup>2</sup> = 0.532 for F-L; r <sup>2</sup> = 0.912 for A3-2).

Fe uptake rates were determined in triplicates plus one formol-killed control at all stations following 24 h incubations with <sup>55</sup>Fe, when a steady state was achieved (Fourquez et al., 2015). To measure intracellular Fe uptake, cells were collected on nitrocellulose filters (Nuclepore) and immediately rinsed first with the Ti–citrate–EDTA washing solution and then with 0.22 µm filtered seawater as detailed in Fourquez et al. (2015). The radioactivity on filters was counted by liquid scintillation after they dissolved for 24 h in 10 mL of Filter-Count scintillation cocktail (PerkinElmer). Fe uptake rates were calculated considering in situ Fe concentrations and corrected for background radioactivity (see Fourquez et al., 2015).

For each uptake experiment, the radioactivity (in disintegration per minute, DPM) was determined using the Tricarb <sup>R</sup> scintillation counter and the DPMs of the controls were subtracted from the live incubations.

#### Statistical Analyses

We used Pearson correlation to determine the relationships between (1) the relative contributions of prokaryotic groups to total abundance and to total substrate active-cells and (2) the percentages of substrate-active cells among different groups. To test whether the percentages of substrate-active cells and their contributions to the substrate-active community were significantly different among prokaryotic groups we applied the Kruskall-Wallis one-way analysis of variance on ranks and the post-hoc Dunn's test if a significant difference was detected by the former. We used this non-parametric test because of the unequal number of observations among prokaryotic groups. We used Pearson correlation to determine the relationships between the environmental parameters measured at each sampling location and the percent of total and group-specific active cells assessed by MICRO-CARD-FISH. This analysis was done for the three substrates of the study. Values for bacterial abundance, Chl a, inorganic nutrients, and bacterial heterotrophic production were log transformed and the percentage of active cells was arcsine transformed for this analysis. Transformed data were according to the Shapiro-Wilk test still not significantly normally distributed, but visual examination of the data revealed that their distribution after transformation was closer to the normal distribution. We performed our test on transformed data using Pearson correlation. We also tested for correlation on both, raw and transformed data using Spearman-rank correlation analysis. No difference in the result was obtained by these statistical approaches.

# RESULTS

#### Environmental Context

The reference Station R-2, located west of the Kerguelen Plateau, revealed low dissolved Fe (DFe, 0.13 ± 0.05 nM) and chlorophyll a concentrations (Chl a, 0.3 ± 0.1 µg L−<sup>1</sup> ) that are typical for surface HNLC waters (**Table 1**). At this station, prokaryotic abundance, prokaryotic biomass production and respiration (Christaki et al., 2014) and prokaryotic growth rates were lower than those in surface waters of the Fe-fertilized stations (**Table 1**). A patchwork of phytoplankton blooms induced by natural Fefertilization developed east of Kerguelen Island. These blooms varied in terms of phytoplankton biomass and community composition (Lasbleiz et al., 2016) reflecting different stages of development. Above the plateau, at Station A3-2, concentrations of DFe in the surface mixed layer (0.16 ± 0.03 nM) were slightly higher than in HNLC waters, and concentrations of Chl a exceeded those in HNLC waters by 6-fold (2.0 ± 0.03 µg L−<sup>1</sup> ), reflecting a well-developed spring bloom at this site, sustained by recent and persistent Fe-fertilization (Trull et al., 2015). The response of heterotrophic microbes was reflected by 6-fold higher bacterial growth rates (0.12 d−<sup>1</sup> ) than in HNLC surface waters (0.02 d−<sup>1</sup> ; **Table 1**). At Station E4-W, DFe concentrations (0.17 ± 0.03 nM) were similar to those at Station A3-2, but Chl a concentrations (1.3 ± 0.1 µg L−<sup>1</sup> ) and prokaryotic growth rates (0.09 d−<sup>1</sup> ) remained lower than at the Plateau station. The highest Chl a concentration (4.0 ± 1.6 µg L−<sup>1</sup> ) was observed at Station F-L, located north of the Polar Front. The concurrent relatively low concentrations of nitrate (20 µM as compared to 25 µM at all other sites) indicate an intense bloom event, induced by recent and brief Fe-fertilization (D'Ovidio et al., 2015). DFe concentrations at Station F-L (0.22 ± 0.06 nM) were among the highest determined in the study region (**Table 1**), and prokaryotic abundance and growth rates (0.21 d−<sup>1</sup> ) exceeded those observed at all other stations in the Fe-fertilized region. Fefertilization appears to be less recent but more persistent within the recirculation feature south of the Polar Front (D'Ovidio et al., 2015). In this regard, the Stations E-1, E-3, E-4E and E-5 were sampled in a quasi Lagrangian manner and can be considered to some extent as a succession of sampling in the same water mass over time (D'Ovidio et al., 2015). Concentrations of Chl a increased from 0.6 to 1.2 µg L−<sup>1</sup> , accompanied by increases in prokaryotic growth rate from 0.07 to 0.12 d−<sup>1</sup> . The unchanged low concentrations of dissolved organic carbon (mean 49 ± 1µM; Obernosterer et al., 2015; Tremblay et al., 2015) across all sampling sites are likely a result of the rapid consumption of phytoplankton-derived DOM.

Bulk uptake rates of leucine ranged between 0.9 and 28 pmol L <sup>−</sup><sup>1</sup> h −1 (n = 8; Christaki et al., 2014) and those of chitin varied between 0.6 and 5.5 pmol L−<sup>1</sup> h −1 (n = 4, **Table 2**). Prokaryotic Fe uptake rates were substantially lower (0.21–0.68 pmol L−<sup>1</sup> h −1 , n = 4; Fourquez et al., 2015, **Table 2**). Similarly, the percent DAPI-cells that were substrate-active was highest for leucine (range 11–56%) and varied between 3 and 10% for chitin and between 1 and 4 % for Fe.

#### Abundance of Prokaryotic Groups

Our CARD-FISH results, using probes at varying phylogenetic levels, reveal an overall similar community composition at all sites (**Figure 2A**). The dominant bacterial groups were SAR11 (range 34–61% of total DAPI-stained cells) and CFB (range 42–48%, except Station E-1: 25% of total DAPI-cells) (**Figure 2B**). Roseobacter (3–10% of total DAPI-cells) and the Gammaproteobacterial groups SAR86 (10–13%, except E-4E: 19% of total DAPI-cells) and SAR92 (0.4–6% of total DAPIcells) revealed lower relative abundances. SAR86 accounted for a large fraction of Gammaproteobacteria (55–95%) at all sites. The relative abundance of Archaea varied between 2 and 19% of total DAPI-cells and Crenarchaeota accounted for 1–63% of Archaeal abundance at the 4 sites considered (**Figure 2B**). Interestingly our analyses revealed higher abundances of SAR92 and Archaea at the Fe-fertilized sites as compared to the HNCL-site R-2.

#### Contribution of Prokaryotic Groups to the Substrate Active Community

The single-cell approach MICRO-CARD-FISH provides two lines of informations: First, the contribution of a prokaryotic group to the bulk substrate active community and second, the fraction of substrate active cells within a prokaryotic group. We observed a pronounced positive relationship between the contribution to abundance and to the leucine-active community (p < 0.0001; r = 0.93; n = 47) across all sites and including all prokaryotic groups (**Figure 3A**). By contrast, different patterns emerged for chitin and Fe. We observed both more (as indicated by data points above the 1:1 line in **Figure 3**) and less (as indicated by data points below the 1:1 line **Figure 3A**) singlecell use of these substrates by the prokaryotic groups than their abundance would suggest. A less pronounced, but still significant correlation was found between these two variables for chitin (p < 0.001; r = 0.51; n = 38), while no significant correlation was found for Fe (p > 0.05; r = 0.32; n = 18).

To better understand this variability among the three substrates, we present the same data set by highlighting the phylogenetic identity of the bacterial and archaeal groups (**Figure 3B**). Despite the overall similar composition of the community among stations (**Figure 2A**), the contribution of distinct bacterial groups and Archaea to the uptake of chitin and Fe varied considerably. As indicated by the positive correlation, the contribution of specific prokaryotic groups to leucine uptake (**Figure 4A**) was overall reflected by their respective contributions to bulk abundance: in both, the total and the leucine active community, SAR11 represented the highest fraction of community members followed by CFB (**Figure 2B**). The contribution of SAR11 to the leucine active community (mean 54 ± 11%, n = 8) was significantly higher than that of all the other prokaryotic groups (mean 2–34%, n = 4–8; p < 0.001), except that of CFB (mean 37 ± 7%; n = 8). The contribution of CFB was significantly higher than the contributions of Roseobacter and Crenarcheaota (p < 0.001).

The chitin and Fe-active communities revealed, however, different patterns than the overall community composition. The chitin active community was dominated by CFB and Archaea, but not by SAR11 (**Figure 4A**). The contribution of CFB (mean 66 ± 15, n = 6) was significantly higher than that of Roseobacter (mean 9 ± 9, n = 6), SAR86 (mean 2 ± 3, n = 5) and Crenarchaeota (mean 4 ± 6, n = 4) (p < 0.001). In contrast, Gammaproteobacteria (mean 53 ± 16, n = 4) and CFB (mean 47 ± 18, n = 4) accounted for the largest fractions of the Fe-active community.

# Fraction of Cells That Are Substrate Active

The fraction of substrate-active cells within a prokaryotic group revealed considerable variations among the sampling sites, and this was most pronounced for leucine (**Figure 4B**). This could be due to the overall higher percentage of DAPI-cells incorporating this substrate (**Table 2**). The mean proportion of leucine-active cells varied between 15% (Crenarchaeota) and 44% (SAR11). The fractions of cells taking up chitin revealed more pronounced differences among groups. It was interesting to note that the fractions of chitin-active Roseobacter, Archaea and Crenarchaeota (means of 8–10%) were higher than those of Gammaproteobacteria, more specifically SAR86, and SAR11 (means of 1–2%). With respect to Fe, Gammaproteobacteria and Roseobacter had slightly higher fractions of active cells (means of 4%) than SAR11 and CFB (means of 1–2%). However, due to the large variability across sites, none of these differences were significantly different.


For DAPI-cells, mean values ± SD of several filter pieces are given. For leucine uptake rates, mean values ± SD of triplicate incubations are given. n.d., not determined. <sup>a</sup>From Christaki et al. (2014).

<sup>b</sup>From Fourquez et al. (2015).

available for the E-sites. The black and the dashed lines across the box represent respectively the median and the mean. The ends of the boxes define the 25th and 75th percentiles and error bars define the 10th and 90th percentiles (data pooled from all stations shown in A). Outliers are plotted in black dots. Values for "other Archaea" were calculated as the difference between Archaea and Crenarchaeota. ARCH, Archaea; CREN, Crenarchaeota; CFB, Cytophaga/Flexibacter/Bacteroides; ROSEO, Roseobacter clade; SAR11, SAR11 cluster; SAR 86, SAR86 cluster; SAR92, SAR92 cluster.

We further explored what factors could explain the variability in the fraction of substrate-active cells across sites. We did not obtain any significant correlation between environmental variables (bacterial abundance, Chl a, inorganic nutrients, and bacterial heterotrophic production) and the substrate-specific single-cell activities of the different groups. However, we did obtain several significant correlations between the percent of active cells among different groups for a given substrate (**Figure 5**). Interestingly, the majority of the correlations were associated with CFB. The number CFB chitin-active cells was positively correlated with that of Gammaproteobacteria and of SAR11, and the number of CFB leucine-active cells was positively correlated with that of Gammaproteobacteria. The number of CFB Fe-active cells was positively correlated with that of Roseobacter.

#### DISCUSSION

By investigating the single-cell activities of several prokaryotic groups, our study provides a new perspective on the microbial cycling of carbon and Fe in the Southern Ocean. We observed that all investigated prokaryotic groups assimilated Fe and the two distinct carbon substrates leucine and chitin. While the contribution of the prokaryotic groups to the leucine-active community was strongly driven by their relative abundances, their contributions to chitin and Fe uptake appear to be uncoupled from abundance. Our results illustrate pronounced differences in the composition of the leucine, chitin and Fe-active communities and also suggest a preferential use of chitin and Fe. We discuss that interactions among groups in the access of these substrates could explain this pattern.

Our observation of the assimilation of the amino acid leucine by all major prokaryotic groups confirms a large number of studies in temperate and polar environments (see for overviews del Giorgio and Gasol, 2008; Kirchman, 2016). The strong correlation between the contribution to abundance and to the bulk leucine-active community could indicate a non-preferential uptake of this substrate and bottom-up control of heterotrophic prokaryotic activity (Cottrell and Kirchman, 2003). The present study took place in early spring when prokaryotic heterotrophic activity is greatly stimulated by the release of phytoplanktonderived DOM in the study region (Christaki et al., 2014), which in turn affects community composition (Landa et al., 2015). Both C and Fe were identified as growth-limiting factors (Obernosterer et al., 2015). Interestingly, the contributions of several bacterial groups at a fine phylogenetic level (>99% identity of the 16S rRNA gene) to abundance and leucine incorporation were also strongly positively correlated during the declining phase of the bloom above the Kerguelen Plateau (Obernosterer et al., 2011), concomitant with high virus-mediated prokaryotic mortality (Malits et al., 2014). These results could suggest that the access to the limiting resource has a stronger shaping effect on the prokaryotic community composition than top-down control in these perennially cold waters.

In contrast to readily bioavailable low molecular weight substrates, the assimilation of polymers requires more specialized pathways, and the lack of these in some prokaryotic taxa could lead to the use of these substrates by a smaller range of groups (Zimmerman et al., 2013). In the present study, we used the dimer N-Diacteylglucosamine, composed of the two amino sugars N-acetylglucosamines (NAG), as a model for an abundant polymeric compound in the ocean (Souza et al., 2011). This substrate is taken up and subsequently cleaved intracellularly in the periplasmatic space or in the cytoplasm by β-N-acetyl-hexosaminidases into NAG (Beier and Bertilsson, 2013). Even though the uptake of the soluble chitin dimer N-Diacteylglucosamine does not prove the active involvement of cells into chitin polymer degradation, an earlier study could demonstrate that members of bacterial groups that incorporated this substrate at high frequencies were typically also highly abundant in the biofilm attached to chitin particles (Beier and Bertilsson, 2011).

Our observation that CFB are key players during the degradation of chitin confirms previous MICRO-CARD-FISH studies in marine (Cottrell and Kirchman, 2003) and freshwater systems (Beier and Bertilsson, 2011). In chitin-amended microcosms performed in the Southern Ocean the growth of Bacteroidetes, more specifically the genus Reichenbachiella, was stimulated (Wietz et al., 2015). Whole genome sequencing of several strains belonging to this phylum provide the metabolic underpinning of these observations (Bauer et al., 2006; Tang et al., 2012; Kabisch et al., 2014). Interestingly, CFB were the only group that had high contributions to both the chitin- and Feactive communities in our study. This feature could represent an ecological advantage of CFB that could in part explain their high abundances in surface waters of the Southern Ocean (present study; Simon et al., 1999; Obernosterer et al., 2011) characterized by low concentrations in bioavailable Fe and C.

A more surprising result is the relatively high fraction of chitin-active Archaea, and thus their potential role in the degradation of this substrate. Chitinase genes were detected in the genomes of several crenarchaeotal and euryarchaeotal strains (Staufenberger et al., 2012; García-Fraga et al., 2014, and

percentiles and error bars define the 10th and 90th percentiles. Outliers are plotted in black dots. ARCH, Archaea; CREN, Crenarchaeota; CFB, Cytophaga/Flexibacter/Bacteroides; ROSEO, Roseobacter clade; SAR11, SAR11 cluster; SAR86, SAR86 cluster; SAR92, SAR92 cluster.

references therein), but to the best of our knowledge not yet in the abundant marine group I Thaumarchaeota or marine group II (belonging to Euryarchaeota) that are likely abundant among the detected cells in the current study. Crenarchaeota and Euryarchaeota were both shown to assimilate a range of organic substrates and CO<sup>2</sup> (Teira et al., 2004; Herndl et al., 2005; Kirchman et al., 2007). Marine group I Thaumarchaeota are considered to be chemolithotrophs and more abundant in deeper layers (Qin et al., 2014; Zhang et al., 2015). In contrast, marine group II Euryarchaeota are mostly heterotrophs and abundant in ocean surface water. Marine group II were suggested to live as particle-attached polymer degraders (Iverson et al., 2012). In line with this current knowledge concerning the ecology of Archaea it seems reasonable to believe that, even if so far no chitinase genes within the three sequenced genomes of marine group II were discovered, some members of this group could be active chitin degraders. Our results demonstrate that members of both Euryarchaeota and Crenarchaeota, and most likely Thaumarchaeota, are involved at least in the final steps of the complex process of chitin degradation.

Copepods are considered the most important source of chitin in the environment (Souza et al., 2011). Exoskeletons, including molts from copepods and marine invertebrate larvae contain chitin (Gooday, 1990) and they could represent one source of this polymer in our study. Mesozooplankton responded to the early stage of the phytoplankton bloom with the production of larvae of copepods and euphausiids (Carlotti et al., 2015). The overall biomasses remained, however, low as compared to those observed during the declining phase of the Kerguelen bloom in late summer (Carlotti et al., 2008). Another potential source that has been given less attention are diatoms, the dominant phytoplankton in the study region (Lasbleiz et al., 2014). In surface waters, heterotrophic prokaryotic activity was strongly driven by phytoplankton-derived organic matter (Christaki et al., 2014; Landa et al., 2015). Whether chitin is provided as growth substrate by diatoms is a particularly interesting question in this context. The genus Thalassiosira was reported to contain chitin synthetase genes, implying chitin production (Durkin et al., 2009). In the present study, small centric diatoms (<25µm) principally composed of several Thalassiosira species were present at all sites, but accounted for about 60% of total diatom biomass at the Polar Front Station F-L (Laurenceau-Cornec et al., 2015; Lasbleiz et al., 2016). At this site, we observed the highest bulk chitin turnover rates concomitantly with high archaeal and crenarchaeotal abundances (19 and 5% of total DAPI-cells, respectively) and cell-specific chitin activities (14 and 16% of chitin active archaeal and crenarchaeotal cells, respectively) that were higher than at any other site. Whether chitin production and consumption favors associations between specific groups of phytoplankton and Archaea as observed for diverse bacterial taxa (Amin et al., 2012; Durham et al., 2015) remains to be explored. Chitin was found to support between 5% to up to 30% of the bacterial production in the Delaware Estuary (Kirchman and White, 1999). These quantitative estimates together with the observation that all investigated prokaryotic groups contributed to chitin uptake further confirm that this compound is a widespread bioavailable source of organic carbon and nitrogen for prokaryotic growth in the ocean.

Whether Fe uptake strategies and demand differ among microbial taxa and how community composition might in turn affect Fe cycling in the Southern Ocean is an essential, but largely unresolved question. A survey of representative marine bacteria revealed the presence of several pathways for Fe acquisition in most genomes, while siderophore biosynthesis genes were reported to be rare (Hopkinson and Barbeau, 2012). Global metagenomic surveys point to spatial differences in Fe metabolism strategies, including uptake and storage and these studies established indirect links to the distribution of microbial taxa (Desai et al., 2012; Toulza et al., 2012). These genomic based observations complement earlier laboratory experiments demonstrating the capability of marine bacterial strains to consume Fe in different forms, such as inorganic Fe, organic ligand-bound Fe (Granger and Price, 1999; Guan et al., 2001; Weaver et al., 2003) or as heme (Hopkinson et al., 2008), and to produce siderophores (Yamamoto et al., 1994; Martinez et al., 2000). Interestingly, Pelagibacter, a member of the most abundant bacterial group in the study region, lacks entirely most Fe transport and siderophore biosynthesis genes, and thus depend on inorganic Fe uptake (Hopkinson and Barbeau, 2012). According to metagenomic and transcriptomic analyses this genera has non-specialized Fe uptake systems, largely lacks the ability to take up Fe chelates directly and would rather use inorganic Fe (Smith et al., 2010; Thompson et al., 2011). In Candidatus Pelagibacter ubique, the presence of Fe3<sup>+</sup> transporters and the lack of TonB-dependent transporters is suspected to be a niche specific adaptation (Smith et al., 2010).

Our study is the first to investigate Fe uptake by different bacterial groups. Because of the low specific activity of <sup>55</sup>Fe, the detection of silver grains is dependent on the cellular <sup>55</sup>Fe quota (Fourquez et al., 2012). This methodological constraint probably explains the overall low fraction of cells associated with silver grains. The Fe-positive fraction could represent cells with highest Fe uptake activities or requirements, or those cells that have potential Fe-storage capacities. We observed in the present study that all investigated bacterial groups assimilated the micronutrient Fe in a range of environmental conditions, but our study highlights Gammaproteobacteria, in particular SAR86, and CFB to be important players. TonBdependent transporters were particularly abundant in members belonging to Gammaproteobacteria and CFB across the GOS data set (Tang et al., 2012). These transporters promote the uptake of diverse substrates present at low concentrations, and they have high affinity with Fe complexes (Moeck and Coulton, 1998). In the region of the present study where Fe and carbon are limiting nutrients (Obernosterer et al., 2015), the presence of TonB-dependent transporters could provide an advantage to members of these bacterial groups. Previous studies report strain-specific differences in the cellular Fe quota under both Fe replete and Fe limited conditions (Granger and Price, 1999; Fourquez et al., 2014). The question, however, of whether Fe quota differ among phylogenetically diverse bacterial groups were not addressed in these studies. How variable this basic physiological characteristic is among taxa and how this affects microbial Fe and carbon cycling needs further investigation.

For the three compounds used in the present study we observed high variability in the fraction of substrate-active cells across sites. None of the environmental variables could explain this observation. Possibly small changes in the concentrations of bioavailable DOM and Fe, not measurable with present chemical techniques, could in part account for the observed variability. Changes in within group community composition could also play a role. Using 454 pyrosequencing, the bacterial communities revealed pronounced differences among sites (Landa et al., 2015), suggesting the metabolic capacities for the uptake of the three substrates to vary at finer phylogenetic levels than those resolved via the probes used in this study. An interesting hypothesis in this context is the possible interaction among bacterial groups as suggested by the positive relationships between substrate-active cells belonging to different bacterial groups. The temporal succession of prokaryotic communities driven by specific metabolic capabilities and substrate preferences could represent synergistic interactions in the degradation of DOM (McCarren et al., 2010; Teeling et al., 2012).

Interactions among taxa might be of particular relevance if "leaky" components are involved that are available extracellularly and provided by the activity of a restricted group of community members (Morris et al., 2012). In contrast to leucine, for both, Fe via the active release of siderophores (Granger and Price, 1999; Martinez et al., 2003; Hopkinson and Barbeau, 2012; Morris et al., 2012) and chitin via the extracellular production of hydrolyses products of the chitin polymer from released chitinases (Beier and Bertilsson, 2011), species interactions may therefore be specifically favored. How these and possibly other types of interactions might facilitate the utilization of the limiting resources carbon and Fe in the Southern Ocean and thereby affect the cycling of these elements is a challenging question for future studies.

#### AUTHOR CONTRIBUTIONS

Contributed to conception, design, and acquisition of data: MF, EJ, RH and IO. Contributed to analysis and interpretation of data: MF, IO, SB, EJ, and RH. Drafted and revised the article: MF, IO, and SB. Approved the submitted version for publication: MF, IO, SB, EJ, and RH.

#### ACKNOWLEDGMENTS

We a grateful to S. Blain, the PI of the KEOPS2 project for providing us the opportunity to participate to this cruise. We thank the chief scientist B. Quéguiner, the captain Bernard Lassiette and the crew of the R/V Marion Dufresne for their enthusiasm and help aboard. S. Bertilsson kindly provided radiolabeled N-Diacteylglucosamine. Thanks to P. Catala and S. Rachik for their help with MICRO-CARD-FISH and microscopic analyses. Two reviewers provided constructive comments on a previous version of this manuscript. We also thank Gustave Chiffard for insightful discussions during the development of this project. This work was supported by the French Research program of the INSU-CNRS LEFE−CYBER (Les enveloppes fluides et l'environnement– Cycles biogéochimiques, environnement et ressources), the French ANR (Agence Nationale de la Recherche, SIMI-6 program), the French CNES (Centre National d'Etudes Spatiales) and the French Polar Institute IPEV (Institut Polaire Paul−Emile Victor).

#### REFERENCES


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Fourquez, Beier, Jongmans, Hunter and Obernosterer. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Viruses and Protists Induced-mortality of Prokaryotes around the Antarctic Peninsula during the Austral Summer

Dolors Vaqué<sup>1</sup> \* † , Julia A. Boras 1 †, Francesc Torrent-Llagostera<sup>1</sup> , Susana Agustí <sup>2</sup> , Jesús M. Arrieta<sup>2</sup> , Elena Lara1, 3, Yaiza M. Castillo<sup>1</sup> , Carlos M. Duarte<sup>2</sup> and Maria M. Sala<sup>1</sup>

1 Institut de Ciències del Mar (CSIC), Consejo Superior de Investigaciones Científicas, Barcelona, Spain, <sup>2</sup> King Abdullah University of Sciences and Technology, Thuwal, Saudi Arabia, <sup>3</sup> Institute of Marine Sciences (CNR-ISMAR), National Research Council, Venezia, Italy

#### Edited by:

Julie Dinasquet, University of California, San Diego, USA

#### Reviewed by:

Urania Christaki, Ministry of Research and Higher Education, France Sigitas Šulcius, ˇ Linnaeus University, Sweden

#### \*Correspondence:

Dolors Vaqué dolors@icm.csic.es † These authors have contributed equally to this work.

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Microbiology

Received: 11 October 2016 Accepted: 03 February 2017 Published: 02 March 2017

#### Citation:

Vaqué D, Boras JA, Torrent-Llagostera F, Agustí S, Arrieta JM, Lara E, Castillo YM, Duarte CM and Sala MM (2017) Viruses and Protists Induced-mortality of Prokaryotes around the Antarctic Peninsula during the Austral Summer. Front. Microbiol. 8:241. doi: 10.3389/fmicb.2017.00241 During the Austral summer 2009 we studied three areas surrounding the Antarctic Peninsula: the Bellingshausen Sea, the Bransfield Strait and the Weddell Sea. We aimed to investigate, whether viruses or protists were the main agents inducing prokaryotic mortality rates, and the sensitivity to temperature of prokaryotic heterotrophic production and mortality based on the activation energy (Ea) for each process. Seawater samples were taken at seven depths (0.1–100 m) to quantify viruses, prokaryotes and protists abundances, and heterotrophic prokaryotic production (PHP). Viral lytic production, lysogeny, and mortality rates of prokaryotes due to viruses and protists were estimated at surface (0.1–1 m) and at the Deep Fluorescence Maximum (DFM, 12–55 m) at eight representative stations of the three areas. The average viral lytic production ranged from 1.0 ± 0.3 × 10<sup>7</sup> viruses ml−<sup>1</sup> d −1 in the Bellingshausen Sea to1.3 ± 0.7 × 10<sup>7</sup> viruses ml−<sup>1</sup> d −1 in the Bransfield Strait, while lysogeny, when detectable, recorded the lowest value in the Bellingshausen Sea (0.05 ± 0.05 × 10<sup>7</sup> viruses ml−<sup>1</sup> d −1 ) and the highest in the Weddell Sea (4.3 ± 3.5 × 10<sup>7</sup> viruses ml−<sup>1</sup> d −1 ). Average mortality rates due to viruses ranged from 9.7 ± 6.1 × 10<sup>4</sup> cells ml−<sup>1</sup> d −1 in the Weddell Sea to 14.3 ± 4.0 × 10<sup>4</sup> cells ml−<sup>1</sup> d −1 in the Bellingshausen Sea, and were higher than averaged grazing rates in the Weddell Sea (5.9 ± 1.1 × 10<sup>4</sup> cells ml−<sup>1</sup> d −1 ) and in the Bellingshausen Sea (6.8 ± 0.9 × 10<sup>4</sup> cells ml−<sup>1</sup> d −1 ). The highest impact on prokaryotes by viruses and main differences between viral and protists activities were observed in surface samples: 17.8 ± 6.8 × 10<sup>4</sup> cells ml−<sup>1</sup> d <sup>−</sup><sup>1</sup> and 6.5 ± 3.9 × 10<sup>4</sup> cells ml−<sup>1</sup> d −1 in the Weddell Sea; 22.1 ± 9.6 × 10<sup>4</sup> cells ml−<sup>1</sup> d <sup>−</sup><sup>1</sup> and 11.6 ± 1.4 × 10<sup>4</sup> cells ml−<sup>1</sup> d −1 in the Bransfield Strait; and 16.1 ± 5.7 × 10<sup>4</sup> cells ml−<sup>1</sup> d <sup>−</sup><sup>1</sup> and 7.9 ± 2.6 × 10<sup>4</sup> cells ml−<sup>1</sup> d −1 in the Bellingshausen Sea, respectively. Furthermore, the rate of lysed cells and PHP showed higher sensitivity to temperature than grazing rates by protists. We conclude that viruses were more important mortality agents than protists mainly in surface waters and that viral activity has a higher sensitivity to temperature than grazing rates. This suggests a reduction of the carbon transferred through the microbial food-web that could have implications in the biogeochemical cycles in a future warmer ocean scenario.

Keywords: viruses, prokaryotes, protists, lysis, lysogeny, mortality, temperature, Antarctic waters

# INTRODUCTION

To understand the functioning of the biogeochemical carbon cycles in extreme marine systems, such as the Antarctic Ocean, it is essential to elucidate the relation between the prokaryotic members of the community and their predators, protists and viruses, since their impact in the carbon cycles are different. While protists channel prokaryotic carbon to higher trophic levels through the grazing process, viruses return dissolved and small particulate prokaryotic carbon forms from the lysed cells (Wilhelm and Suttle, 1999) to the water column (the viral shunt), and may modify the efficiency of the carbon pump (Suttle, 2007). Indeed, viruses contribute to generate new substrates for prokaryotes, increasing respiration (Eissler et al., 2003), nutrients regeneration (Gobler et al., 1997), and being crucial for Fe regeneration, an important trace element for phytoplankton growth (Strzepek et al., 2005; Evans and Brussaard, 2012).

The ecology of prokaryotes and protists in Antarctic waters has been well studied since the 1980's (e.g., Karl et al., 1991; Kuparinen and Bjørsen, 1992; Leaky et al., 1996; Vaqué et al., 2002a,b; Christaki et al., 2008) but less is known on the role of viruses (Danovaro et al., 2011). The few studies focused on viral infection of prokaryotes were carried out mainly in waters adjacent to the Antarctic Peninsula (Guixa-Boixereu et al., 2002), in coastal regions of Antarctica (Pearce et al., 2007), the Weddell Sea and Polar Frontal zones, and in the Southern Ocean including Sub-Antarctic areas (Bonilla-Findji et al., 2008; Evans et al., 2009; Evans and Brussaard, 2012; Malits et al., 2014). All those studies point toward a high viral lytic production and suggest a major role of viruses in prokaryotic mortality. However, if we aim at elucidating which is the main source of prokaryotic mortality both viral production and protistan grazing should be measured simultaneously, what has been done only in two studies: Guixa-Boixereu et al. (2002) in surface samples around the Antarctic Peninsula and Christaki et al. (2014) in Sub-Antarctic waters. In both cases they reported that mortality rates caused by viruses were higher (at least in some periodes of the year) than those caused by protists, suggesting that the viral shunt would drive the destiny of the prokaryotic carbon in those regions.

The impact of viruses and protists on prokaryotes reported in Guixa-Boixereu et al. (2002) during the Austral summer in the Bellingshausen Sea and the Gerlache Strait was very high, removing ≥ 50% d−<sup>1</sup> of its biomass and more than 100% d−<sup>1</sup> of its heterotrophic production, which suggests that prokaryotes were top-down controlled. Then, to maintain a sustainable prokaryotic heterotrophic production and biomass, it would imply that DOC should not be a limiting factor for prokaryotes to grow. Indeed, Pedrós-Alió et al. (2002), point out that during the Austral Summer in the Antarctic Ocean prokaryotic production was not constrained by DOC.

Finally, according to the available literature, values of prokaryotic production and mortality rates in Antarctic waters do not differ much from those found in temperate systems (e.g., Guixa-Boixereu et al., 2002; Boras et al., 2009), although microbial processes in the Antarctic occur at very low temperatures. However, an increase of temperature in marine ecosystems, including Polar Regions, enhances heterotrophic prokaryotic production and mortality rates (Vázquez-Domínguez et al., 2007; Vaqué et al., 2009; Lara et al., 2013). Sarmento et al. (2010) compiled data of prokaryotic production and grazing rates from temperate systems to Antarctic waters, and have shown a higher sensitivity to temperature for prokaryotic production than for mortality rates. The same was found by Maranger et al. (2015), who included also the rate of lysed prokaryotes for the Arctic Ocean into the dataset. It is then important to elucidate the sensitivity of these processes in the Antarctic Ocean, in which this information is lacking, since they may have consequences in the fate of the microbial carbon in a warmer ocean.

In the present study, we aimed to test the following working hypotheses: (a) since high viral production rates were detected previously in polar systems, the mortality caused by viruses should be at least as important or higher than the protistan impact in Antarctic waters; (b) taking into account that DOC is not a limiting factor for prokaryotic production during the Austral summer, we expected that prokaryotes are top-down controlled at this time of the year; (c) prokaryotic production, viral infection and grazing by protists would show different sensitivity to temperature changes as suggested by study in the Arctic Ocean. To verify these hypotheses we evaluated prokaryotic losses due to viruses and protists in three different areas around the Antarctic Peninsula (the Weddell Sea, the Bransfield Strait and the Bellingshausen Sea). We elucidated the impact of predators on the prokaryotic community (topdown control) and we assessed the sensitivity of the microbial processes (prokaryotic production and mortalities) to the different temperatures recorded at the three visited areas. This study will contribute to add new information on the fate of prokaryotic carbon within the microbial food-web, and how the impact of temperature on heterotrophic prokaryotes production and losses, which will affect the biogeochemical cycles, could generate new hypothesis in a warmer ocean scenario.

#### MATERIALS AND METHODS

#### Sampling Sites

A cruise was carried out on board of the R/V BIO Hespérides from January 28 to February 25, 2009, around the Antarctic Peninsula. The studied area and sampling sites are shown in **Figure 1** and Table 1SM. We sampled three representative Antarctic areas: (1) the Weddell Sea (WS); (2) the western basin of the Bransfield Strait (BrS); and (3) the Bellinghausen Sea (BeS). These areas were characterized according to their location, physicochemical properties as well as water circulation (Garcia et al., 2002; Gòmis et al., 2002; Hellmer et al., 2011).

#### Physicochemical Variables and Microbiological Biomasses

Profiles of salinity, temperature, and units of fluorescence (as a proxy of chlorophyll a concentration) were obtained using a CTD EG&G model MkIIIC WOCE between 1 and 100 m depth (Table 1SM). Samples for the microbiological parameters were taken from 0.1 or 1 m to 100 m, at seven depths: from two to three

above the DFM (deep fluorescence maximum), one at the DFM, and from two to three below the DFM (Table 1SM). Samples from 0.1 m were collected directly from a rubber boat, and for the other six depths with 12 L Niskin bottles attached to a rossette sampler system. Due to weather conditions, no measurements were done for 0.1 m at stations 4, 11, 15, 23, and 27. Viruses and prokaryotes abundances were measured at all depths and at all stations (Table 1SM). Subsamples for viral abundances (2 ml) were fixed with glutaraldehyde (0.5% final concentration), quick frozen in liquid nitrogen as described by Brussaard (2004) and stored at −80◦C. Subsamples for prokaryote abundances (2 ml) were fixed with paraformaldehyde (1% final concentration). Virus samples were stained with SYBR- GreenI and analysed as described in Brussaard (2004). Prokaryote samples were stained with SYTO13 according to the described in Gasol and del Giorgio (2000), and were run using 0.92-µm yellow-green latex beads as an internal standard. Viral and prokaryotes counts were made on a FACSCalibur (Becton & Dickinson) flow cytometer, back in the Institut de Ciències del Mar (ICM) lab. Virusprokaryote ratio VPR was calculated dividing the in situ viral abundance by in situ prokaryotic abundance. Viral biomass was calculated using the factor 1 × 10−<sup>16</sup> g C virus <sup>−</sup><sup>1</sup> described by Børsheim et al. (1990), and prokaryotic biomass was estimated using the carbon-to-volume relationship equation derived by Norland (1993) from the data of Simon and Azam (1989): pg C cell−<sup>1</sup> = 0.12 pg x (µm<sup>3</sup> cell−<sup>1</sup> ) 0.7. We assumed an average prokaryote cell volume of 0.047 µm<sup>3</sup> cell−<sup>1</sup> measured in similar Antarctic waters (Vaqué et al., 2002a, 2009). Heterotrophic (HF) and phototrophic (PF) pico/nanoflagellates abundances were measured at eight selected stations, at depths where prokaryotic mortality rates were recorded (Table 1SM). They were counted by epifluorescence microscopy (Olympus BX40-102/E at 1,000X) back to the ICM lab. Subsamples (50 ml) were fixed with glutaraldehyde (1% final concentration), filtered through 0.6 µm black polycarbonate filters and stained with DAPI (4,6-diamidino 2-phenylindole) at a final concentration of 5 µg ml−<sup>1</sup> (Sieracki et al., 1985). PF could be distinguished from HF under blue light, as the presence of plastidic structures with red fluorescence in PF could be observed. At least 50–100 cells for HF and PF were counted per filter. They were grouped into 2 size classes: ≤5 µm and >5 µm. Biomass of HF and PF was calculated using a volume to carbon ratio of 0.22 pg C µm−<sup>3</sup> (Børsheim and Bratbak, 1987). Cell volumes were estimated assuming spherical shapes and a diameter of 3 µm, taking into account that 95% of HF and PF had a diameter smaller than 5 µm. Microphytoplankton samples were collected at all stations at the surface (1 m) and at the DFM. 30 ml of each sample was filtered onto 2 µm pore-size black polycarbonate filters, fixed with glutaraldehyde (1% final concentration) and stored frozen at −80◦C until counting under an epifluorescence microscope (Zeiss© Axioplan Imaging) back to the IMEDEA lab. Two majors groups, dinoflagellates and diatoms, were identified. The average cell volume for each phytoplankton group identified during the study was computed using the geometrical approximation of their forms, and the biovolume (µm<sup>3</sup> L −1 ) of the different phytoplankton groups in each sample was calculated as the product of the cell density (cell L−<sup>1</sup> ) multiplied by average cell volume (µm<sup>3</sup> cell−<sup>1</sup> ).

#### Prokaryotic Heterotrophic Production

Prokaryotic heterotrophic production (PHP) was determined at all stations and depths (Table 1SM). It was estimated by the radioactive <sup>3</sup>H-leucine incorporation technique (Kirchman et al., 1985), with the modifications established for the use of microcentrifuge vials (Smith and Azam, 1992). The vials were counted in a Beckman scintillation counter on board. PHP was calculated according to the equation:

$$\text{PHP} = \text{LexxF} \left( \mu \text{gCL}^{-1} \text{day}^{-1} \right),$$

where Leu is the <sup>3</sup>H-leucine incorporation (pmol l−<sup>1</sup> h −1 ) and CF is the conversion factor (1.5 kg C mol Leu−<sup>1</sup> , Kirchman, 1992).

#### Prokaryotic Mortality due to Protists

Grazing rates on prokaryotes by protists (GZ) were evaluated at eight representative stations (Table 1SM) and at two selected depths: surface (0.1 or 1 m) and DFM, using the fluorescently labeled bacteria (FLB) disappearance method (Vázquez-Domínguez et al., 1999). The FLB used in this study were prepared with a culture of Brevundimonas diminuta provided by the Spanish Type Culture Collection (www.uv.es/ cect). The size of these bacteria (0.064 µm<sup>3</sup> cell−<sup>1</sup> ) is comparable to the natural marine prokaryotes recorded in Antarctic waters (0.040–0.070 µm<sup>3</sup> cell−<sup>1</sup> ) (Vaqué et al., 2002a, 2009). The FLB work solutions were prepared in the ICM lab and stored frozen at −20◦C until use. Grazing rate experiments were run in 2-L polycarbonate bottles, in duplicates (1 L of seawater each) plus one prokaryote-free control (1 L of seawater filtered with a 30 kDa cartridge). All bottles (control and duplicates) were inoculated with FLB at 20% of the in situ prokaryote concentration (assessed for this purpose on the on board epifluorescence microscope, BX40-102/E at 1,000X, with the previous staining with DAPI). The added FLB oscillated between 4.1 × 10<sup>4</sup> and 1.6 × 10<sup>5</sup> cells ml−<sup>1</sup> (final concentration). Bottles were incubated in a thermostatic chamber that simulated the in situ temperature, in the dark, for 48 h. Samples for prokaryotes and FLB abundances were taken at the beginning and at the end of the incubations. The abundances were determined with epifluorescence microscopy (Olympus BX40-102/E; 1,000 X magnification) back to the ICM lab, after filtering the sample (20 ml) through 0.2-µm black polycarbonate filters and staining with DAPI at a final concentration of 5 µg ml−<sup>1</sup> (Sieracki et al., 1985). Natural prokaryotes were identified by their blue fluorescence when excited with UV radiation, while FLB were identified by their yellow-green fluorescence when excited with blue light. Protists grazing rates were calculated following the equations of Salat and Marrasé (1994). For details see Boras et al. (2010).

# Viral Production and Prokaryotic Mortality due to Viruses

Viral production (VP) and prokaryotic losses due to viruses (rate of lysed cells, RLC) were measured at the same stations and depths as GZ (Table 1SM). VP was determined following the virus-reduction approach (Weinbauer et al., 2002). This method distinguishes between the production of lytic (VPL), and lysogenic phages (VPLyso), by inducing lysis with mitomycin C. To perform the VP measurements, one liter of seawater was filtered tangentially on the VIVAFlow 200 cartridges to obtain the prokaryote concentrate (40 ml) and the virus-free water as described in details in Boras et al. (2010). A mixture of virusfree water (160 ml) and prokaryote concentrate (40 ml) was prepared and distributed into four sterile 50-ml falcon plastic tubes. Two of the tubes were kept as controls to measure viral lytic production, while mitomycin C (Sigma) was added to the other two tubes as the inducing agent of the lytic cycle in prophages (1 µg ml−<sup>1</sup> final concentration). The tubes were incubated in a thermostatic chamber simulating in situ temperature, in the dark for 12 h. Samples for viral and prokaryotic abundances were collected at time zero and every 4 h of incubation, fixed with glutaraldehyde (0.5% final concentration) and stored as described before for viruses abundance. Viruses and prokaryotes from viral production incubations were counted by flow cytometry, as described above, back to the lab in the ICM. Calculations of viral lytic and lysogenic production were made according to Weinbauer et al. (2002). As part of the prokaryotes is lost during the prokaryotic concentration process, the VP<sup>L</sup> and VPLyso were multiplied by the prokaryote correction factor (Winget et al., 2005, from 1.7 to 10 in our study) to enable the comparison of the VP from different incubations. The number of viruses released by a prokaryote cell (burst size, BS) was estimated from viral production incubations, as in Jiang and Paul (1996) and Boras et al. (2009), and ranged from 12 to 126 viruses per cell.

The RLC was calculated by dividing VP<sup>L</sup> by BS as is described in Guixa-Boixereu (1997):

$$\text{RLC (cells layered ml}^{-1}\text{d}^{-1}) = \text{VP}\_{\text{L}}/\text{BS}$$

The RLC was used also to calculate the % of the prokaryotic standing stock that was lysed by viruses (PSSRLC):

$$\text{\%PSS}\_{\text{RLC}} \text{\(d}^{-1}\text{)}=\text{RLC} \times 100/\text{PSS}$$

Where PSS is the prokaryotic standing stock abundance.

#### Sensitivity of Microbial Processes to Temperature

The temperature sensitivities of the different microbial processes were obtained by estimating the activation energy (Ea) of each one of them, using the Boltzmann-Arrhenius model:

$$\mathbf{B} = \mathbf{B}\_0 \times \mathbf{e}^{\text{(Ea/kT)}}$$

where, B is the basal metabolic rate (i.e., of PHP, GZ, and RLC) and B<sup>0</sup> is a normalization constant independent of body-size and temperature.

The term e(Ea/kT) is the Boltzmann factor that describes the temperature (T, in Kelvin degrees) dependence of a metabolic rate, where k is the Boltzmann's constant (8.62 × 10−<sup>5</sup> eV k−<sup>1</sup> ) and Ea the activation energy of the given rate (West et al., 1997; Brown et al., 2004).

This equation can also be written as:

$$
\ln \mathcal{B} = \ln \mathcal{B}\_0 + \text{Ea} \times 1/\text{kT}
$$

Ea is the slope; obtained when plotting lnB against 1/kT. The steeper is the slope the greater is the sensitivity to temperature changes.

#### Statistical Analyses

All data except temperature and salinity were log transformed. To estimate differences of physicochemical and biological parameters among the Antarctic areas we carried out oneway ANOVA analyses and the subsequent post-hoc Tukey test. Student's t-test was applied to test differences of mortality rates between surface and DFM, and between grazing rates and lysed rates for each station and depth. Regression analyses were carried out between different microbiological parameters. All statistical analyses were performed with the JMP 8.0 program.

#### RESULTS

# Physicochemical and Biological Parameters

The water column was always stratified showing a DFM between 20 and 50 m in the Weddell Sea, 15 and 30 m in the Bransfield Strait, and 12 and 55 m in the Bellingshausen Sea (Table 1SM). ANOVA analyses revealed that in the Weddell Sea temperature and salinity had significant lower values (−0.49◦C ± 0.09, 27.86 ± 0.07, respectively) (**Table 1** and Table 2SM) than the other two areas. Units of Fluorescence (UF, a proxy of phytoplankton biomass), reached the highest value in station 7 (the Weddell Sea, 14.60), but significant differences were only found between the Bransfield Strait (1.66 ± 0.18) and the Bellingshausen Sea (1.10 ± 0.15, **Table 1** and Table 2SM). However, the phototrophic pico/nanoflagellates (PF) did not show significant differences among the three areas (**Table 1**). Microphytoplankton dominated the biomass in the Weddell Sea (66% of total biovolume), and phototrophic nanoflagellates dominated in the other two zones, the Bellingshausen (56% of total biovolume) and the Bransfield (66% of total) areas. Centric diatoms prevailed along the cruise, while pennate forms, mainly represented by Pseudo-Nitzschia Seriata, Pseudo-Nitzschia Delicatissima and Navicula sp., displayed similar biovolumes. We detected in the Weddell Sea an increase in the biovolume of centric diatoms, being mostly represented by Thalassiosira sp. while Corethron sp. and Coscinodiscus sp. were predominant forms in the Bransfield and the Bellingshausen zones.

In the Weddell Sea both, abundances of prokaryotes (3.08 ± 0.16 × 10<sup>5</sup> cell ml−<sup>1</sup> ) and viruses (4.11 ± 0.73 × 10<sup>6</sup> virus ml−<sup>1</sup> ) showed significantly lower values (**Table 1**, Table 2SM) than in the other two areas. In contrast, heterotrophic pico/nanoflagellates (HF) displayed significantly higher values for the Weddell Sea (1.75 ± 0.34 × 10<sup>3</sup> cell ml−<sup>1</sup> ) than the Bellingshausen Sea (0.97 ± 0.01 × 10<sup>3</sup> cell ml−<sup>1</sup> ) (**Table 1** and Table 2SM). Prokaryotic heterotrophic production (PHP) and viral lytic production (VPL) did not show differences among the three areas (ANOVA, p > 0.05), although the highest values for PHP were registered in the Weddell Sea (6.83 × 10<sup>5</sup> cells ml−<sup>1</sup> d −1 ), and for VP<sup>L</sup> in the Bransfield Strait (4.43 × 10<sup>7</sup> viruses ml−<sup>1</sup> d −1 ) (**Table 1**, **Figure 2**). Lysogeny (VPLyso) in the Weddell Sea was registered twice, at the DFM of station 5 (78% of VP) and at the surface of station 7 (83% of VP). In the Bransfield Strait it was found at all sampled stations at surface and DFM, except at surface of station 2. In the Bellingshausen Sea was only detected once, at station 15 at DFM (50% of total VP) (Table 1SM, **Figure 2**). The average percentage of prokaryotes consumed by grazers showed significantly higher averaged value in the Bransfield Strait (29.10 ± 2.87% d−<sup>1</sup> ) in respect to the Bellingshausen Sea (15.06 ± 4.91% d−<sup>1</sup> ), while the averaged values of the lysed prokaryote cells due to viruses was higher, although not significantly, in the Bransfield Strait (32.92 ± 18.50% d−<sup>1</sup> ) than the other two areas (**Table 1**, Table 1SM).

When pooling all data for the three areas, we observed that PHP was significantly related with UF (indicator of phytoplankton biomass), and with HF and viral abundances (main prokaryote mortality agents, Figure 1SM, **Table 2**). However, when considering each area separately, the tightest relationship between PHP and UF was obtained in the Weddell Sea, where UF explained a 75% of PHP variability. Furthermore, HF and PHP achieved the strongest relationships in the Bransfield Strait, where PHP explained a 69% of the HF abundance variability (**Table 2**). Finally, the strongest relationship between viral abundance and PHP was found in the Bellingshausen Sea, where PHP explained 40% of the variability of viral abundance (**Table 2**).

#### Mortality Rates of Prokaryotes at the Surface and in the DFM

Comparison of grazing rates of protists between surface and DFM, tended to be higher at the surface (**Table 1**, Table 1SM, **Figure 3**), except for stations 5 and 23 (Weddell and Bellingshausen Seas, respectively). But, we did not find significant differences between surface and DFM among areas (t-test, n = 16, p > 0.05), neither within each area. Whereas, the rate of lysed prokaryotes and percentage of lysed cells due to viruses always reached significantly higher values at the surface than in the DFM among areas (t-test, n = 16, p < 0.02), being significantly different within each area in the Weddell Sea (n = 4, p < 0.02), the Bransfield Strait (n = 6, p < 0.02) and in the Bellingshaussen Sea, when excluding station 23, (n = 4, p < 0.05) (**Table 1**, **Figure 3**).

Comparison between viral to protistan impact on prokaryotes at the surface showed that averages of the rate of lysed cells was always higher than grazing rates in all three areas, while in the DFM, averages of grazing rates was higher than rates of lysed cells except in the Bellingshausen Sea (**Table 1**, **Figure 3**). Looking closely to each single station, in surface, the viral impact was similar or higher than grazing rates, except at station 15 in the Bellingshausen Sea (**Figures 3E–F**). Significantly higher values of lysed cells were detected at stations 2, 7, 17, and 23 (**Figures 3**, t-test n = 2, p < 0.05). Conversely, the impact of grazers was

#### TABLE 1 | Depth-averaged and minimum and maximum values for the three areas, n: number of data.


VPR (virus-prokaryote ratio), HF and PF (heterotrophic and phototrophic pico/nanoflagellates), PHP (prokaryote heterotrophic production), VP<sup>L</sup> (viral lytic production), VPLyso (lysogeny), GZ (grazing rates on bacteria), RLC (rates of lysed prokaryotic cells), %PSS (percentage of prokaryotes removed by protists or by viruses). In gray averages of viral and protists processes in surface and in the DFM. nd: no detected. Number in parenthesis: cases where VPLyso was not detected. In bold averaged values.

FIGURE 2 | Viral lytic and lysogenic production in each station and depth. Full circles: lytic production at surface; empty circles: lytic production at DFM; full diamonds: lysogenic production at surface; empty diamonds: lysogenic production at DFM. Outlier value for lysogeny at surface of station 7 is indicated.

TABLE 2 | Regression equations (y = a + bx) between: prokaryotic heterotrophic production (PHP, <sup>µ</sup>g C L−<sup>1</sup> <sup>d</sup> −1 ) and Units of Fluorescence (UF); heterotrophic pico/nanoflagellate (HF, cells ml−<sup>1</sup> ) and PHP; viral abundance (VA, virus ml−<sup>1</sup> ) and PHP.


Dp, Dependent variable; Indp Var, Independent variable. WS, Weddell Sea; BrS, Bransfield Strait; BeS, Bellingshausen Sea. Log, logarithm; ns, no significant.

similar or higher than viruses in the DFM, except at stations 17 and 23 in the Bellingshausen Sea (**Figures 3E–F**). Specifically, significant higher values for protistan activity in the DFM were observed at stations 2, 5, and 13 (**Figure 3**, t-test n = 2, p < 0.05).

#### Top-down vs. Bottom-up Control

We applied the approach of Ducklow (1992), using prokaryotic biomass (PB, µg C L−<sup>1</sup> ) plotted against prokaryotic heterotrophic production (PHP, µg C L−<sup>1</sup> h −1 ) to detect whether prokaryotes were bottom-up (substrate limitation) or top-down (grazers and viruses pressure) controlled. Regarding these regressions, slopes lower than 0.4 indicate weak control by resource supply, between 0.4 and 0.6 indicate a moderate resource-limitation, and higher than 0.6 denote a stronger control by substrate. Following this convention, considering all data pooled and also the data for each of the three different areas separately, all the relationships between PB and PHP were significant and the slopes were always lower than 0.4 (all data: log PHP = 0.99 + **0.22**∗log PB, n = 54, R <sup>2</sup> = 0.19, p < 0.001; Weddell Sea: log PHP = 1.02+ **0.37**∗log PB, n = 13, R <sup>2</sup> = 0.502, p < 0.005; Bransfield Strait: log PHP = 0.99 + **0.17**∗log PB, n = 21, R <sup>2</sup> = 0.407, p < 0.01; Bellingshausen Sea: log PHP = 1.15 + **0.25**∗log PB, n = 20, R <sup>2</sup> = 0.179, p = 0.05). According to these slopes (<0.4), the prokaryotes in our area of study were top-down controlled by protists and viruses.

#### Sensitivity to Temperature

The relationships between temperature and prokaryotic heterotrophic production, grazing rates and rates of lysed cells measured in the three areas formed two groups: one that included data for the Bransfield Strait and the Bellingshausen Sea, and the other for the Weddell Sea. Both groups showed similar slopes with temperature, but different intercepts (**Figure 4**). PHP and the rate of lysed cells showed similar steeper slopes (Ea), 7.69 ± 0.85 and 9.60 ± 4.23, respectively, than grazing rates, 4.66 ± 0.67 (**Figure 4**). Consequently, PHP and RLC have higher sensitivity to temperature than grazing rates.

#### DISCUSSION

In the present study we verify our hypotheses about the impact of the predators on bacterial communities in Antarctic waters and provide new information on mortality rates of prokaryotes due to viruses and protists at surface and DFM during the Austral summer, as well as their control on prokaryotic communities, and their sensitivity to temperature.

Viral lytic production observed during this study was at the levels of the reported by Evans and Brussaard (2012) in the Weddell Sea, also during the Austral Summer, but lower than in the East sub-Antarctic zone (Evans et al., 2009). Lysogeny presented a large variation (from not detectable to 16.5 × 10<sup>7</sup> virus ml−<sup>1</sup> d −1 ) between the sampling points (**Figure 2**), as is also indicated in Evans and Brussaard (2012). Part of this variability could be attributed to the time span between sampling stations within the same region (2–10 days). In addition it is probable that we might have sampled different water masses. Noticeably, in the Weddell Sea lysogeny was very high (∼80% of total viral production, **Table 1**), and at the same time we detected there the highest values of UF, with a phytoplanktonic community dominated by diatoms, suggesting that we could be in a situation of bloom or an initiation of bloom. This is contrary to the findings of Brum et al. (2016), who sampled a coastal area of the Western Antarctic Peninsula from November to February, and showed that in summer months (January and

February) chlorophyll a concentration achieved high values and lysogeny was almost negligible. These authors hypothesized that lysogeny was linked to different phytoplankton bloom phases. Thus, during the initiation of the phytoplankton bloom lysogens would be induced in response to the improved trophic conditions (Stewart and Levin, 1984). This would cause the decline of lysogeny and an increase of cell lysis, which would increase the pool of dissolved organic matter. It would be also reasonable

to expect that the induction of lysogens would persist along the summer and in a post-bloom situation. This would favor the availability of the pool of DOM by cells lyses together with the DOC derived from senescent phytoplankton cells, increasing prokaryotic abundance, heterotrophic production and cell lysis. Furthermore, in the Bransfield Strait and in the Bellingshausen Sea we observed low UF values, suggesting a post-bloom situation. Thus, in the Bransfield Strait, we detected lysogeny in 5 of six cases, where grazing rates achieved higher values than rates of lysed cells in the DFM (**Figure 3**) and similar in surface for stations 10 and 13 (**Figure 3**), and this was reflected by the strongest relationship between PHP and HF. In contrast, in the Bellingshausen Sea, lysogeny was only detected in the DFM at station 15, and in the other two stations we obtained high rates of lysed cells at the surface and DFM, high prokaryotic and viral abundances as well as a significant relationship between PHP and viral abundances (**Table 2**).

In the studied Antarctic waters, mortality rates of prokaryotes due to viral activity achieved the maximum values, and significantly higher than those by protistan grazing in several surface samples (**Figure 3**). These results are in agreement to those in Guixa-Boixereu et al. (2002) where viral activity was higher than grazing by protists in the Bransfield Strait, the Gerlache Strait and the Bellingshaussen Sea. In contrast, a similar or higher impact of grazers than viruses was found in the DFM. A possible explanation is that HF and mainly PF abundances were similar or higher in the DFM than in the surface (Table 1SM). It might be likely that part of the PF considered in this study as phototrophic microorganisms, might be mixotrophs, which is a common feature in Antarctic waters, representing 10% of all chloroplastidic nanoflagellates in the water column (Moorthi et al., 2009).

As a result of viral and protist activities, during the Austral summer prokaryotes were top-down controlled according to the approach of Ducklow (1992). This is corroborated by the fact that both predators removed on average a high percentage of prokaryotic biomass (46.60 ± 9.24% d−<sup>1</sup> ) and production (116.09 ± 13.01% d−<sup>1</sup> ) and these losses would not be sustainable if there was not enough DOC to maintain the growth of the population. Brum et al. (2016) and Evans and Brussaard (2012) stated that the release of DOM from prokaryote lysates in the onset of blooms could also contribute to maintain the sufficient levels of substrate for prokaryotic growth. Also Morán and Estrada (2002) in the Bransfield Strait measured the supply of photosynthate by phytoplankton, showing that the DOC released by primary producers overcome the prokaryotic requirements, while Guixa-Boixereu et al. (2002) in the same cruise reported high mortality rates on prokaryotes, corroborating the top-down control. In contrast, in Franklin Bay in the Arctic Ocean prokaryotic community in spring was bottom-up regulated, suggesting that prokaryotic growth was controlled only by substrate and escaping predation (Vaqué et al., 2008).

In order to summarize in a big picture the main fluxes of carbon in the studied Antarctic areas (notice that respiration was not measured in this study), we performed a rough estimation of the carbon biomass and fluxes (PHP, rate of lysed cells and grazing) averaging all values from each area (**Table 3**). According


TABLE 3 | Biomass of prokaryotes (PB), viruses (VB), heterotrophic pico/nanoflagellates (HFB ≤5 µm), phototrophic pico/nanoflagellates (PFB), prokaryotic heterotrophic production (PHP ± SD), mortality rates due to grazers (GZ ± SD), and to viral lyses (RLC ± SD) at the surface and deep fluorescence maximum (DFM).

Biomasses are expressed in µg C L−<sup>1</sup> and rates in µg C L−<sup>1</sup> d −1 . WS, Weddell Sea; BrS, Bransfield Strait; BeS, Bellingshausen Sea. St, station; Z, depth (surface and DFM). In bold averaged values ± SE.

to that, the viral shunt of prokaryotic carbon to the water column provided as much as 1.4 ± 0.5 µg C L−<sup>1</sup> d −1 in the Weddell Sea, 1.7 ± 1.0 µg C L−<sup>1</sup> d −1 in the Bransfield Strait, and 2.0 ± 0.6 µg C L−<sup>1</sup> d −1 in the Bellingshausen Sea. At the same time, the carbon ingested by small bacterivores, HF ≤ 5 µm, was similar or lower to that released due to viral lysis (0.8 ± 0.2 µg C L−<sup>1</sup> d −1 in the Weddell Sea, 1.4 ± 0.1 µg C L−<sup>1</sup> d −1 in the Bransfield Strait and 0.9 ± 0.2 in the Bellingshausen Sea), while PHP values ranged from 3.4 ± 0.9 µg C L−<sup>1</sup> d −1 in the Weddell Sea to 1.9 ± 0.2 µg C L−<sup>1</sup> d −1 in the Bellingshausen Sea. In terms of carbon biomass, average prokaryotic carbon was considerably higher (10.6 ± 2.2 µg C L−<sup>1</sup> ) and the carbon biomass of bacterivorous HFB ≤ 5 µm was the lowest (2.7 ± 0.5 µg C L−<sup>1</sup> ) in the Bellingshausen Sea (**Table 3**). This global estimation suggests a significant contribution of viral activity in the Austral Summer and agrees with the results obtained by Christaki et al. (2014) in the Kergelen Area (Sub-Antarctic ocean). These authors found that the absolute amount of bacterial carbon channeled toward the higher trophic levels was the same in spring and summer, but the underlying mechanisms were contrasting. In spring, the PHP was relatively low and most of it was channeled to the higher trophic levels due to predation, while in summer, the PHP was higher, but most of it was returned to the dissolved phase through viral lysis. It is plausible that higher lysis in summer might denote the result of the induction of lysogens, agreeing with the findings of Brum et al. (2016) that viral activity was more important during the Austral Summer, when the phytoplankton bloom prevails.

Temperature played an important role in the studied microbial processes (PHP, GZ and RLC) and PHP and RLC responded quicker to small changes than grazing rates (**Figure 4**). In contrast, Maranger et al. (2015) showed that PHP and GZ were more sensitive to temperature than viral activity in Arctic regions. This could be due to the fact that in the Arctic, during the summer season, the range of temperature is wider (−1.87 to 9.0, Maranger et al., 2015) than in Antarctic waters (−1.87 to 4.0, Danovaro et al., 2011). Thus, Arctic HFs were probably well adapted to this wide range, and responded quickly to small increases of temperature. In contrast, HFs in Antarctic waters are used to live in a narrower range of cold temperatures and do not respond as quickly when the temperature increases (Vaqué et al., 2009). Furthermore, Sarmento et al. (2010) in a study considering data of PHP and GZ from temperate to Antarctic systems also found that PHP (Ea, slope = 0.67) was more sensitive to temperature than GZ (Ea, slope = 0.47), although they did not have viral activities to compare with. Indeed, our results suggest a decrease of the fraction of PHP taken by protists and an increase of that recirculating through viral lysis in a warmer ocean. Such a change in the carbon fluxes would imply lower microbial carbon efficiency in the channeling of organic matter to higher trophic levels, and increase of respiration. Together with the few studies on the different sensitivity of these two mortality processes, our work may help to predict the destiny of the microbial carbon in a warmer ocean. However, it may be necessary to carry out more studies on the effect of temperature on microbial processes in order to confirm our empirical observations.

### CONCLUSIONS

Our results indicate that grazers and viruses controlled the prokaryotic community in the visited Antarctic areas during the Austral summer 2009. Viruses had the highest impact on prokaryotes mainly at the surface, while protists activity played an important role in the DFM. In the big picture, the contribution of the prokaryotic carbon fuelled to the water column due to viral activity was higher than the prokaryotic carbon transferred to higher trophic levels by grazers. Prokaryotic heterotrophic production and viral activity were more sensitive to temperature than protists grazing, suggesting a lower flux of C through the microbial loop. This fact should be confirmed in further field and experimental studies in Antarctic waters to understand future changes in microbial dynamics in a warmer ocean scenario.

#### AUTHOR CONTRIBUTIONS

DV and JB designed the study, collected samples for viruses, prokaryotes and protists, performed mortality experiments, analyzed the results and wrote the manuscript; SA coordinated the phytoplankton sampling, its analysis and identification, and provided constructive criticisms to the manuscript; FT, YC, and EL processed the abundances of viral and protists profiles, as well as viral and prokaryotic abundances in experimental samples, and also helped with data analysis; JA performed heterotrophic prokaryotic production and prokaryotic abundances in natural samples, adding clever comments to the text; CD coordinator

#### REFERENCES


of the ATOS project, provided a creative environment and added constructive criticisms throughout the study; MS provided constructive comments, revised and edited the manuscript. All authors commented and discussed the obtained results, and suggested improvements on the manuscript

#### FUNDING

This study was supported by the following projects: ATOS (POL2006-00550/CTM, P.I.: CD.) funded by the Spanish Ministerio de Ciencia e Innovación, YC work was supported by a Ph.D. fellowship from the MINECO (FPI grant).

#### ACKNOWLEDGMENTS

We are grateful to Sebastian Lasternas that actively contribute to phytoplankton identification and counts, and to the staff of UTM (CSIC), colleagues of the ICM and IMEDEA and the crew of the R/V Bio-Hespérides for their help during the cruise. We also thank to three anonymous reviewers for their helpful comments that have contributed to improve the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.00241/full#supplementary-material


during Austral summer 1995/96. Deep-Sea Res. Part II 49, 603–621. doi: 10.1016/S0967-0645(01)00114-X


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Vaqué, Boras, Torrent-Llagostera, Agustí, Arrieta, Lara, Castillo, Duarte and Sala. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Microbial Community Response to Terrestrially Derived Dissolved Organic Matter in the Coastal Arctic

Rachel E. Sipler<sup>1</sup> \*, Colleen T. E. Kellogg<sup>2</sup> , Tara L. Connelly<sup>3</sup>† , Quinn N. Roberts<sup>1</sup> , Patricia L. Yager<sup>3</sup> and Deborah A. Bronk<sup>1</sup>

<sup>1</sup> The Virginia Institute of Marine Science, College of William & Mary, Gloucester Point, VA, United States, <sup>2</sup> Department of Microbiology & Immunology, University of British Columbia, Vancouver, BC, Canada, <sup>3</sup> Department of Marine Sciences, University of Georgia, Athens, GA, United States

#### Edited by:

Ingrid Obernosterer, Observatoire Océanologique de Banyuls sur Mer (OOB), France

#### Reviewed by:

Daniel P. R. Herlemann, Leibniz Institute for Baltic Sea Research (LG), Germany Andrea Niemi, Fisheries and Oceans Canada, Canada

#### \*Correspondence:

Rachel E. Sipler rachelsipler@gmail.com

#### †Present address:

Tara L. Connelly, Department of Ocean Sciences, Memorial University of Newfoundland, St. John's, NL, Canada

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Microbiology

Received: 18 January 2017 Accepted: 22 May 2017 Published: 09 June 2017

#### Citation:

Sipler RE, Kellogg CTE, Connelly TL, Roberts QN, Yager PL and Bronk DA (2017) Microbial Community Response to Terrestrially Derived Dissolved Organic Matter in the Coastal Arctic. Front. Microbiol. 8:1018. doi: 10.3389/fmicb.2017.01018 Warming at nearly twice the global rate, higher than average air temperatures are the new 'normal' for Arctic ecosystems. This rise in temperature has triggered hydrological and geochemical changes that increasingly release carbon-rich water into the coastal ocean via increased riverine discharge, coastal erosion, and the thawing of the semipermanent permafrost ubiquitous in the region. To determine the biogeochemical impacts of terrestrially derived dissolved organic matter (tDOM) on marine ecosystems we compared the nutrient stocks and bacterial communities present under ice-covered and ice-free conditions, assessed the lability of Arctic tDOM to coastal microbial communities from the Chukchi Sea, and identified bacterial taxa that respond to rapid increases in tDOM. Once thought to be predominantly refractory, we found that ∼7% of dissolved organic carbon and ∼38% of dissolved organic nitrogen from tDOM was bioavailable to receiving marine microbial communities on short 4 – 6 day time scales. The addition of tDOM shifted bacterial community structure toward more copiotrophic taxa and away from more oligotrophic taxa. Although no single order was found to respond universally (positively or negatively) to the tDOM addition, this study identified 20 indicator species as possible sentinels for increased tDOM. These data suggest the true ecological impact of tDOM will be widespread across many bacterial taxa and that shifts in coastal microbial community composition should be anticipated.

Keywords: dissolved organic matter, bacterial diversity, Arctic, Chukchi Sea, microbial community composition, tDOM

#### INTRODUCTION

As much as 50% of the world's terrestrial organic carbon (C) pool is stored in the northern hemisphere as permafrost and approximately half of that is located within the upper 3 m (Gorham, 1991; Tarnocai et al., 2003, 2009). As atmospheric temperatures rise, the once semi-impermeable layers of permafrost begin to thaw, mobilizing a portion of this C pool by increasing the active layer depth, and then increasing riverine discharge and coastal erosion (Froese et al., 2008; Schuur et al., 2008; Feng et al., 2013; Holmes et al., 2013). This change in hydrology releases water, C, and nutrients, once trapped within the ice, into the coastal ocean. The amount of terrestrially derived dissolved organic matter (tDOM) supplied to the coastal ocean is hard to constrain due to complex environmental factors including changes in climate, hydrology and vegetation (Amon et al., 2012;

**240**

Kicklighter et al., 2013). However, it is estimated that mobilization of organic C has increased by 3 – 6% from 1985 to 2004 (Feng et al., 2013) and will continue to increase with a warming climate (Amon et al., 2012). Bacterial production in the coastal Arctic appears to be controlled by the bottom up supply of C (Garneau et al., 2008) and the composition of C and nitrogen (N) pools can shape microbial community structure. These microbial responses can impact higher trophic levels, biogeochemical fluxes, and climate feedbacks (McCarren et al., 2010; Vincent, 2010). We hypothesize that tDOM increases will be important to the overall productivity of the Arctic Ocean and suggest that responsive microorganisms could be important sentinels of environmental change. Therefore, understanding what fraction of tDOM is bioavailable, how coastal microbial communities will respond to increases in tDOM and what impact these changes will have on microbial community composition and nutrient cycling is paramount.

Most of the freshwater runoff (both by water volume and DOM contribution) into the coastal Arctic Ocean happens during the short-window of freshet. The term freshet describes the pulse of freshwater that occurs in spring upon the melting of winter snow and ice accumulations. On the Alaskan Arctic coast, >50% of annual freshwater river discharge occurs within a 2 week period in spring (McClelland et al., 2014). This ephemeral fluvial discharge is accompanied by tremendous loads of tDOM and inorganic nutrients (McClelland et al., 2014). As a result, Arctic rivers are highly enriched in dissolved organic C (DOC) and N (DON) (Opsahl et al., 1999; Mathis et al., 2005; Holmes et al., 2008; Frey and Mcclelland, 2009), exposing coastal microbial communities to concentrated pulses of tDOM annually. The tDOM supplied by these Arctic rivers was once thought to be highly refractory (Opsahl et al., 1999; Dittmar and Kattner, 2003; Amon and Meon, 2004; Tank et al., 2012; Xie et al., 2012). Several studies, however, have found that a portion (10 – 40%) of this riverine tDOM is bioavailable on time scales of weeks to months (Hansell et al., 2004; Holmes et al., 2008; Alling et al., 2010; Mann et al., 2012; Vonk et al., 2013). This suggests that much of the bioavailable tDOM is being degraded within estuaries and the coastal ocean before it reaches off-shore waters.

Compared to more temperate regions of the world, study sites in the coastal Arctic are rare. Much of the existing research has focused on the major Arctic rivers, and comparatively little attention has been paid to the impacts of smaller rivers. This lack of knowledge impedes our ability to predict the impact of climate change on local Arctic food webs and biogeochemical cycles (Thingstad et al., 2008). In an effort to fill this knowledge gap, this study used tDOM from the Meade River, Alaska to address two main research objectives: (1) assess the bioavailability of tDOM to microbial communities in the coastal Chukchi Sea and (2) determine how pulsed, high concentration additions of tDOM, impact microbial community composition. These objectives were achieved through a series of bioassay experiments where extracted Meade River tDOM was supplied to microorganisms collected from coastal waters of the Chukchi Sea during ice-covered (April) and open-water (August) conditions. Biological [chlorophyll a (Chl a) and bacterial community composition] and chemical [ammonium (NH<sup>4</sup> <sup>+</sup>), nitrate (NO<sup>3</sup> <sup>−</sup>), DON, DOC, and phosphate (PO<sup>4</sup> <sup>3</sup>−)] changes were monitored throughout the experiments. This study provides the first evidence that some tDOM-responsive bacterial taxa could provide early indications of environmental changes in the coastal Arctic.

# MATERIALS AND METHODS

#### tDOM Collection and Extraction

Approximately 100 L of water was collected from a thermokarst draining into the Meade River, Alaska, at its mouth (70◦ 540 39.100N, 156◦ 070 25.900W), on 29 August 2010. The pH was 6.5, the salinity was 4 and the temperature was 7◦C. The salinity of the samples is indicative of its proximity to the coast. The thermokarst water was collected into acid washed (10% HCl) high-density polyethylene (HDPE) carboys and transported in freeze safes (∼2 h transit) to the Barrow Arctic Research Center (BARC). The sample was filtered sequentially through pre-rinsed 5-µm polycarbonate filters, combusted (450◦C for 4 h) Whatman Glass fiber filters (GF/F; nominal pore size 0.7 µm), and 0.2-µm Supor <sup>R</sup> (PALL corp) filters. A 10% HCl solution was added to the filtered water to reduce its pH to 2. Samples were stored at 4 ◦C until being shipped in freeze safes to the Virginia Institute of Marine Science (VIMS). Data loggers (HOBO TidbiT v2) were used to monitor water temperature at 60-s intervals during both the transit to BARC and VIMS. Sample temperatures did not increase more than 0.7◦C during transit.

Two different extraction methods were compared to identify the method with the highest DOC recovery prior to the large-scale extraction efforts. Sub-samples of the source water were extracted using Superlite DAX-8 resin, described by Aiken (1985) for Amberlite XAD-8 resin or PPL solid phase extraction (Dittmar et al., 2008). Both extraction methods isolate and concentrate the DOM source retaining a fraction of the DOM pool but allowing salts (including inorganic nutrients) to pass through as filtrate. The DOC recovery using the DAX-8 resin was higher than the PPL extractions (61 and 48%, respectively) for the same aqueous sample. Because of its higher recovery, Superlite DAX-8 resin (Aiken, 1985) was used to extract the DOM used in this study. Once isolated, the salinity of the DAX extracted tDOM was adjusted to 30 using a combination of baked (500◦C for 4 h) sodium chloride, magnesium sulfate and sodium bicarbonate. The change in salinity causes a portion of the tDOM to precipitate out of solution, mimicking the salinity induced flocculation that occurs within estuaries and deltas. After the salinity was adjusted, the higher salinity tDOM solution was re-filtered (0.2 µm) to remove any particulates that formed through this flocculation process, thus providing a more realistic geochemical signature of the tDOM. The proportion of total riverine DOM that is removed via salinity induced flocculation ranges between 3 and 11% (Sholkovitz, 1976). We found that ∼17% of the DOC was removed during the flocculation of the extracted tDOM (humic) fraction. Prior to use in the bioassay experiments, the tDOM source was brought to the in situ temperature of the experimental waters (−1 ◦C or 4.5◦C) to ensure additions of tDOM did not shock the receiving communities.

# Field Sample Collection and Bioassays

Seawater samples were collected from a site located 2.5 km northwest of Barrow, Alaska (71◦ 200 39.600N, 156◦ 410 25.000W). Whole (unfiltered) Chukchi Sea water was collected from 4 m depth in April and from a depth of 8 m in August. In April, the site was covered by 1.5 m of ice. The site was ice-free in August. Samples were collected in acid washed (10% HCl), 20 L HDPE carboys. The sampling procedures are described in detail in Baer et al. (2017). Bioassays were performed in April (26 April – 1 May 2011) and August (17 – 21 August 2011) and incubated in walk-in chambers located at the BARC in Barrow, Alaska. Temperatures were maintained at −1 ◦C in April and 4.5 ◦C in August, corresponding to ambient water temperatures at the time of collection. Both of these experiments were conducted under ambient light conditions, which was greater in August (∼50 µmol quanta m−<sup>2</sup> s −1 ) than in April (4.7 µmol quanta−<sup>1</sup> m−<sup>2</sup> s −1 ), so both autotrophic and heterotrophic processes were apparent.

Pulsed DOM addition studies (Bioassays) are particularly applicable for the coastal Arctic. This is based on the well documented large pulses of high tDOM waters to the coast via Arctic rivers (Amon et al., 2012; Holmes et al., 2012, 2013; McClelland et al., 2014). While freshet represents the largest pulse of DOC to the coastal Arctic annually, the flow of DOM via rivers continues throughout the summer and can include pulses of DOM release (e.g., Holmes et al., 2008, 2012). Although DOC data in coastal waters during pulse events are limited, April DOC concentrations in the coastal Beaufort Sea range from ∼100 to 400 µmol C L−<sup>1</sup> and August DOC concentrations range between ∼75 and 300 µmol C L−<sup>1</sup> (Dunton and Crump, 2014). Our additions were 400 µmol C L−<sup>1</sup> tDOM in both seasons. The same concentration and volume was used in April and August to reduce experimental variability and thus microbial response among the study seasons. Although 400 µmol C L−<sup>1</sup> tDOM additions may be a bit high for summer incubations, with increasing atmospheric temperatures and the known C reserves contained within permafrost (Gorham, 1991; Tarnocai et al., 2003, 2009), the addition is realistic.

Bioassays were prepared by adding 1.25 L of substrate (extracted tDOM or artificial seawater in the Control treatments) to 8.75 L of whole seawater in acid washed (10% HCl), 10-L polycarbonate carboys, resulting in a 12.5% dilution of the ambient community. The addition of artificial seawater to the control ensured that the dilutions of the microbial community and ambient nutrient pools were consistent among both the control and tDOM addition treatments. This replication of dilution allowed us to specifically target the microbial response to the extracted tDOM. All bioassays were run in duplicate. Bioassays were sub-sampled for Chl a, bacterial community composition, and nutrient concentrations [DOC, total dissolved N (TDN; used to calculate DON), NO<sup>3</sup> <sup>−</sup>, nitrite (NO<sup>2</sup> −), NH<sup>4</sup> <sup>+</sup>, and PO<sup>4</sup> <sup>3</sup>−], at time points 0, 1, 2, 4, and 6 days in April and 0, 1.25, 2.5, and 4 days in August. The duration of the April incubation was longer (6 days) than the August incubation (4 days) due to a shorter field opportunity created by extensive weather delays in summer. Nutrient samples were filtered through combusted (450◦C for 2 h) Whatman GF/F

filters (nominal pore size 0.7 µm) and stored frozen until analyzed. Bacterial abundance was measured for ambient site samples in both April and August but bacterial abundance and production measurements within the bioassays were only made during the April sampling. The August bioassay bacterial abundance samples were collected but degraded prior to analysis and complications with shipping the radioisotopes to Barrow precluded production measurements in August.

#### Nutrients and Dissolved Organic Matter Assessment

The colorimetric phenol-hypochlorite method was used to manually measure NH<sup>4</sup> <sup>+</sup> concentrations in triplicate (Koroleff, 1983). Concentrations of NO<sup>3</sup> <sup>−</sup>, NO<sup>2</sup> <sup>−</sup>, and PO<sup>4</sup> <sup>3</sup><sup>−</sup> were measured using a Lachat QuikChem 8500 autoanalyzer (Parsons et al., 1984). Triplicate TDN and DOC concentrations were measured by high temperature combustion on a Shimadzu TOC-V TNM autoanalyzer (Hansell, 1993; Sharp et al., 2002, 2004). Deep-sea and low-carbon reference water from the University of Miami consensus reference material program were used as a quality control standard for DOC and TDN samples (Hansell, 1993). Concentrations of DON were determined as the difference between the TDN and dissolved inorganic N substrates; the errors from the TDN, NH<sup>4</sup> <sup>+</sup>, and NO<sup>3</sup> <sup>−</sup> + NO<sup>2</sup> − measurements were propagated to provide a standard error for DON.

Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS) was used to assess the composition of the extracted tDOM. The extracted tDOM was analyzed in positive ionization mode using a hybrid 7 Tesla linear ion trap (LTQ) FT-ICR mass spectrometer equipped with an electrospray ionization (ESI) inlet system (LTQ FT Ultra, Thermo Electron Corp., Wood Hole Oceanographic Institute Mass Spectrometer Facility). Samples were analyzed in the positive ionization mode according to methods described in Sipler et al. (2013) and data were processed and internally calibrated according to Bhatia et al. (2010). FT-ICR MS data are reported as mass-to-charge ratios (m/zs), because the data presented here represent singly charged compounds; an individual m/z generally represents a single compound. The mass range analyzed was 100 – 1000 m/z. Compound classifications were assigned based on general hydrogen to C (H:C) and oxygen to C (O:C) molar ratios. General classifications and their corresponding ranges are described in Hockaday et al. (2009) and listed in **Table 1**. Multiple studies have used H:C and O:C molar ratios to relate the thousands of compounds generated using FT-ICR to more common compound classifications (e.g., Kim et al., 2003; Hockaday et al., 2009; Bhatia et al., 2010; Ohno et al., 2010; Sleighter et al., 2010; Sleighter and Hatcher, 2011). These classifications provide a general understanding of composition; however, the specific ranges and nomenclature used to distinguish among these classifications continues to evolve (e.g., Kim et al., 2003; Hockaday et al., 2009; Bhatia et al., 2010; Ohno et al., 2010; Sleighter and Hatcher, 2011). For example, the terms lignin, tannin and terrestrial have all been used to describe overlapping ranges (e.g., Hockaday et al., 2009; Bhatia et al., 2010; Killberg-Thoreson et al., 2013). Here we follow the

TABLE 1 | Elemental ratios and compound classification characteristics of masses detected in extracted tDOM source using FT-ICR MS.


Molecular formulas were assigned to 96% (5343 m/z) of the 5543 m/zs detected. Data represent the classification, range of H:C and O:C molar ratios, the number of m/zs per classification, and proportion of total m/zs detected in positive ionization mode using Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS). Classifications described in more detail in Hockaday et al. (2009).

ranges and nomenclature outlined by Hockaday et al. (2009). It should also be noted that although we have provided H:C and O:C ratios for each classification, a high degree of overlap among classifications is expected (Hockaday et al., 2009). We use these general classifications as a tool to group and visualize the potential composition and thus origins of the detected compounds. The aromaticity index (AI) was also calculated for all m/zs with molecular formula assignments (Koch and Dittmar, 2006; Šantl-Temkiv et al., 2013).

#### Biological Assessment

Concentrations of Chl a were measured fluorometrically (Parsons et al., 1984). Bacterial abundance was measured from triplicate samples taken from each incubation bottle and fixed with 4% formaldehyde. Abundance samples were stained with SYBR Green and counted on a flow cytometer (FACSCalibur, Becton-Dickinson). Fluorescing beads were used to approximate the number of cells mL−<sup>1</sup> and at least 10,000 cells were counted per replicate. Bacterial production rates were measured via the incorporation of <sup>3</sup>H-leucine into protein as a proxy for growth (Kirchman et al., 1985; Smith and Azam, 1992; Ducklow, 2000; Kirchman, 2001). Briefly, triplicate aliquots of 1.5 mL seawater from each treatment bottle were incubated in the dark in a water bath set to temperatures mimicking ambient temperatures for 4 h with 25 nmol L−<sup>1</sup> of <sup>3</sup>H-leucine (Williams et al., 2016). Incubations were stopped by rinsing each aliquot with 0.1 mL of 100% trichloroacetic (TCA). Protein was extracted by rinsing each aliquot with 1 mL of ice-cold 50% TCA and then by rinsing with 1 mL of 80% ethanol. Samples were centrifuged between each rinse. Parallel incubations for killed controls were done by adding <sup>3</sup>H-leucine only after killing cells with 0.1 mL of 100% TCA. Activity measured in killed controls was subtracted from the sample values, eliminating any isotope adsorption to particulate protein.

#### Bacterial Community Composition

Whole water from the sample sites and each incubation bottle was filtered onto a 0.2-µm Supor filter and frozen at −80 ◦C. Before extraction, 900 µL of DNA extraction buffer (Fortunato and Crump, 2011), lysozyme (2 mg/mL final concentration) and proteinase K (0.2 mg/mL final concentration) were added to the filters. Samples were then subjected to three freeze-thaw cycles. This was followed by enzymatic lysis in a 30 min 37◦C incubation and then continued lysis at 65◦C for 1 – 2 h after the addition of SDS (1% final concentration). Two phenol:chloroform:isoamyl alcohol (25:24:1) extractions were then carried out to isolate nucleic acids. Nucleic acids were then precipitated using 100% isopropanol (0.6 × volume of the resulting supernatant) for 2 h up to overnight, pelleted at 13,000 rpm for 30 min, and then rinsed and re-pelleted twice with 70% ethanol before being dried in a roto-evaporator. Once dry, samples were resuspended in 250 mL of nuclease-free water.

The V4 region (515F, GTGCCAGCMGCCGCGGTAA and 806R, GGACTACHVGGGTWTCTAAT) of the 16S rRNA gene for prokaryotic composition was amplified using Earth Microbiome Project protocols<sup>1</sup> , but with only 30 total cycles. This 806R primer has a bias against SAR11 sequences (Parada et al., 2015) and thus the relative abundance of these taxa and Alphaproteobacteria in general are likely underestimated in our study. Sample libraries were sent to Argonne National Lab for 2 × 150 bp sequencing on the Illumina MiSeq platform and reads were paired using fastq-join (Aronesty, 2013). Sequences that successfully joined were quality filtered, dereplicated (derep\_fulllength) and abundance sorted (sortbysize) using UPARSE v 7.0.1001 (fastq\_filter) (Edgar, 2013) with an expected error rate of 0.5. Singleton sequences were removed in the latter step in order to prevent them from seeding clusters when clustering operational taxonomic units (OTU). Reads were then clustered (cluster\_otus in UPARSE pipeline) at 97% similarity. Reference-based chimera filtering was performed using UPARSE (uchime) with the Gold Database<sup>2</sup> as the reference database. Reads (including singletons) were subsequently mapped back to OTUs using UPARSE (usearch\_global) and an OTU table created. Taxonomy of the representative sequences was assigned in QIIME v 1.8 (assign\_taxonomy.py; Caporaso et al., 2010) using the RDP classifier trained to the Greengenes database (v. 13.8<sup>3</sup> ) for 16S amplicons. Any remaining singletons and OTUs occurring in only one sample, chloroplast, mitochondrial and archaeal sequences were removed in QIIME (filter\_otus\_from\_otu\_table.py). Sequences passing these quality control filters were uploaded to the NCBI Short Read Archive (SAMN06175504 to SAMN06175533) under the BioProject PRJNA310254. Samples were subsampled to 3037 sequences per sample (the minimum number of sequences per sample in the dataset) before subsequent analyses.

#### Statistical Analyses

We used ANOVAs (Nelson et al., 2013) and Tukey's HSB test, which controls for multiple comparisons, to determine if an OTU significantly changed (p < 0.05) over the course of the bioassay (among any of the treatments) and the treatments and time points between which these OTUs changed. These analyses

<sup>1</sup>http://www.earthmicrobiome.org/emp-standard-protocols/

<sup>2</sup>http://www.genomesonline.org/

<sup>3</sup>http://greengenes.secondgenome.com/

were only run on OTUs that were present in four or more samples and were abundant or became abundant during at least one of the time points. Here we define an abundant OTU as an OTU that reached > 1% of the sample at some point over the course of the bioassay. Replicates were collected from each sample bottle and were grouped by treatment and time point for these statistical analyses. We then calculated the fold-change between the relative abundance of these OTUs at 4 days and in the starting community to determine the magnitude of change in the relative abundance of these significant OTUs over the course of the experiments. Relationships between the average relative abundance of the OTUs that changed throughout the bioassays and average nutrient, Chl a and organic matter concentrations over the course of the bioassays were assessed using Spearman rank correlations. Non-metric multidimensional scaling (MDS) plots were computed in R, using metaMDS in the VEGAN package (Oksanen et al., 2016) and the stress for this ordination (0.0362) was calculated using nmds.monte in the BIOSTATS collection of R functions (Kevin McGarigal<sup>4</sup> ).

#### RESULTS

#### Site Characteristics

The thermokarst water sample contained no detectable concentrations of NH<sup>4</sup> <sup>+</sup> or NO<sup>3</sup> <sup>−</sup>, but was highly enriched (relative to seawater) in DOC and DON (**Table 2**). Sixty-one percent of the thermokarst DOC and 56% of the DON pools were retained during tDOM isolation. The molar DOC:DON ratio within the thermokarst water and extracted tDOM were 24 and 26, respectively. Molecular formulas were assigned to 96% (5343 m/z) of the 5543 m/zs detected in the extracted tDOM sample by FT-ICR MS (**Figure 1**). Masses were detected across the entire analytical range (100–1000 Da.) but the average molecular weight of compounds detected in this sample was 590 Da. The average H:C of detected m/zs was 1.39 and the average O:C of detected m/zs was 0.33 which is similar to other lignin and tannin-rich natural DOM samples from temperate swamps (e.g., Sleighter and Hatcher, 2008; Hockaday et al., 2009; Ohno et al., 2010) and other Arctic permafrost thaw streams (Spencer et al., 2015). Based on the H:C and O:C molar ratios of the individual m/zs, the tDOM was dominated by lignin-like compounds which accounted for 57% of the m/zs for which molecular formulas could be assigned (**Table 1**). Of the remaining masses, protein, lipid and unsaturated hydrocarbonlike compounds represented 13, 9, and 5% of the compound pool, respectively. Approximately 13% of compounds for which molecular formulas were assigned could not be categorized under the general H:C and O:C molar ratios for known classifications (**Table 1**). For comparison, we also calculated the AI of the compounds for which molecular formulas could be assigned (Koch and Dittmar, 2006; Šantl-Temkiv et al., 2013). Using this approach we found that approximately 37% (1960 m/zs) of compounds could be categorized as moderately unsaturated, and aromatic and condensed aromatic each accounted for 16 and 15%, respectively. Approximately 64% of compounds that were identified as lignin-like were also identified as moderately unsaturated showing similarities between these two distinctions.

Distinct chemical and biological differences in seawater were observed between April and August. Ambient Chl a and bacterial abundance concentrations were 3- to 4-fold higher in August than April (**Table 2**). Dissolved inorganic nutrient concentrations were higher in April than August, while DOM concentrations were lower (**Table 2**). The largest difference in ambient nutrient stocks occurred within the NO<sup>3</sup> <sup>−</sup> pool, which was 26-fold higher in April than in August.

### Seasonal Differences in In Situ Bacterial Community Composition

Distinct seasonal differences were observed across multiple taxonomic levels. The ambient communities used in the April and August bioassays shared fewer than 50% of their OTUs. The most abundant taxa classes were Gammaproteobacteria, Betaproteobacteria, Alphaproteobacteria, and Deltaproteobacteria within the phylum Proteobacteria, as well as phyla Bacteroidetes, Verrucomicrobia, and Actinobacteria (**Figure 2**). Gammaproteobacteria was the dominant group in both April and August, accounting for approximately 42% and 39% of the populations, respectively. OTUs related to family Oceanospirillaceae were the largest contributors to this class and constituted approximately 22 and 8% of the April and August communities, respectively. Sequences closely related to oligotrophic marine taxa (e.g., SAR86, SAR92 and OM182; Cho and Giovannoni, 2004) constituted only 4.5% of the April population but 13.5% of the starting populations in August.



The thermokarst water sample was collected on 29 August 2010, and marine site water samples were collected on 26 April 2011 and 17 August 2011. A full survey of this system can be found in Baer et al. (2017). Numbers reported are the average of triplicate samples ± the standard deviation. BD = below detection or < 0.03 µmol N L−<sup>1</sup> . ND = not determined.

<sup>4</sup>http://www.umass.edu/landeco/teaching/ecodata/labs/biostats.pdf

Betaproteobacteria were more prevalent in April (7%) than in August (∼3%), with dominant OTUs identified within the families Comamonadaceae and Burkholderiaceae. Although the total contribution of Alphaproteobacteria was similar between the two seasons (15.7 and 16.4% in April and August, respectively), contributions from unclassified Alphaproteobacteria OTUs were relatively more abundant in April and order Rhodobacterales OTUs had higher relative abundances in August. Bacteroidetes, dominated by Polaribacter spp., comprised only 1% of the April community compared to ∼13.7% in August. Verrucomicrobia (especially Coraliomargarita sp.) was also less abundant in April (<2%) than in August (∼10%).

Less abundant members of the starting bacterial communities, including the class Deltaproteobacteria (SAR324), and the phyla Actinobacteria (Acidimicrobiales sp., SVA0996 clade), and Marine Group A (SAR406), also exhibited distinct seasonal patterns. The proportion of SAR324, was 20-fold greater in April (4.3%) than in August (0.18%). Likewise, proportions of Acidimicrobiales were greater in April (2%) than in August (0.5%). Sequences closely related to SAR406 made up 2.7% of the April population but constituted a negligible proportion of the August community. In August, however, ∼2.5% of the Actinobacteria sequences were closely related to candidate genus "Aquiluna," a genus that was rare in April.

#### Biological and Chemical Changes in Bioassays

Additions of tDOM affected both ice-covered (April) and openwater (August) microbial communities. Bacterial abundance at the end of the April experiment was 3-fold higher when tDOM was added compared to the Control (**Figure 3A**). In April, bacterial production increased 2-fold in the Control and 55-fold in the tDOM treatment over the 6-day incubation (**Figure 3B**). Chl a concentration doubled in both the Control and tDOM treatments in April (**Figure 4**). In August, however, increases in Chl a concentration were nearly 2-fold higher (1.6 ± 0.0 µg Chl a L −1 ) in the tDOM treatment than in the no addition Control (0.9 ± 0.1 µg Chl a L −1 ; **Figure 4**).

Little change in NH<sup>4</sup> <sup>+</sup> concentration was observed in both the Control and tDOM treatments in April, but NH<sup>4</sup> + concentrations decreased by approximately 0.5 µmol N L−<sup>1</sup> in both the Control and tDOM treatments in August (**Figure 5**). Concentrations of NO<sup>3</sup> <sup>−</sup> did not change in either treatment or season, even though NO<sup>3</sup> <sup>−</sup> concentrations were an order of magnitude higher in April than August (**Figure 5**). DON concentrations in the tDOM treatments decreased by 7 ± 1 µmol N L−<sup>1</sup> in April and by 8 ± 1 µmol N L−<sup>1</sup> in August (**Figure 5**), compared to no change in the Control treatments. DOC concentrations in the tDOM treatments decreased by 29 ± 2 µM and 37 ± 3 µmol C L−<sup>1</sup> in April and August, respectively (**Figure 5**), compared to no change in the Control treatments (**Figure 5**). Concentrations of PO<sup>4</sup> <sup>3</sup><sup>−</sup> decreased by 0.15 ± 0.04 µmol P L−<sup>1</sup> in April and 0.3 ± 0.05 µmol P L−<sup>1</sup> in August in the tDOM treatments, compared to no change in the Controls.

# Bacterial Community Response to tDOM Amendments

A clear shift in community composition was observed in response to tDOM amendments (**Figure 2**), resulting in decreased bacterial diversity in all tDOM treatments (Supplementary Figure S1) and indicating that tDOM selected for and against certain species and clades. Bacterial diversity also decreased in control bioassays, but to a lesser degree, suggesting that bottle effects may also play a part in bringing about the observed declines in diversity. The tDOM bioassays diverged from the Control communities after 4 days in April and after 2.5 days in August (**Figure 2** and Supplementary Figure S2). A total of 138 bacterial OTUs (out of 1656) significantly (p < 0.05) increased or decreased during either April or August. Approximately 42 of these 138 OTUs (e.g., 30%) reached a relative abundance of > 1% in at least one sample, and belonged to phyla Actinobacteria, Bacteroidetes, Proteobacteria, or Verrucomicrobia (**Figure 6**). Sixteen of these 42 OTUs (38%) changed in either the Control or tDOM bioassays or both (**Figure 6**). The change in 14 of these 16 OTUs showed significant correlations (ρ > 0.5 or < –0.5 n = 16, p < 0.05) with either Chl a or inorganic nutrient concentrations indicating

a response to phytoplankton growth or ambient nutrient stocks (Supplementary Figure S3). While 5 of these 16 OTUs were also correlated with DOC concentration, the correlations with Chl a were greater (ρ > 0.8 or < –0.7, p < 0.05). For 7 out of the 16 OTUs that changed in both Controls and treatments, the change in relative abundance in the tDOM bioassays was at least 2-fold higher in the tDOM treatments than in the Controls, indicating that at least some of the increase in relative abundance was due directly or indirectly to the tDOM addition.

Here we highlight 20 OTUs that increased in the tDOM bioassays, but not in the Controls, and therefore, may be sentinels for increased terrestrial inputs to coastal Arctic systems (**Figure 7**). Specifically, 10 OTUs responded positively to the addition of tDOM and 5 of these were positively correlated with DOC concentrations (ρ > 0.56, n = 16 for each OTU, p < 0.05, **Figure 7**). The 10 OTUs that increased in relative abundance come from 2 bacterial phyla (**Figure 7**): Bacteroidetes (Polaribacter spp.) and three classes within the

FIGURE 3 | Bacterial response to tDOM additions. Time series of bacterial abundance (A) and bacterial production (B) data for the April bioassays where near coastal Arctic microbial communities were amended with terrestrially derived dissolved organic matter (tDOM). Data points are the mean (n = 2) ± half the range of duplicate samples. The tDOM treatment is depicted by the solid line and the Control treatment, which contains no additional DOM, is depicted by the dashed line.

Proteobacteria (Gammaproteobacteria, Alphaproteobacteria, and Epsilonproteobacteria).

Of the Gammaproteobacteria that significantly increased in the presence of tDOM, four OTUs, which increased in both April and August, were closely related to the family Colwelliaceae [OTU\_118, OTU\_593, OTU\_26099, OTU\_16026; (**Figure 6**)]. Three of these OTUs were closely related to the genus Colwellia (**Figure 7**). Another gammaproteobacterial taxa that increased in response to tDOM, but only in August, was an OTU related to Glaciecola sp. (OTU\_121). While we observed a decline in some groups of marine Roseobacter clade (e.g., Octadecabacter sp.; OTU\_8) in the tDOM bioassays, discussed in greater detail below, we observed an increase in other taxa, especially one related to order Rhodobacterales (OTU\_14). The relative abundance of an OTU closely related to Arcobacter, an Epsilonproteobacteria, increased in response to tDOM in both seasons (**Figure 7**).

While the aforementioned OTUs benefited from tDOM, the relative abundance of 10 other OTUs decreased in the tDOM treatment, but not in the Controls. The relative abundance of half of these OTUs were inversely correlated with DOC concentrations (ρ < –0.54, n = 16, p < 0.05), suggesting increased DOM inputs impede the growth of these OTUs relative to other members of the community. The negatively affected OTUs included taxa within Gammaproteobacteria, Alphaproteobacteria, Bacteroidetes, Betaproteobacteria, and Actinobacteria.

Several marine clades of Gammaproteobacteria, including SAR86 (OTU\_39, OTU\_21), OM182 (OTU\_63) and an Oceanospirillales-related OTU (OTU\_6) declined significantly (p < 0.05) with the addition of tDOM. Three Alphaproteobacteria OTUs (OTUs 8, 12, and 23) showed a significant decline in relative abundance in the tDOM-amended bioassays, including an OTU belonging to the family Pelagibacteraceae (OTU\_23), an Octadecabacter sp. (OTU\_8), and a member of the SAR116 clade (OTU\_12). Within the Bacteroidetes phylum, an OTU closely related to family Cryomorphaceae (OTU\_224, with 100% identity to the Antarctic fosmid ANT39E11; Grzymski et al., 2006) was found to decline throughout the course of the April bioassay.

#### DISCUSSION

Rivers, estuaries and coastal regions are particularly important for the transformation and degradation of tDOM (Raymond and Bauer, 2000; Battin et al., 2008). Approximately half of the DOC released by Arctic rivers is removed over the continental

shelves (Letscher et al., 2011). This project explored the effects of increased tDOM on coastal Arctic microorganisms under ice-covered and ice-free conditions. By investigating the effect of tDOM on microbial community composition we found that additions of tDOM selected for certain species and against others, altering community composition and decreasing bacterial diversity. Increases in tDOM discharge to the coastal Arctic will change both the biogeochemistry and microbial composition of coastal regions.

# Bulk Microbial Response to and Lability of tDOM

Concentrations of Chl a increased in all treatments indicating that the autotrophic communities were not nutrient limited at the time of collection. The lower initial Chl a concentration and low light level (4.7 µmol quanta−<sup>1</sup> m−<sup>2</sup> s −1 ) may have contributed to the lack of an additional response to the tDOM supplement in April. The divergence between the Chl a concentrations in the Control and in the tDOM treatments in August occurred

sequences in at least one sample. The OTU ID is noted to the left of the heatmap while the annotation, given as the lowest taxonomic level, is to the right of the heatmap. The bioassay (April or August), treatment (Control or tDOM) and timepoint (in days) are noted above the heatmap. Finally, for each OTU, asterisks within the heatmap denote a sample or pair of samples (duplicates) that have a relative abundance that is significantly greater or less than the Control.

after incubating for 3 days. Although phytoplankton may have been capable of directly using tDOM, the 3 day delay in growth may indicate that regenerated nutrients or critical microbial byproducts contributed to the observed change in Chl a. Tank et al. (2012) estimated that remineralized riverine N may support up 0.5 to 1.5 Tmol C yr−<sup>1</sup> of primary production.

To further support the importance of remineralized nutrients in Arctic microbial growth, we found that reduced forms of N (i.e., NH<sup>4</sup> <sup>+</sup> and DON) were more labile to the receiving community than NO<sup>3</sup> <sup>−</sup> which did not change in concentration in either treatment or season on the timescale tested, even though it was an order of magnitude higher in ambient waters in April than in August (**Figure 5**). This hypothesis is supported by high regeneration rates (15.2 nmol N L−<sup>1</sup> h −1 ) and NH<sup>4</sup> <sup>+</sup> uptake rates (14.1 nmol N L−<sup>1</sup> h −1 ) observed in a parallel study for this specific microbial community (Baer et al., 2017). Concentrations of NH<sup>4</sup> <sup>+</sup> decreased by approximately the same concentration (0.5 µM N) in both the Control and tDOM treatments in August, however, DON may have been remineralized into NH<sup>4</sup> <sup>+</sup> and then quickly used by the community with no net change in

concentration. Changes in nutrient concentrations are often not observed in tightly coupled regenerative systems (Sipler and Bronk, 2015).

Approximately 7 – 8% of the DOC supplied in this study was removed during the incubations. Because no change in the bulk DOC concentration was observed in the control treatments from either season it is likely that the majority of DOC lost during the incubation in the tDOM treatments was from the tDOM pool. The lability of the tDOM source to the summer community is possibly underestimated as phytoplankton growth in the tDOM treatment was observed for the final time point and plankton derived DOC production may have masked additional tDOM removal. The timing restrictions of our field season limited the duration of our incubations but it is likely, based on changes in bacterial abundance in spring and increasing Chl a concentrations at the final time point in summer, that additional degradation would occur with longer incubation times.

The 7 – 8% lability presented here is lower than riverine tDOM lability reported by other incubation studies (17 – 62%; Holmes et al., 2008; Mann et al., 2012; Vonk et al., 2013; Spencer et al., 2015). Some of this difference may be due to seasonal or regional differences in composition. For example, a study focusing on the lability DOC from Alaskan Arctic rivers showed DOC collected between July and August was ∼2 –9% labile when incubated in the dark for 90 days (Holmes et al., 2008). This proportion is similar to the 7 – 8% that we observed. Another study conducted in the Russian Arctic found that >50% of DOC in permafrost thaw streams collected in September was bioavailable to riverine bacterial communities on short 7 day time scales (Spencer et al., 2015). The distinct difference in lability over similar time scales between our study and Spencer et al. (2015) could be due to differences in the target microbial communities or DOM composition. For example, our study investigated the response of coastal marine communities while the other aforementioned studies focused on to freshwater riverine communities. The high bacterial production rates show that the community in our study was primed to use DOM but based on the complex compositional characteristics of the tDOM source many of the compounds supplied were likely less easily degraded than simple amino acids used for production estimates. Therefore it is expected that the rate of degradation decreased as the most labile fractions are removed (Cherrier et al., 1999).

Unfortunately, data showing changes in DOM composition over the incubation period for our study are not available

and direct comparisons of the FT-ICR MS data between our study and Spencer et al. (2015) are difficult as samples were analyzed in different ionization modes. What we can determine from comparing these studies is that a large proportion of compounds in both sources showed elemental characteristics of being moderately unsaturated and lignin-like, and both showed district clustering of some of the compounds in the lipid/ proteinlike or aliphatic region of the van Krevelen diagram (**Figure 1**). What cannot be determined from our current data is how many of these compounds are shared among both sites/ regions and if the same compounds are labile to both communities.

Another difference between our study and the previous studies is that we did not exclude phytoplankton from our incubations and incubated in the presence of light. The inclusion of primary production could mask the reduction of riverine DOC by supplying an additional source of DOC to a closed system but their presence also provides a more inclusive microbial response to tDOM additions by including both autotrophic and heterotrophic processes. We acknowledge that, like most bioassay studies, our conclusions are limited by the exclusion of higher trophic levels but feel that important information can be gained about potential impacts on the base of the coastal food web from our work assessing the response of the phytoplankton and bacterial community.

#### Seasonal Differences in In Situ Microbial Community Composition

The extreme physical and chemical differences observed between April and August likely contributed to the distinct variations between the April and August ambient microbial communities. Large seasonal changes in organic matter composition are observed throughout the Arctic (e.g., Crump et al., 2003; Connelly et al., 2015; Mann et al., 2016), and can influence the composition (Crump et al., 2003; Judd et al., 2006) and physiology of the ambient microbial community (Sipler et al., 2017). An evaluation of offshore sites found differences in community composition between seasons to be insignificant (Kirchman et al., 2010), which when combined with our results, further supports the importance of terrestrial influence (tDOM) in shaping coastal microbial communities. The observed variations in community composition are likely influenced by the dramatic seasonal changes in their physiochemical environment. Given the notable differences in the ambient microbial community compositions in April and August, it follows that the microbial response to tDOM will differ seasonally as well.

Gammaproteobacteria were the dominant phylum in both April and August. The largest contributors in April were OTUs related to the Oceanospirillaceae family. Members of this family have been shown to support phytoplankton growth in polar waters through the synthesis of cobalamin (vitamin B12) (Bertrand et al., 2015). Sequences closely related to open-ocean marine taxa (e.g., SAR86 and OM182 (Cho and Giovannoni, 2004) were more prevalent in the nutrient-poor August waters, similar to observations observed by Ghiglione et al. (2012) in Arctic and Antarctic waters.

Betaproteobacteria were more prevalent in April than in August. Members of Betaproteobacteria have been observed throughout Arctic surface waters, especially in the summer (Garneau et al., 2006; Ghiglione et al., 2012) and are often indicators of freshwater inputs. Therefore, their higher relative abundance in April, before the spring freshet, was unexpected. However, a recent study in the Beaufort Sea found that Betaproteobacteria composed up to 20% of bacterial cells sampled from ice-covered waters in February (Alonso-Sáez et al., 2014), suggesting the niche for different members of Betaproteobacteria is wider and potentially more diverse than previously thought.

Proportions of Alphaproteobacteria were similar between the two seasons. Phaeobacter sp. (Roseobacter clade), which was more dominant in August, has been shown to grow in association with phytoplankton (Brinkhoff et al., 2008). They promote algal growth through the synthesis and secretion of antibiotics and algal growth stimulants (Seyedsayamdost et al., 2011).

Bacteroidetes, dominated by Polaribacter spp., comprised ∼13.7% of the August community. Polaribacter are abundant and active members of coastal Arctic and Antarctic bacterial communities (Malmstrom et al., 2007; Ghiglione et al., 2012; Grzymski et al., 2012; Nikrad et al., 2012). Summer conditions appear to favor taxa belonging to the Polaribacter genus. For example, the proportion Polaribacter active in the uptake of simple DOC substrates is 7- to 10-fold higher in summer compared to winter (Nikrad et al., 2012). Several Polaribacter taxa have been observed to have gas vesicles (Gosink et al., 1998) and contain proteorhodopsin (González et al., 2008; Grzymski et al., 2012). Given their high relative abundances and notable activity in polar summer waters as well as their buoyancy-related adaptations, members of this taxa likely play a key role in summer C cycling in coastal Arctic ecosystems, perhaps benefiting from long days.

Verrucomicrobia (especially Coraliomargarita sp.) was another phylum that dominated August samples. The majority of Verrucomicrobia sequences in our samples belong to subdivision 4, which are observed in surface waters globally (Freitas et al., 2012). Verrucomicrobia have been shown to increase in relative abundance in response to phytoplankton-derived DOM inputs (Landa et al., 2014) and may be important polysaccharide degraders in marine environments (Martinez-Garcia et al., 2012).

Although constituting a smaller proportion of the starting communities, Deltaproteobacteria (SAR324), Actinobacteria (Acidimicrobiales sp., SVA0996 clade), and SAR406 also exhibited distinct seasonal patterns. The proportions of two taxa, SAR324 and Acidimicrobiales, that are important in winter N cycling in these waters (Connelly et al., 2014), were greater in April than in August. The higher abundance of SAR324 in April is not surprising as it is known to be a metabolically flexible taxa able to take advantage of diverse nutrient sources (Sheik et al., 2014). SAR406 was more abundant in April but negligible in August. Sequences of SAR406 have been recovered under the sea-ice east of our Barrow field site in the Beaufort Sea (Collins et al., 2010) and in surface waters elsewhere in the Arctic and Antarctic (Ghiglione et al., 2012). Both SAR324 and SAR406 tend to be present in higher abundances in deep waters, yet have also been observed in winter polar surface waters (Ghiglione

et al., 2012 and this study), suggesting that their distribution is not limited to deep waters, especially in polar environments. A metagenomic assessment of Antarctic surface waters also highlighted the dominance of chemolithoautotrophic pathways characteristic of these taxa in winter surface waters (Grzymski et al., 2012). The actinobacterial sequences detected in August were closely related to candidate genus "Aquiluna," a genus that was rare in April. Originally thought to be a freshwater genus (Hahn, 2009), sequences closely related to "ca. Aquiluna" have been recovered from marine environments, including from the Arctic (Kirchman et al., 2010; Kang et al., 2012) and have been observed to contain actinorhodopsin, suggesting the potential for photoheterotrophy by this taxa (Kang et al., 2012).

### Bacterial Response to tDOM Amendments

Additions of tDOM selected for certain bacterial taxa and against others, altering community composition and decreasing bacterial diversity. Members of Actinobacteria, Bacteroidetes, Proteobacteria, and Verrucomicrobia significantly responded to the addition of tDOM. Twenty OTUs were identified as potential sentinels for increased terrestrial inputs in coastal Arctic systems (**Figure 7**). Polaribacter spp. within Bacteroidetes, taxa within Gammaproteobacteria (Colwelliaceae family and Glaciecola spp.), a member of the alphaproteobacterial family Rhodobacteraceae, and an Epsilonproteobacteria related to Arcobacter sp. all increased in relative abundance in the presence of tDOM (**Figure 7**). Many of the taxa that responded positively to the tDOM additions are common within polar waters, are known to grow in association with phytoplankton, and thrive in high nutrient environments. Common in polar waters in both spring (Malmstrom et al., 2007) and summer (Ghiglione et al., 2012; Grzymski et al., 2012), Polaribacter have been linked to the uptake and decomposition of complex polymeric DOM (Kirchman, 2002). Members of this clade are also often observed in association with polar phytoplankton blooms (Teeling et al., 2012; Landa et al., 2016). Here we extend evidence of their success to high-tDOM environments. Given their ability to thrive in the presence of phytoplankton-derived DOM or tDOM, the Polaribacter genus may be resilient to the ongoing changes in coastal Arctic waters.

The Gammaproteobacteria that significantly increased in the presence of tDOM were closely related to the family Colwelliaceae (**Figure 6**), specifically the genus Colwellia (**Figure 7**). Although this family was not very abundant in the starting communities in this study, it is clear that they may become dominant members of polar microbial communities during times of high DOM inputs. Colwellia has a diverse array of metabolic strategies aiding its survival in Arctic waters including blooming in response to phytoplankton-derived DOM (Landa et al., 2016), assimilating bicarbonate (Alonso-Sáez et al., 2010), production of extracellular enzymes capable of degrading high molecular weight compounds (Methé et al., 2005), degrading sinking organic matter (Delmont et al., 2014), and responding to tDOM inputs as described in this study. Glaciecola sp. (Gammaproteobacteria) also increased in response to tDOM. This genus is capable of growing in association with complex DOM sources. For example, this genus has been found in sea ice (Brakstad et al., 2008) and sea water contaminated with crude oil (Chronopoulou et al., 2015), appearing to be able to survive in systems with high hydrocarbon concentrations. Arctic peats naturally contain hydrocarbons (Yunker et al., 1993); therefore, it is not surprising that 16% of the DOM compounds within the tDOM source had elemental characteristics similar to Aromatics and 15% as condensed Aromatics that could have provided a source of DOM for Glaciecola sp.

Roseobacters are a metabolically diverse clade of Alphaproteobacteria (Buchan et al., 2005; Newton et al., 2010). While we observed a decline in some groups of Roseobacter (e.g., Octadecabacter sp.) in the tDOM bioassays, discussed in greater detail below, we observed an increase in other taxa, especially an OTUs related to Rhodobacterales, further highlighting the metabolic breadth of this important clade of marine bacteria. Our results suggest that some taxa within the globally important Roseobacter clade will thrive in a changing Arctic Ocean.

Arcobacter, an Epsilonproteobacteria, also increased in response to tDOM. In marine environments, Arcobacter-related sequences are often associated with metazoans (Gugliandolo et al., 2008; Fontanez et al., 2015), yet, the Arcobacter that responded to tDOM was more closely related (96% similar) to a sulfur-oxidizer (Candidatus Arcobacter sulfidicus) (Wirsen et al., 2002). Because waters between 1 and 18 m depth account for 17% of Arctic shelves (including the site where our seawater samples were collected), and because wave action is expected to increase in the coastal Arctic due to sea ice loss (Overeem et al., 2011), taxa that thrive on tDOM or near sediments (e.g., sulfur-oxidizers) may become more dominant in the water column in the future.

While several OTUs benefited from tDOM, others, including taxa within Gammaproteobacteria, Alphaproteobacteria, Bacteroidetes, Betaproteobacteria, and Actinobacteria, were negatively affected by the presence of tDOM. More than half of the OTUs that decreased in relative abundance in the tDOM bioassays were inversely correlated with DOC concentrations. This suggests that increased DOM inputs will impede the growth of these OTUs relative to other members of the community.

Several marine clades of Gammaproteobacteria that declined with the addition of tDOM have previously shown diverse responses to nutrients and DOM sources. For example, SAR86, which has been shown to have mixed responses to phytoplankton blooms (Obernosterer et al., 2011; Tada et al., 2011; Rinta-Kanto et al., 2012; Wemheuer et al., 2014; Landa et al., 2016), are found in higher concentrations in the open ocean rather than in Arctic coastal waters (Ghiglione et al., 2012). Likewise, members of the OM182 clade exhibit higher growth rates in lower C environments (Cho and Giovannoni, 2004). Given their preference for oligotrophic waters, it is not surprising that SAR86 and OM182 OTUs decreased in the tDOM bioassays. Similar to SAR86, members of the Oceanospirillales order have shown mixed responses to DOM amendments suggesting global variation in how these taxa, or ecotypes within these taxa, respond to tDOM inputs. Some studies observed a

response of Oceanospirillales to both phytoplankton-derived and plant-derived organic matter (Mou et al., 2008), but other studies show that some members of this order decrease in relative abundance in response DOM amendments (Landa et al., 2013). The Oceanospirillales-related OTU in our study was negatively correlated with DOC concentration and significantly decreased in both the April and August tDOM incubations.

This study highlights the ecological complexity of Arctic microbes. As discussed above, some OTUs that are known to respond positively to phytoplankton blooms and presumably phytoplankton derived DOM decreased in relative abundance while others increased. Having access to preferred forms of DOM does not mean that the other biogeochemical conditions remain favorable enough for certain species to take advantage of the resource. High levels of DOM, and the metals that they chelate may (1) inhibit some species that benefit from the lower concentration tDOM additions or (2) species that benefit from tDOM additions may simply outcompete or actively inhibit other species neutral to tDOM additions. The underlying mechanisms that control microbial community composition are not fully understood but these mechanisms are likely much more complex than C availability alone.

Alphaproteobacteria, including OTUs belonging to the Pelagibacteraceae family, an Octadecabater sp., and a member of the SAR116 clade, also showed a significant decline in relative abundance in the tDOM amended bioassays. Pelagibacteraceae taxa specialize in the degradation of fresh, low molecular weight dissolved organic matter (Malmstrom et al., 2005; Poretsky et al., 2010), while the tDOM in this study represents a complex mixture of predominantly lignin -like compounds (**Table 1**), that may not favor the growth of Pelagibacteraceae or other oligotrophs. Similarly, Octadebacter sp. appear to lack the machinery needed to degrade aromatics (Newton et al., 2010), an important characteristic for utilizing many of the tDOM compounds. Paralleling our observations, SAR116 activity was observed to be low in eutrophic temperate coastal waters (Campbell et al., 2009) and ribotypes of this clade have been observed to have higher abundances in open ocean waters during highly oligotrophic periods (Morris et al., 2005).

Within the Bacteroidetes phylum, an OTU closely related to the Cryomorphaceae family (with 100% identity to the Antarctic fosmid ANT39E11; (Grzymski et al., 2006) was found to decline throughout the course of the ice-covered bioassay despite members of this family being known to thrive in high organic matter environments. Members of this taxa, as well as SAR86, the order Oceanospirillales and the Pelagibacteraceae family, have been previously shown to thrive in high DOM environments but the observed decrease in our samples highlights the importance of not generalizing an organism's preference for high organic matter environments to all types or sources of organic matter. Even microorganisms that thrive in high DOM waters may not do well in polar waters with a high fraction of terrestrial or tundra-derived organic matter.

Although we have focused on OTUs that only responded positively or negatively to the tDOM additions, approximately half of the OTUs that changed in both treatments, increased more in the tDOM treatments than the Controls, indicating that at least a fraction of their increase in relative abundance was due directly or indirectly to the tDOM addition and not solely attributed to bottle effects. For example, commonly considered generalist taxa (Mou et al., 2008), OTUs belonging to the gammaproteobacteria family Colwelliaceae increased in both Control and tDOM bioassays in our study, but the increase in the tDOM bioassays far exceeded that in the Control bioassays (**Figure 6**), highlighting their potential importance in taking advantage of increasing inputs of tDOM in a changing Arctic Ocean. Because phytoplankton (Chl a) increased in the August tDOM treatments it is hard to differentiate the direct impacts of tDOM or the indirect influence of tDOM promoting phytoplankton growth/ exudates. We argue that this study addresses the overall response, direct and indirect, of target coastal microbial communities to tDOM additions. Several of the OTUs that changed in both the Control and tDOM bioassays significantly correlated with Chl a concentrations or inorganic nutrient concentrations validating their response to phytoplankton growth or ambient nutrient stocks (Supplementary Figure S3). Similarly, while alpha diversity declined in both the Control and tDOM treatments, the magnitude of these decreases was greater in tDOM treatments. This suggests that tDOM amendments enhanced selection for bacterial taxa beyond that which was observed in the controls alone indicating that this selection was not only the result of bottle effects.

Seasonal differences in how quickly microbial communities responded to tDOM were apparent, with August tDOM-amended community composition diverging from control communities more rapidly than April communities (Supplementary Figure S2). It is possible that some of these differences may be temperature driven, given that small increases in temperature can have notable effects microbial metabolism (e.g., Kirchman et al., 2005), but our experimental design does not allow for this effect to be quantified. Nevertheless, our results suggest that summertime microbial communities may be poised to respond more rapidly to pulses of tDOM than communities in early spring. Regardless of season, the addition of tDOM generally shifted bacterial communities toward more copiotrophic and away from more oligotrophic taxa. No one order was found to respond universally (positive or negative) to the tDOM addition. These data suggest that not all sources of DOM (phytoplankton DOM vs. tDOM) are created equal and generalization about the potential impact of a source (e.g., DOM) to a specific phylum, family or clade could be misleading.

# Implications for a Warmer More tDOM Rich Arctic

We know that Arctic atmospheric temperatures are rising (Serreze and Francis, 2006; Screen and Simmonds, 2010), that permafrost reserves are thawing (Smith et al., 2005, 2010), and that tDOM inputs to the coastal Arctic are increasing (Fichot et al., 2013). Here we provide evidence that these increases in tDOM to coastal waters of the Arctic Ocean will likely yield a shift in microbial community composition. What remains unknown is how all of these changes will impact global C, N, and phosphorus (P) cycles.

The most obvious impact for the global C cycle is the release of carbon dioxide associated with the respiration and degradation of this abundant and increasing source of organic C. However, a less obvious consequence for the C cycle includes a potential changes in autotrophic and heterotrophic C-fixation. Several studies suggest that increased freshwater discharge will cause an overall reduction in coastal production (e.g., Balch et al., 2012, 2016; Wikner and Andersson, 2012); however, the present experiment provides evidence that some late season Arctic phytoplankton may benefit from tDOM additions through regenerated nutrients. No such benefit was observed in spring when light levels are low. Therefore, the positive Chl a response observed at the final summer time point may be indicative of a seasonal or community dependent response but that the overall impact on production may be negative.

The light absorption by the tDOM ranges into the visible spectrum, which decreases light needed for photosynthesis (Arrigo and Brown, 1996). The light attenuation associated with tDOM may be more physiologically consequential in spring when light is already limiting productivity (Sipler et al., 2017). A companion study Sipler et al. (2017) found that the addition of tDOM reduced C-fixation during a 24 h incubation. Additionally, several of the above-described Alpha- and Gammaproteobacterial taxa, including Octadecabacter and Oceanospirillales spp., that declined in abundance when tDOM was added, are important heterotrophic dissolved inorganic C-fixers (DeLorenzo et al., 2012). While other heterotrophic C-fixers, such as Colwellia sp. (Alonso-Sáez et al., 2012) or Polaribacter sp. (DeLorenzo et al., 2012), were observed to increase in response to tDOM, indicating increasing tDOM inputs may have mixed effects on heterotrophic C-fixation. Although autotrophic C-fixation dominates Arctic primary productivity, heterotrophic C-fixation may be important in seasons or regions (depth) when light is limiting (Alonso-Sáez et al., 2010; Herndl and Reinthaler, 2013).

Indirect impacts of DOM on primary production may also include promoting (Bertrand et al., 2015; Mishamandani et al., 2016) or potentially inhibiting co-existent or mutualistic bacterial species essential for phytoplankton growth. For example, reductions in specific taxonomic groups (e.g., Oceanospirillaceae sp.) may lead to reductions in bacterially synthesized compounds, including cobalamin, that are essential for phytoplankton growth (Bertrand et al., 2015). Thus, a decline in the abundance and activity of these organisms in an Arctic Ocean faced with increased inputs of tDOM may impact not only the cycling of organic C but also the sequestration of inorganic C.

Nutrient budgets will also be affected by increases in tDOM. A recent study investigated the potential for regenerated terrestrial-dissolved organic nutrients to balance Arctic Ocean inorganic nutrient budgets (Torres-Valdés et al., 2016). Although DON inputs appear insufficient in supplying enough N to offset the rather large denitrification deficit (Chang and Devol, 2009) in the Arctic Ocean, terrestrial DON and DOP sources do play an important role in Arctic and North Atlantic N and P cycles. Because microbial communities mediate the magnitude and rate of DOM regeneration, understanding how DOM and microbial pools will change in relationship to one another is critical to predicting the impacts of future environmental perturbations and trends (Kujawinski et al., 2016).

The extent and magnitude at which tDOM inputs increase in the Arctic may determine the future productivity of this sensitive ecosystem. In this interconnected system (Wassmann et al., 2011), changes in phytoplankton and bacterial community composition have enormous consequences for higher trophic levels locally and across the Arctic. Beyond the community changes we observed in response to tDOM amendments, more work is necessary to elucidate how these changes affect community metabolism, ecosystem function, and process rates important in the cycling of globally important elements.

#### CONCLUSION

Changes in heterotrophic production and microbial community composition are expected as substantial reserves of tDOM, released from once stable permafrost, become available to coastal microbial communities. Unlike sea ice melt, which has been speculated to cause short term increases in primary production, a large scale loading of a C-rich source may increase carbon dioxide release via respiration and impact the cycling of both organic and inorganic nutrients. Here we have shown that the addition of tDOM promoted certain species and inhibited others, decreased bacterial diversity, thus shifting the microbial community composition. At least 7% of the terrestrial-DOC supplied during this study was consumed by both ice-covered and ice-free microbial communities. The terrestrial-DON fraction was proportionately more labile with ∼ 38% of the supplied DON being removed during the relatively short (4 – 6 day) incubations. Through this effort we have identified 20 OTUs that significantly changed in response to the tDOM addition that may serve as sentinels for future environmental change. Our work is a step forward in identifying microbial shifts that may indicate larger future impacts. Additional studies are needed to investigate temporal and spatial trends of these species as they relate to coastal humification, to confirm their role as true sentinels of change in the Arctic.

#### AUTHOR CONTRIBUTIONS

RS, DB, and TC designed the study; RS, DB, TC, and QR collected the samples; RS, CK, TC, and QR analyzed the data; RS, CK, and TC wrote the paper; all authors helped interpret the data and edited the paper.

# FUNDING

This work was supported by the United States National Science Foundation grants ARC-0909839 and 0910252. We also acknowledge the funding sources of the WHOI FT-MS Users Facility (National Science Foundation MRI program OCE-0619608 and the Gordon and Betty Moore Foundation) for assistance with MS data acquisition.

#### ACKNOWLEDGMENTS

fmicb-08-01018 June 8, 2017 Time: 16:6 # 16

We thank M. Frischer, K. Sines, S. Baer, Z. Tait, and L. Killberg-Thoreson for their help in the field. We also thank CH2MHill, the UMIAQ science support staff and the Barrow Whaling Captains Association for their field and logistical support. We thank B. Crump for generously providing the resources for extracting DNA from and sequencing the samples involved in this project. We also acknowledge E. Kujawinski, M. Soule, K. Longnecker and the funding sources of the WHOI FT-MS Users Facility (National Science Foundation MRI program OCE-0619608 and the Gordon and Betty Moore

#### REFERENCES


Foundation) for assistance with MS data acquisition. This work is supported by NSF grant# ARC-0909839 to DB and ARC-0910252 to PY. This paper is Contribution No. 3638 of the Virginia Institute of Marine Science, College of William & Mary.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.01018/full#supplementary-material


Antarctic marine planktonic bacteria. Appl. Environ. Microbiol. 72, 1532–1541. doi: 10.1128/AEM.72.2.1532-1541.2006


fmicb-08-01018 June 8, 2017 Time: 16:6 # 17



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Sipler, Kellogg, Connelly, Roberts, Yager and Bronk. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Carbon Bioavailability in a High Arctic Fjord Influenced by Glacial Meltwater, NE Greenland

Maria L. Paulsen<sup>1</sup> \*, Sophia E. B. Nielsen2, 3, Oliver Müller <sup>1</sup> , Eva F. Møller 4, 5 , Colin A. Stedmon<sup>2</sup> , Thomas Juul-Pedersen<sup>6</sup> , Stiig Markager <sup>4</sup> , Mikael K. Sejr <sup>5</sup> , Antonio Delgado Huertas <sup>7</sup> , Aud Larsen<sup>8</sup> and Mathias Middelboe<sup>3</sup>

<sup>1</sup> Department of Biology, University of Bergen, Bergen, Norway, <sup>2</sup> National Institute for Aquatic Resources, Technical University of Denmark, Charlottenlund, Denmark, <sup>3</sup> Marine Biological Section, University of Copenhagen, Helsingør, Denmark, <sup>4</sup> Department of Bioscience, Aarhus University, Roskilde, Denmark, <sup>5</sup> Arctic Research Centre, Aarhus University, Aarhus, Denmark, <sup>6</sup> Climate Research Centre, Greenland Institute of Natural Resources, Nuuk, Greenland, <sup>7</sup> Instituto Andaluz de Ciencias de la Tierra (CSIC-UGR), Armilla, Granada, Spain, <sup>8</sup> Uni Research Environment, Bergen, Norway

#### Edited by:

Ingrid Obernosterer, FR3724 Observatoire Océanologique de Banyuls sur Mer (OOB), France

#### Reviewed by:

Emma Jane Rochelle-Newall, Institut de Recherche Pour le Développement, France Xosé Anxelu G. Morán, King Abdullah University of Science and Technology, Saudi Arabia

\*Correspondence:

Maria L. Paulsen maria.l.paulsen@uib.no

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Marine Science

Received: 18 October 2016 Accepted: 19 May 2017 Published: 08 June 2017

#### Citation:

Paulsen ML, Nielsen SEB, Müller O, Møller EF, Stedmon CA, Juul-Pedersen T, Markager S, Sejr MK, Delgado Huertas A, Larsen A and Middelboe M (2017) Carbon Bioavailability in a High Arctic Fjord Influenced by Glacial Meltwater, NE Greenland. Front. Mar. Sci. 4:176. doi: 10.3389/fmars.2017.00176 The land-to-ocean flux of organic carbon is increasing in glacierized regions in response to increasing temperatures in the Arctic (Hood et al., 2015). In order to understand the response of the coastal ecosystem metabolism to the organic carbon input it is essential to determine the bioavailability of the different carbon sources in the system. We quantified the bacterial turnover of organic carbon in a high Arctic fjord system (Young Sound, NE Greenland) during the ice-free period (July-October 2014) and assessed the quality and quantity of the 3 major organic carbon sources; (1) local phytoplankton production (2) runoff from land-terminating glaciers and a lowland river and (3) inflow from the ocean shelf. We found that despite relatively low concentrations of DOC in the rivers, the bioavailability of the river–DOC was significantly higher than in the fjord, and characterized by high cell-specific bacterial production and low C:N ratios. In contrast, the DOC source entering via inflow of coastal shelf waters had high DOC concentrations with high C:N and low specific bacterial production. The phytoplankton production in the fjord could not sustain the bacterial carbon demand, but was still the major source of organic carbon for bacterial growth. We assessed the bacterial community composition and found that communities were specific for the different water types i.e., the bacterial community of the coastal inflow water could be traced mainly in the subsurface water, while the glacial river community strongly dominated the surface water in the fjord.

Keywords: bacterial carbon demand, bacterial diversity, dissolved organic matter, runoff, glacial meltwater, high arctic ecosystems, young sound

# INTRODUCTION

Carbon consumption and mineralization by pelagic heterotrophic bacteria play a key role in marine ecosystems. Increasing temperatures, with pronounced effects at high latitudes, have raised questions about how reduced ice-cover and increased runoff can affect the ecosystem metabolism i.e., the balance between respiration and primary production. Bacterial carbon turnover was traditionally suggested to be limited by low temperature in high latitude systems, and consequently play only a minor role in turning over primary production (Pomeroy and Deibel, 1986; Pomeroy et al., 1991). Several high latitude studies during the past decade have however reported bacterial production rates similar to those reported in lowlatitude non-oligotrophic systems (Børsheim, 2000; Rysgaard and Nielsen, 2006; Sejr et al., 2007). Two seasonal studies found annual ranges in bacterial production of 5–42 mg C m−<sup>2</sup> d −1 in Kobbefjord (64◦N) (Middelboe et al., 2012) and 90–165 mg C m−<sup>2</sup> d −1 in Kongsfjorden, Svalbard (78◦N) (Iversen and Seuthe, 2011), with maximum values during spring, coinciding with the spring phytoplankton bloom.

Estimates of bacterial carbon cycling is often based on measurements of net bacterial production, however variability in the factors used to convert radioisotope (e.g., <sup>14</sup>C-leucine or <sup>3</sup>H-thymidine) incorporation to carbon production affects the growth estimates and potentially complicates comparison of studies. Measurements of the bacterial respiration (BR) are required in order to estimate the bacterial carbon demand (BCD) and growth efficiency (BGE). Such measurements in Arctic systems are few and the BGE reported are highly variable, but all in the low end of those reported in other aquatic systems (del Giorgio and Cole, 1998). It has been hypothesized that low temperatures limit substrate uptake and consequently argued that Arctic bacteria need relatively higher concentrations of carbon to grow at low temperatures (Pomeroy and Wiebe, 2001). Other studies have demonstrated an inverse relationship between BGE and temperature (Rivkin and Legendre, 2001; Apple et al., 2006) leading to speculations that the low efficiency found in the Arctic may instead be a result of poor quality carbon sources (Middelboe et al., 2012). While the concentration of organic matter alone does not reflect the carbon quality (Kirchman et al., 2005), the elemental ratios of the dissolved organic matter (DOM) i.e., the C:N ratio has provided insight on the DOM bioavailability (del Giorgio and Cole, 1998; Pradeep Ram et al., 2003; Kragh and Søndergaard, 2004).

The activity of bacteria is tightly coupled to DOM bioavailability (Amon and Benner, 1996; Kragh and Søndergaard, 2004). Bioavailable dissolved organic carbon (BDOC) has been estimated to constitute on average <1% of the oceanic DOC pool, however elevated in the surface waters (Hansell, 2013). Studies in the Greenland Sea, Fram Strait, and Kobbefjord have shown that BDOC constitutes 13–36% of total DOC in surface water (Middelboe and Lundsgaard, 2003; Middelboe et al., 2012; Jørgensen et al., 2014). Glacial meltwater from both Alaskan (Hood et al., 2009) and Alpine glaciers (Singer et al., 2012) contained highly bioavailable (>60%) DOC. As the Greenland Ice Sheet is melting at record speed (Nghiem et al., 2012) and the melt is projected to continue increasing (Keegan et al., 2014) it poses the question whether coastal bacterial carbon turnover will increase and drive the fjord systems toward more heterotrophic conditions in the future.

The bioavailability of different DOM types is influenced by a number of factors including composition of substrate, availability of mineral nutrients and the bacterial communities and their enzymatic capabilities (Middelboe and Lundsgaard, 2003; Kritzberg et al., 2010; Traving et al., 2016). Only few studies have tried to directly connect specific bacterial groups to different types of DOM (Kirchman et al., 2007; Baña et al., 2013; Osterholz et al., 2016). Meltwater from glaciers has been found to significantly modify the structure of microbial communities in the connected fjord (Gutiérrez et al., 2015). Consequently, increased runoff associated with warming climate will not only affect the transport of organic matter, it may also change the dynamics of coastal bacterial communities toward a higher influence of riverine bacteria and thus potentially changes in BGE and DOM degradation of the coastal bacterial community (Fortunato et al., 2013). Exploring links between the bacterial community composition and the various DOM sources, are therefore highly relevant in Arctic environments. Young Sound receives most of its runoff from the Greenland Ice Sheet via land terminating glaciers (Citterio et al., 2017), resulting in a clear gradient of allochthonous sources of both organic matter and silt throughout the fjord (Murray et al., 2015). The organic carbon sources in the fjord comprise two allochthonous carbon sources 1) meltwater from land-terminating glaciers and a lowland river and 2) coastal water that contains traces of DOM from the Arctic Ocean (Amon and Budéus, 2003).

Based on previous findings we hypothesize that the glacial runoff in Young Sound contains highly bioavailable DOM compared to the local production and inflowing coastal water. In order to understand the response in ecosystem metabolism we evaluate the importance of each of the 3 carbon sources as substrate for bacterial carbon degradation and examine associations between the bacterial DOM degradation and the genetic diversity.

# MATERIALS AND METHODS

#### Study Site and Sampling

The study was conducted in the high arctic fjord Young Sound, NE Greenland (74.2–74.3 ◦N, 19.7-21.9 ◦W). A sill of 45 m depth separates the deeper parts of the fjord from the coastal shelf waters, which are influenced by the East Greenland Current (for more info see Rysgaard et al., 2003). Sampling was conducted at four stations located along a length section from the inner fjord (St. 1) to the shelf waters on the outer side of the sill (St. 4) (**Figure 1**). The stations 1, 2, 3, and 4 are located according to stations monitored yearly by the Greenland Ecosystem Monitoring (GEM) MarinBasis Zackenberg programme in which they are named Tyro 05, YS 3.18, Standard St. and GH 05, respectively. The fjord stations were each sampled approximately every 10th during the early ice-free period (July 15–August 7) and the late summer period before new ice formation (September 4–October 4) (Figure S1).

The first sampling was conducted prior to the sea-ice breakup through a hole in the ice at St. 3, when only the central part of the fjord was still ice covered (see satellite photos **Figure 1**). The ice broke up in the central part on July 15 and the fjord rendered ice-free within 24 h. The remaining sampling was carried out from the research vessel Aage V. Jensen using mini rosette with 12 × 1.7 L Niskin bottles from 6 standard depths (1, 10, 20, 30, 40, and 100 m) and 1–2 additional depths at the deep chlorophyll maximum (DCM) when this did not overlap with one of the standard depths. The DCM was determined

prior to every sampling using a Satlantic Free-falling Optical Profiler (Murray et al., 2015). A Seabird SBE 19+ CTD profiler was deployed at every sampling occasion and recorded vertical profiles of temperature (◦C), salinity (ratio; no units), chlorophyll fluorescence (fluchl, relative; no units), turbidity (FTU), and photosynthetically active radiation (PAR, µmol m−<sup>2</sup> s −1 ) at every sampling occasion. The light attenuation was estimated from the CTD-profiles using a two-phase Weibull function as described in Murray (2015).

Glaciers cover ca. 33% of the drainage area of the fjord and land-terminating glaciers contribute 50–80% of the annual terrestrial runoff with highest contribution in the inner fjord (Bendtsen et al., 2014; Citterio et al., 2017). Three of the major rivers discharging into the fjord were sampled as long as sufficient water was flowing (last sampling was on September 10). The meltwater in the Tyroler river (R1) and Clay Bay river (R2) flows from glaciers trough rocky sediment basins with close to zero vegetation for a distance of ca. 0.5 and 2 km, respectively, before they reach the fjord. A model study estimate the residence time of river water in the fjord to be about 2 weeks in July and up to a month in August (Bendtsen et al., 2014).

R1 receives water directly from the Greenland Ice Sheet and has the largest catchment area (Bendtsen et al., 2014), while R2 receives meltwater from smaller local glaciers (**Figure 1**). The largest river, R3 (Zackenberg river) has the second largest catchment area and is connected to 2 lakes. It flows through lowland permafrost soils covered with vegetation types like dwarf shrub heath (Salix arctica) and grasses (e.g., Arctagrostis latifolia), however the riverbed is rocky and without vegetation (Elberling et al., 2008). River water was collected just below surface in 5 L plastic bottles. A 10-year time series of temperature and the organic and inorganic particulate biomass and the dissolved organic carbon (DOC) recorded from the Zackenberg river by the GeoBasis programme by Greenland Ecosystem Monitoring are included in **Figure S3**.

A total of 25 profiles were sampled at the fjord stations and the rivers were sampled each 3 times. Based on the salinity and temperature (**Figure 2**) five water types were defined (**Table 1**). Bacterial abundance, production and chemical parameters (nutrients, DOC, DON, and chl a) were measured in total 174 times each. A total of 42 samples were collected in the rivers and at 1 m and DCM for "extra" analysis of carbon bioavailability, particle associated bacterial production and community composition analysis. These "focus samples" are marked with large symbols in **Figure 2**. Environmental data associated with these 42 focus samples are given in **Table S1**. Note no focus samples were collected from the water mass defined as Shelf water.

#### Chlorophyll a and Primary Production

Concentrations of chlorophyll a (chl a) were determined according to Jespersen and Christoffersen (1987). Triplicates of 250 mL water was filtered onto GF/F, 2 and 10µm polycarbonate filters and chl a was extracted in 5 mL 96% ethanol for 12– 24 h and analyzed on a Turner Design Fluorometer calibrated against a chl a standard. The measurements were done in

Shelf water. The season is split into two periods: July-August (orange) and September-October (blue). All data points from CTD casts throughout the fjord is shown as small triangles, while larger symbols (circles and triangles) represent the "focus depths" which include measurements of bioavailability, particle associated production and bacterial community composition. The river samples (triangles) are labeled according to origin; Tyroler river (R1), Clay Bay river (R2), and Zackenberg river (R3) and their salinity is set to 0.

TABLE 1 | Thermohaline properties describing the classification of water types.


triplicates. Primary production (PP) was measured as <sup>14</sup>C-uptake (Nielsen, 1952) according to Markager et al. (1999) and measured for 3 size fractions; dissolved (<0.7µm), small phytoplankton (0.7–10µm) and lager phytoplankton (>10µm). Samples were collected at 1 m depth and at one or two additional depth with a notable DCM (26 samples in total). The areal primary production was calculated according to Lyngsgaard et al. (2014). The daily area production was estimated by integrating over 24 h and with depth. The light intensity at each depth was calculated from the light attenuation and the surface light measured at the nearby Zackenberg research station as part of the GEM Programme.

# Flow Cytometry

The abundance of bacteria was determined on an Attune <sup>R</sup> Acoustic Focusing Flow Cytometer (Applied Biosystems by Life technologies) with a syringe-based fluidic system and a 20 mW 488 nm (blue) laser. Samples were fixed with glutaraldehyde (0.5% final conc.) and kept dark at 4◦C until analysis within 12 h. Samples were stained with SYBR Green I (Molecular Probes, Eugene, Oregon, USA) for min 0.5 h at low flow rate of 25 µL min−<sup>1</sup> following the protocol of Marie et al. (1999). It should be noted that the bacterial abundance in the rivers was not easily counted by flow cytometry as the inorganic particle signal was high and obscured the counts of free-living bacteria. To reduce the problem the samples were diluted x10 with TEbuffer.

# Nutrients

Unfiltered seawater was filled directly from the Niskin bottles into 30 mL acid washed HDPE bottles and stored at −20◦C. Nitrite and nitrate (NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 ), phosphate (PO3<sup>−</sup> 4 ) and silicic acid (H4SiO4) were measured on a Smartchem200 (by AMS Alliance) autoanalyser following procedures as outlined in Wood et al. (1967) for NO<sup>−</sup> <sup>3</sup> <sup>+</sup>NO<sup>−</sup> 2 , Murphy and Riley (1962) for PO3<sup>−</sup> 4 and Koroleff (1983) for the determination of H4SiO4. Concentration of NH<sup>+</sup> <sup>4</sup> was determined directly in fresh samples using orthophthaladehyde according to Holmes et al. (1999).

#### Organic Matter Concentration

DOC and DON samples for determining the initial concentration of DOC were collected in 60 mL acid washed HDPE (highdensity polyethylene) bottles and stored frozen (−20◦C) until analysis. DOC is here considered to equal total organic carbon (TOC) according to (Anderson, 2002). DOC concentrations were determined by high temperature combustion (720◦C) using a Shimadzu TOC-V CPH-TN carbon and nitrogen analyser calibrated using a standard series of acetoanilide and the accuracy of the instrument was evaluated using seawater reference material provided by the Hansell CRM (consensus reference material) program. DON was calculated by subtraction of inorganic nitrogen. For particulate organic carbon and nitrogen (POC and PON) a total of 85 samples were collected. 15 L samples from the rivers and 30 L samples from the fjord (collected at 1 m, DCM and 100 m) were filtered onto 47 mm pre-combusted GFF filters using a peristaltic pump. The filters were placed in a desiccator containing concentrated 37% HCl for 12–14 h to remove inorganic carbon. POC and PON was measured using a Carlo Elba NC1500 (Milan, Italy) CHN elemental analyser following the method of Hedges and Stern (1984).

# Bacterial Production

Bacterial production was estimated from incorporation of <sup>3</sup>Hthymidine (Riemann et al., 1982). From each water sample, four replicates of 10 mL unfiltered seawater samples were transferred to 20 mL plastic vials. One replicate was immediately amended with 500 µL of 100% trichloroacetic acid (TCA) and served as control. Samples were incubated with 10 nM <sup>3</sup>H-thymidine (final concentration) for 3–5 h at in situ temperature and stopped by addition of 500 µL 100% TCA. Samples were filtered onto 0.2 µm cellulose-nitrate filters, which were subsequently washed 10 times with ice-cold 5% TCA. Filters were transferred to 6-mL plastic vials and stored at −20◦C until analysis. In the laboratory 5 mL of scintillation liquid was added and the radioactivity was counted on a Perkin Elmer Liquid Scintillation Analyzer Tri-Carb 2800TR. The measured thymidine incorporation was converted to cell production assuming 2.0 x10<sup>18</sup> cells produced per mole <sup>3</sup>H thymidine incorporated (Fuhrman and Azam, 1980). Bacterial population growth rate (GR) was calculated as cell production (cells mL−<sup>1</sup> d −1 ) divided by the cell abundance (cells mL−<sup>1</sup> ). Cell production was converted to bacterial carbon production (BP) assuming 2.0 × 10−<sup>14</sup> g C cell−<sup>1</sup> (Lee and Fuhrman, 1987). For the calculation of area-integrated bacterial carbon consumption, the bacterial production was depth-integrated across the upper 100 m, and divided by the BGE measured in July (see below). Additionally, at the "focus depths" and in the rivers the particulate bacterial production was measured as the fraction associated with particles larger than 3 µm and calculated as the total BP subtracted the <3 µm fraction.

# Bacterial Respiration and Growth Efficiency (BGE)

Bacterial respiration (BR) was measured as oxygen consumption for ∼48 h at constant temperature in water from 1 m and DCM in triplicate 12 mL gas tight Exetainers equipped with an optical sensor. Water was pre-filtered through a 3 µm- polycarbonate filter to reduce bias from eukaryotic cell respiration and grazing. A sample with 20 µL HgCl served as control. Oxygen was measured every 5 min for >24 h using a 4-channel Fiber-Optic Oxygen Meter (FireSting, Pyroscience) using the program Pyro Oxygen Logger Software version 2.37 (PyroScience). Respiration rates were calculated as the decrease in oxygen concentration (µM) over the incubation time after subtracting control values. Conversion from oxygen consumption to carbon respiration was done assuming a respiratory quotient (RQ) of 0.82 (Søndergaard and Middelboe, 1995). BGE was calculated from measurements of net bacterial production (NBP) and bacterial respiration (BR) in the 3 µm-filtered samples as: BGE (%) = BP/ (BP+BR) × 100. BGE was measured at all four stations at 1 m depth and DCM, and from the three rivers, however only BR measurements that fulfilled the three following criteria were used: (1) constant incubation temperature (± 0.1◦C) for > 24 h, (2) low abundance (<5,000 cells mL−<sup>1</sup> ) of small phytoplankton (<3 µm) (as we saw indications of possible respiration contamination from these), and finally (3) the measured respiration rate should not exceed the total respiration rate of the system (measured by T. Dalsgaard unpublished).

# Long Term BDOC Experiments

The quantity of DOC available for bacterial degradation (BDOC) was measured in long term (126–148 d) oxygen consumption experiments. A total of 36 incubations were established from the focus depths and rivers (**Table S1**). 2 L were 0.22 µm filtered (Millipore <sup>R</sup> Sterivex) into acid washed HDPE bottles (the filters were later used for extraction of nucleic acids). Note we thus did not measure the bioavailability of the particulate organic matter. The bacterial inoculum was prepared by GF/F filtering 100 mL into 2 × 50 mL falcon tubes. The 0.22 µmfiltered water and bacterial inoculum was stored cold (2◦C) until experimental set-up. The oxygen consumption experiment was set up with five replicate 65 mL Winkler glass bottles equipped with an optical oxygen sensor for each sample and incubated in dark at 8◦C for 148 days (samples collected in July and August) or 126 days (samples collected September and October). Prior to incubation NO<sup>−</sup> 3 and PO3<sup>−</sup> 4 (final conc. 5 µM and 1 µM, respectively) were added along with the bacterial inoculum (10% vol.) to ensure that N or P was not limiting C-degradation during incubation. Consequently, these measurements do not reflect in situ conditions neither the bacterial communities at the time of sampling, but are rather quantitative measure of the bioavailable DOC pool. All Winkler bottles contained a magnet to ensure mixing, and were incubated in a water bath in order to minimize oxygen contamination. In addition, parallel bottles containing 100 and 0% air saturated seawater were measured to correct oxygen measurements for deviations in the 100% control. As control incubations, triplicates of sample water were incubated without bacterial inoculum as well as a sample with 20 µL HgCl. The change in % air saturation over time was monitored every 7–14 days using a Fibox 3 fiber optic patch oxygen sensor (Presense) calibrated with 100% and 0% air saturation using the program OxyView– PST3-V6.02. As for the BGE measurements the change in oxygen concentration over time was converted to carbon consumption assuming an RQ of 0.82 (Søndergaard and Middelboe, 1995). DOC concentration was measured initially and by the end of the incubation.

# Nucleic Acids Extraction, Amplification, and Amplicon Sequencing

All environmental samples for molecular analysis were collected by filtering water onto 0.22 µm pore size Millipore <sup>R</sup> Sterivex filters (the filtrate was used for bioavailability measurements described above). Note that samples were not prefiltered i.e., also the particle associated bacteria are included in this analysis. The filters were immediately frozen and stored at –80◦C until nucleic acid extraction. DNA and RNA were extracted simultaneously using the AllPrep DNA/RNA Mini Kit (Qiagen, Hilden, Germany) according to manufacturer's instructions with modifications for extraction from Sterivex filters as in Paulsen et al. (2016). RNA was subsequently treated with the DNAfree DNA Removal kit (Invitrogen, CA, USA) and reverse transcribed using the SuperScript III First-Strand Synthesis System for RT-PCR (Invitrogen). Amplification of cDNA and DNA was performed using a two-step nested PCR approach with primers 519F (CAGCMGCCGCGGTAA; Øvreås et al. (1997) and 806R (GGACTACHVGGGTWTCTAAT; Caporaso et al. (2011) targeting the bacterial (and Archaeal) 16S rRNA gene V4 hypervariable region. For the first PCR step triplicate samples were amplified in reaction volumes of 20 µL including 10 ng DNA or cDNA, 10 µL HotStarTaq Master Mix (Qiagen), 500 nM of each primer and nuclease free water. PCR cycles consisted of an initial denaturation of 15 min at 95◦C, followed by 25 cycles of 95◦C for 20 s, 55◦C for 30 s and 72◦C for 30 s and a final extension step of 72◦C for 7 min. Triplicate PCR products were pooled, purified using the DNA Clean & Concentrator-5 kit (Zymo Research Corporation, CA, USA) and quantified using the Qubit 3.0 Fluorometer. For the second PCR step, 10 ng of pooled PCR product was used in a reaction mixture containing 25 µL HotStarTaq Master Mix, 500 nM of each nested primer with an unique eightnucleotide barcode (total of 96 combinations) and nucleasefree water to bring the mixture to the total volume of 50 µL. Thermal cycles had an initial denaturation for 15 min at 95◦C, followed by 15 cycles at 95◦C for 20 s, 62◦C for 30 s, 72◦C for 30 s, and a final extension step of 72◦C for 7 min. PCR products were purified using Agencourt AMPure XP Beads (Beckman Coulter Inc., CA, USA) and prepared for sequencing by pooling the amplicons in equimolar amounts. The quality and concentration of the amplicon pool were assessed by agarose gel electrophoresis and by using a Qubit 3.0 Fluorometer, before sending to the Norwegian Sequencing Centre (Oslo, Norway) for High-Throughput Sequencing on a MiSeq platform (Illumina, CA, USA) using the MiSeq Reagent Kit v2 (Illumina). Sequencing data is available at "The European Bioinformatics Institute" under study accession number PRJEB16067 (http://www.ebi.ac.uk).

#### 16S rRNA Gene Sequence Analysis

Paired-end sequences were processed using different bioinformatic tools incorporated on a qiime-processing platform (Caporaso et al., 2010). FASTQ files were quality end-trimmed at a phred quality score ≥24 using Trimmomatic (Bolger et al., 2014) and merged using PANDAseq (Masella et al., 2012), while all reads <200 bp were removed. Prokaryotic OTUs were selected at a sequence similarity threshold of 97% using a de novo uclust (Edgar, 2010) OTU clustering method with default parameters and taxonomy assigned using the Silva 111 reference database (Quast et al., 2013). OTUs with a taxonomic identification were assembled to an OTU table providing abundances for each sample excluding singletons and rarefied to the number of sequences of the smallest samples (5,000 sequences). A total of 4,096,371 sequences were retrieved from the Illumina sequencing of the 16S rRNA gene V4 hypervariable region from total RNA across 52 samples. After removal of singletons, unassigned OTUs and chloroplast reads, sequences were rarefied to 5,000 reads per sample, with a total of 15,922 unique OTUs at 97% sequence identity. Multivariate statistical analysis was performed on basis of the rarefied OTU matrix to explain variations in the data and test for multivariate environmental correlation with the prokaryotic community structure. Bray– Curtis resemblance, ANOSIM, principal component analysis and redundancy analysis were calculated using primer-e version 6 (Plymouth, UK) and Canoco 5 (Ter Braak and Šmilauer, 2012).

# Source Tracker Analysis

To illustrate the spread of bacterial communities in the fjord the SourceTracker 0.9.5 software (Knights et al., 2011) was applied in QIIME. It is designed to track the relative contribution of predefined microbial sources in sink samples using a Bayesian approach, as done in Storesund et al., in review. The relative abundance of 16S rRNA genes of all focus samples was used as input data. The OTU table, comprising all OTUs with a taxonomic identification (excluding singletons and chloroplast reads) was rarefied to 1,000 sequences and filtered to only include OTUs that were abundant in more than 3 samples. The rivers (R1, R2 and R3) and as a proxy for inflowing coastal water St. 4 DCM samples were used as "source populations." All remaining stations during the sampling period from July until October were defined as the "sink samples." The result is given as % likely origin from the 4 defined sources. The remaining are categorized as "unknown source". The result were visualized as fjord transects with weightedaverage extrapolation between points using Ocean Data View (Schlitzer, 2016).

#### Paulsen et al. C-Bioavailability in an Arctic Fjord

# RESULTS

#### Hydrography

During the first period (July-August) the fjord was stratified by a strong halocline at 5–6 m depth with low salinity (<20) in the surface and more saline bottom water (>30). The stratification was strongest in the inner fjord and the salinity of the surface water increased eastwards toward the shelf (Figure S1). The later period (September-October) was in general colder (surface water <2 ◦C) and frequent storms mixed the upper layer, which deepened to 30 m at the two inner stations and down to 50 and 80 m at St. 3 and 4, respectively (**Figure S1**). Five water types were defined for this study based on the thermohaline properties (**Table 1** and **Figure 2**). This was done to facilitate the interpretation of the results as the dominant carbon sources can be expected to differ across these water types. The "Shelf water" represented the mixing of waters from the East Greenland Current (T∼1.5◦C, S < 34) with warmer and more saline Atlantic water (S > 34.4). "Surface water" is influenced by runoff and the seasonal surface heating, while "Subsurface fjord" water represents the shelf water that has entered the fjord passing over the sill and gradually being mixed with the "Surface water." We defined the "River plume waters" to be the fjord surface waters that were under direct influence of the river discharge. The change between the two periods is apparent in the T-S plot (**Figure 2**). The runoff was strongest in July and during the last river sampling on September 10 the flow from the glacial rivers (R1 and R2) had almost terminated, while Zackenberg river (R3) was still flowing until end-September. The inner stations were strongly affected by the river silt, with a high turbidity of surface water and the photic zone was therefore initially shallow at St. 1 and 2 (<5 and 10 m, respectively), however later in September it deepened to 25–35 m as turbidity decreased. The opposite trend was observed at the outermost station where the photic depth decreased from 35 to 10 m from July to October, due to the decreasing irradiation.

# Bacterial Abundance and Chlorophyll a

While the central fjord was still ice-covered (**Figure 1**), the inner part was ice-free and a reduction in nutrients measured below the ice at St. 3; NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 , PO3<sup>−</sup> 4 , SiO4, in the surface (0.01, 0.33, 1 µM) compared to 100 m values (3.2, 0.7, 6.4 µM) indicated that a phytoplankton spring bloom had already initiated in the inner fjord prior to current sampling program. Chl a was highest (up to 3.1 µg L−<sup>1</sup> ) in the outer region of the fjord in July and August with a deep maximum at 20–40 m (**Figure 3D**), while in the late period chl a was highest in the inner fjord (**Figure 3A**). The average bacterial abundance (BA) in the surface water was 3.2 ± 1.0 × 10<sup>5</sup> mL−<sup>1</sup> and generally peaked at 0–20 m in the first period, with a maximum abundance in the outer fjord. BA was significantly higher in the upper 20 m for 19 out of 25 profiles, thus the BA maxima were decoupled from the chl a, especially in the first period where the chl a max was deep (20–40 m) (**Figure 3**). Overall only a weak significant linear correlation was found between BA and chl a (r <sup>2</sup> = 0.03, p = 0.03, n = 164). At the innermost station BA did not correlate with chl a at any time, but there was a positive correlation between BA and turbidity (r <sup>2</sup> = 0.757, p < 0.01, n = 31). At the outermost station chl a correlated with BA during the study (r <sup>2</sup> = 0.44, p < 0.01). At the two mid-fjord stations (2 and 3), correlation between BA and chl a was found only after the runoff had ceased (St. 2: r 2 = 0.83, p < 0.01 and St. 3: r <sup>2</sup> = 0.64, p < 0.01). BA in the rivers was significantly lower than in the fjord surface waters (**Table 3**). Abundances ranged from a minimum of 1.3 × 10<sup>5</sup> in the glacial rivers (R1 and R2) to 6.4 × 10<sup>5</sup> cells mL−<sup>1</sup> in R3 (**Table 2**). Note that only the free-living bacteria were enumerated.

# Bacterial Production and Particle Association

Bacterial production (BP) was highest in the beginning of the sampling period (17–21 July) with a maximum of 2 µg C L−<sup>1</sup> d −1 in the river plume water (St.1 + 2, 1 m) (**Table 2**, **Figures 4A,B**). The third highest measure of BP (1.7 µg C L−<sup>1</sup> d −1 ) was found in the fresh water layer just below the ice sampled on July 11. BP was lowest within the Shelf water at St. 4 where it remained below 0.09 µg C L−<sup>1</sup> d −1 throughout the entire study period, despite BA being relatively high. BP in the rivers was on average 0.39 ± 0.07 µg C L−<sup>1</sup> d −1 and 0.32 ± 0.06 µg C L−<sup>1</sup> d −1 in the first and the second period, respectively, and thus generally higher than those measured in the fjord despite low BA (**Table 3**). We found a larger fraction of BP in the R3 to be free-living (**Table 2**), which is possibly due to the presence of lakes acting as sedimentation basins. The contribution of particle-associated (>3 µm) bacterial production was considerable in both the fjord and rivers. In 17 out of 25 cases particle association was higher in the DCM sample than in the associated 1 m sample, however there was no significant difference between the %particle associations at the two depths (**Figure 4**).

# Bacterial Carbon Demand and Primary Production

Only two of the BR measurements fulfilled all criteria for solid measurements of oxygen consumption throughout the incubation. These gave growth efficiency values of 7.3 ± 1.0 and 6.4 ± 2.0%, during incubation at the in situ temperature of −1.1◦C (St. 3, July) and at 3◦C (St. 4), respectively. The average BGE of 6.9% was applied for determination of bacterial carbon demand (BCD) to allow comparison with the total amount of carbon fixed by planktonic primary production (PP). When integrated over the photic zone the estimated BCD:PP was on average 1.7 ± 1.2 across the sampling period, suggesting that bacterial carbon demand could not be sustained by the local phytoplankton production. There were no clear spatial or temporal trends in the BCD:PP ratio (**Figure 5**), and estimates were similar when integrated to 100 m. PP was highest initially and could support a high BCD at St. 1 and 3, while bacteria were not sustained by fresh PP initially at St. 2 and 4 in this period. Toward the end of the open water period different patterns developed as bacteria in the inner fjord could be sustained by fresh PP, while BCD was decoupled (up to 5 times higher) from PP at the outermost station (**Figure 5**). The dissolved fraction (<0.7 µm) of PP was in general high in the fjord and contributed 39–52% in the first period and less, 27-36%, in the late period.

The largest fraction contributed most at the outer station (max 32%) and in the last period it was never higher than 13%, whereas the fraction 0.7–10 µm became dominant in the late period (**Figure 5**).

# Organic Matter Concentrations and C:N Ratios

DOC concentrations showed little systematic vertical variability except for the surface samples at 1 m, which had significantly lower concentrations in 19 out of 25 profiles (**Figure S2**). On average, the DOC concentration decreased from the outer part to the inner fjord from 130 ± 16 µM at St. 4 to 106 ± 30 µM at St. 1 (**Figure 6**). From the first to the second sampling period where the mixed layer deepened, the DOC concentration at 1 m increased, while at the DCM there was a slight decrease. The DOC concentration was highest in the deep water types characterized as Subsurface fjord water (101 ± 20 µM, n = 49) and slightly lower in the Surface water (97 ± 27 µM, n = 97). The River water had significantly lower DOC (40 ± 13 µM, n = 24), and thus the River plume water was diluted to an averaged concentration of


Where there is no SD, only one sample was available.

TABLE 3 | Water type characteristics and properties shown as average ± SD of the 5 water types for each of the two periods.


Note Shelf water was only found at 100 m depth. Where there is no SD, only one sample was available.

67 ± 13 µM, n = 7. DOC:DON was also lowest in the River water (8 ± 2) and the River plume water (11 ± 2), while the Shelf water had a significantly higher C:N ratio (21 ± 8) (**Table 3**).

Maximum POC concentration was found in the rivers with up to 14 µM, while in the fjord a maximum of 4 µM POC was found at the DCM. In the fjord the particulate fraction (%) of the total organic matter was in general minor (avg. 2.3 ± 1, max = 5%),

(DCM) from July to October at the 4 stations.

while in the rivers the contribution of POC was significant (avg. 13 ± 8, max = 35%) (**Tables 2, 3**). This gave a significant negative relationship between salinity and POC (r <sup>2</sup> = 0.3, p < 0.001, n = 86). POC did not correlate with chl a, turbidity or to particulate bacterial production. The C:N ratio of the particulate matter was significantly lower in the rivers than in the fjord (p = 0.047, Ftest). In the fjord, the C:N ratio increased from the first period to the end of the open water season (p = 0.0023, F-test), whereas no trend was observed in the rivers.

Inorganic nitrogen (NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 ) was the limiting inorganic nutrient for primary production and was reduced to 0.4 µM in the surface water during the entire period, while the background level in the Shelf water was ca. 6 µM (**Table 3**, **Figure S2**). Ammonium (NH<sup>+</sup> 4 ) was only measured at St. 3 and ranged between 0.05 and 0.4 µM, with a maximum at 40 m (below DCM).

#### Bioavailability of DOM

Despite the increase in average DOC concentration from St. 1 to 4, there was a slight decrease in the concentration of BDOC from 19 ± 10 to 11 ± 8 µM from St. 1 and 4, respectively (when values were averaged for 1 m and DCM over the entire period). The fraction of BDOC relative to total DOC (%BDOC) thus decreased from the inner to the outer part of the fjord (**Figure 6**).

BDOC concentrations in the rivers were relatively high (avg. 18 ± 8 µM), resulting in significantly higher %BDOC in the rivers than in the fjord (p < 0.005, t-test) (**Figure 6**, **Table 3**). Averaging %BDOC within each water type revealed a decrease as river water was mixed with the fjord water, i.e., River water : River plume : Surface water : Subsurface water (**Table 3**). There was a significant negative correlation between the relative abundance of C to N (DOC:DON) and %BDOC (r <sup>2</sup> = 0.17, p = 0.08, n = 42). C:N ratios were generally higher in the last period and the highest C:N of 29 ± 3 was found in the Shelf water.

#### Bacterial Community Composition

In general, Proteobacteria were the most abundant bacteria phylum in both fjord and river samples (∼86% of the bacterial phylum). Differences in community composition were observed at class level, with fjord DCM samples containing more Alphaproteobacteria than the rivers and the Subsurface

water being strongly dominated by Gammaproteopbacteria (80%), opposed to surface samples and rivers samples with a higher share of Betaproteobacteria (**Table 3**). Cyanobacteria were present in all rivers and constituted up to 2.7 ± 1.7% of the microbial community in the innermost stations. Alpha diversity (described by the Shannon index) ranged from 4.9 to 9 (**Table S2**) indicating a greater diversity in the rivers and lowest in fjord samples from July and August (**Figure 7**). A redundancy analysis (including spatial and temporal parameters) was performed to identify factors that significantly affect bacterial community composition. This explained 47.2% of the total variation in the OTU data, with "water type" (19.5%; river (13%), p = 0.002) and time of sampling expressed as "month" (10.8%; September (8.5%), p = 0.002) being the most significant variables. As "station" generally explained <5% this variable was not included. Samples from all three rivers with overall higher species richness clustered together. The majority of the September fjord samples clustered tightly together, independent of their respective water types and differences between surface and DCM were minimal. Samples from July and August however clustered according to sample depth (i.e., surface or DCM) (**Figure 7**). Explanatory variables (water type, month and station) explain in total less than 50% of the variation in diversity for the entire data set.

To elucidate changes in community composition throughout the fjord the complexity of the dataset was reduced by constraining a phylogenetic analysis to only include the most abundant bacterial taxa (relative abundance >1%). Further the temporal factor was reduced by restricting the analysis to cover a 10-day period in early August (**Figure 8**). The heat map shows that all three rivers were very different from the fjord community, and that each river had unique taxa as the most abundant (blue = high, red = low relative abundance). While bacteria found in R1 (with the closest connection to the Greenland Ice Sheet) all could be found in the two other rivers, the R3 (that runs through vegetation-covered catchment area) included some unique families e.g., Granulosicoccaceae, Alcaligenaceae, and an unknown family from the order vadinha64 (the latter presented as "uncultured\_bacterium" from the class Opitutae in **Figure 8**). In early August the majority of the most abundant river taxa were absent in the fjord samples, with the exception of R1 that shared a great number of its most abundant taxa with the nearby Station 1 surface sample which had more unique taxa than any other station (**Figures 1**, **8**). Stations 2, 3, and 4 showed a higher number of shared taxa indicating a gradient from the inner to the outer fjord. The opposite was observed for the deeper DCM samples, which showed a gradual decrease of certain taxa from the outer St. 4 toward the innermost St. 1.

The SourceTracker analysis (including the full dataset at OTU level) showed clear contribution of especially the R1 community to the surface fjord stations with maxima of 75, 63, and 66% similarity at St. 1, 3, and 4, respectively in July-August and ca. 40% in September (**Figure 9**). The coastal community (source set as St. 4 DCM) entering the outer part of the fjord dominated strongly in the deeper fjord samples throughout the entire study (**Figure 9** and **Table S1**). However, the coastal communities also dominated at the surface waters of St. 2 in the first period, where R1 only contributed 15%. As the R1 community was also present in R2 and R3, overall river contribution is incorporated in the R1 plot (**Figures 9A,B**). The community unique to R2 and R3 only contributed to minor degree (7–15%) to the fjord surface community.

In order to investigate the influence of environmental parameters exerted on the bacterial community structure, we performed canonical correspondence analysis (CCA) between numerical factors, such as bacterial production, chl a fluorescence, temperature, salinity, DOC, and %BDOC, (**Figure S2**). Due to the complexity of the dataset no strong correlations were found with this analysis. However, when single taxa were correlated to specific environmental parameters we found that especially genera from the class Gammaproteobacteria showed correlations with specific DOM characteristics e.g., a strong positive correlation was found between the genus Glaciecola (order: Alteromonadales) and bacterial production (r <sup>2</sup> = 0.5283; p < 0.0001) with a maximum relative abundance of 25% when BP was highest. Another unknown genus from the order Alteromonadales showed a positive correlation with %BDOC (r <sup>2</sup> = 0.2720; p < 0.0075).

#### DISCUSSION

#### Bioavailability of Allochthonous DOM Sources

River input in marine systems are often a source of high DOM concentrations and therefore DOM often correlate negatively to salinity (Cauwet, 2002). The Young Sound system however deviates from this trend due to two factors. Firstly the freshwater input in fjord has low DOC concentrations due to the dominance of glacial meltwater and limited catchment vegetation. Secondly, we hypothesize that the coastal shelf waters entering the fjord are characterized by high levels of terrestrial organic matter that originates predominantly from Siberian rivers (we did however not sample sufficiently deep at St. 4 to capture the pure Polar water). These rivers discharge into the Arctic Ocean (Amon and Budéus, 2003) and the terrestrial DOC is retained in the surface waters exiting the Arctic Ocean via the Fram Strait as part of the East Greenland Current (Granskog et al., 2012). A large fraction of the terrestrially derived DOM transported by the major rivers Ob and Yenisei is found to be refractory (Meon and Amon, 2004), which explain the low bacterial activity and low DOC bioavailability we found in the Shelf water despite high concentrations. These conditions therefore result in a positive correlation between DOC and salinity (r = 0.51, p < 0.001, df = 175) in Young Sound.

In addition to the quantitative difference, the qualitative C:N ratio of the two allochthonous DOM sources differ. The ratio between bioavailable DOC and inorganic nitrogen exceeded the Redfield ratio (C:N:P = 106:16:1) by thousand fold, whereas the BDOC:PO3<sup>−</sup> 4 ratio was 45 ± 30. This emphasizes the importance of DON as the main source of nitrogen for bacteria. The low C:N ratios in the river water resemble those reported from Alaskan glacial rivers, where the relatively high source of DON is explained by microbial production of protein-rich DOM in the subglacial environment (Hood and Scott, 2008). The concentration of bioavailable DOM in the rivers entering Young Sound was slightly lower than values obtained in Alaskan glacier outflow (Hood et al., 2009), but very similar to other studies from the Greenland ice sheet meltwater (Lawson et al., 2014). In contrast to the river DOC, the allochthonous DOC entering the fjord from the open ocean had high C:N ratios, similar to Arctic surface water (Benner et al., 2005). Together, our results demonstrate that open ocean DOC is less labile than the DOC produced in the fjord and that supplied from the rivers.

As expected the load of particulate organic carbon was relatively high in the rivers (Hood et al., 2015). However a high POC-signal was not traceable in the surface water of the inner fjord stations as was the case e.g., for the silt particles (measured as turbidity) and silicate. This indicates that the POC has been lost from water column either by sedimentation, dissolution or bacterial degradation. Bioavailability of POC was not determined, and BDOC values thus potentially represent an underestimation of labile organic carbon concentration. However, since POC on average accounted for 2.3 ± 1.0% and 13 ± 8% of total DOC in the fjord and rivers, respectively, the contribution of bioavailable POC to total bioavailable carbon is probably relatively small. The spatial gradient in %BDOC along the fjord transect suggests a gradual consumption of the labile DOC entering the fjord via the rivers. Consumption of riverine DOC in the fjord is supported by the high BP in the river and river plume water (**Table 3**) and the negative correlation of BP to salinity. The differences in DOM concentration and composition has been suggested as a driver for diversification of bacterial communities (Crump and Hobbie, 2005; Blanchet et al., 2016; Roiha et al., 2016). Our study


suggests that the labile character of the river-DOM may have a role stimulating and shaping the activity and structure of bacterial communities in the fjord, by favoring fast growing bacteria. In samples where the community structure was analyzed, high BP was negatively correlated to salinity (r <sup>2</sup> = 0.2743; p < 0.0103). Further, high bacterial production was associated with growth of specific taxa, such as a positive correlation between BP and the relative abundance of the genus Glaciecola.

### Bacterial Community Composition

In general, our results suggest that the bacterial community is largely structured by the different water sources to the fjord and changes along the salinity gradient, as also found in other coastal environments (del Giorgio and Bouvier, 2002; Crump and Hobbie, 2005; Gutiérrez et al., 2015). Especially in the early period large differences were observed between 1 m and DCM bacterial communities, which were explained by the

between points using Ocean Data View (Schlitzer, 2016).

strong stratification at 5–7 m in this period (**Figure S1**). This resulted in lower species richness during stratification in July and August (**Figure 8**), possibly because bacterial communities exhibit environmental niche partitioning when the water masses remain separated in this period (Morris et al., 2005; Delong et al., 2006; Chow et al., 2013; Salter et al., 2014). As river runoff decreased and mixing events increased (September) the resemblance between communities at 1 m and DCM naturally increased. One specific strong storm that lasted for several days (21–26 September) may explain the high similarity of 1 m surface samples from St. 1 and 2 to the river samples (**Figure 8**), as samples were collected immediately after the storm where the disturbance apparently caused transport of terrestrial bacteria to the fjord (Crump and Hobbie, 2005).

In temperate and Polar Regions Flavobacteriia often dominate during phytoplankton bloom (Wilson et al., 2017). We also found this group to dominate when chl a concentrations were high (i.e., DCM samples at St. 4), while they contributed only marginally (<0.1%) to the total bacterial community in the remaining fjord and particularly little in the rivers (**Figure 8**). Cyanobacteria were relatively abundant in the glacial runoff and can be considered as freshwater tracer, since marine cyanobacteria are usually not found in these Arctic regions (Paulsen et al., 2016). The bacterial communities in the rivers were highly specific to each of them (**Figure 8**), which we suggest is due to their difference in origin and catchment area; close connection (0.5 km) to the Greenland Ice Sheet (R1), longer distance (2 km) from smaller local glacier (R2), and the lowland vegetation rich river with lake connection (R3). This is in agreement with a recent study from West Greenland that similarly show distinct communities in rivers and a proglacial lake over a 2 km distance (Hauptmann et al., 2016), and find less terrestrial species in river samples with a more direct connection to glaciers.

In the present study R1 has the closest connection to the glacier and all the bacterial families found in R1 were also found in the two other rivers. The SourceTracker analysis (**Figure 9**) revealed that the R1 community had the far most influence on the surface water, despite not being the largest river. The species unique to R3 hardly contributed to the fjord community, despite being the largest river in terms of volume. This suggests that there is a higher potential for the glacial bacterial community to persist in the fjord than terrestrial communities. Gutiérrez et al. (2015) also found persistence of specific bacterial communities in the surface water of a glacier influenced fjord in Patagonia, and suggested that the glacial meltwater community in the surface was maintained by the competitive advantage of tolerating the cold fresh conditions. Given the averaged doubling time of bacteria in the surface water of Young Sound of 11 ± 7 days (fastest doubling time of 0.9 days) and a transport time of 14–30 days from innermost to the mouth of the fjord, the persistence of the specific communities may be a result of fast transport within the surface water and limited mixing. The differentiation between surface and subsurface communities was lost when the thermal stratification broke down in present study and in Gutiérrez et al. (2015).

#### Annual Estimations of Carbon Production and Turnover in Young Sound

Due to the strong stratification, high turbidity and input of allochthonous carbon sources, relatively heterotrophic conditions were expected to prevail in the fjord system as concluded in Nielsen et al. (2007). The seasonal and spatial resolution of bacterial carbon demand and primary production allowed a rough, but more robust than previous estimates of the annual carbon budget in Young Sound (Rysgaard and Nielsen, 2006; Nielsen et al., 2007), as previous studies cover a shorter period of the productive season and use a literature BGE of 33%. However, the present study did not cover the early productive season during ice cover in June. As an attempt to account for that we extrapolate the carbon uptake and primary production during the ice-covered productive period using the values obtained on July 11 at St. 3 during ice-cover and by assuming a productive period of 125 days. Based on these assumptions, the annual bacterial production amounted to 1.3 g C m−<sup>2</sup> year−<sup>1</sup> , corresponding to a carbon demand of 18 g C m−<sup>2</sup> year−<sup>1</sup> in the upper 100 m. This BP value is ∼3 times lower than a previous estimate of net annual bacterial production in Young Sound of 4.2 g C m−<sup>2</sup> year−<sup>1</sup> , based on measurements conducted during ice-cover at St. 3 (Rysgaard and Nielsen, 2006), emphasizing that the conditions at St. 3 in June are not representative for the late open water period (August-October) or for the entire Fjord system.

The assessment of bacterial carbon demand rely greatly on estimates of BGE, which is known to depend on a number of factors including DOC composition and lability, bacterial community and temperature, thus it is important to emphasize that BGE is likely to vary across time and space in Young Sound. The BGE we found (6.9%) is low, but still in line with other Arctic studies ranging from 2.2 ± 2.1% in the Chucki Sea (Cota et al., 1996), 6.3 ± 3% in the Fram Strait (Kritzberg et al., 2010), 19.1 ± 9.5% in the rivers Ob, Yenisei and the adjacent Kara Sea (Meon and Amon, 2004) and 9.5 ± 8.7% in Kobbefjord, Greenland (Middelboe et al., 2012).

The bacterial carbon demand on average exceeded primary production during the study period (**Figure 5**) and indicated that only about 30% of the PP is dissolved and thus potentially available for bacterial uptake. Further, protist and copepod grazing of phytoplankton consumed a substantial fraction of the particulate PP (Middelbo et al. submitted; Arendt et al., 2016). Even though some of the particulate primary production are eventually degraded by pelagic bacteria as detrital matter, our results strongly indicated that primary production could not sustain bacterial carbon demand in the fjord within the productive 125 days and much less on an annual basis. Consequently, the allochthonous DOM sources contributes significantly to the bacterial carbon turnover in Young Sound.

The current study presents an overview of the organic carbon sources and their turnover in the high Arctic Young Sound. Further it highlights that the meltwater associated bacterial community from the glacial rivers persists and is actively transforming the river-DOM within the freshwater lens. The calculations of annual carbon production and turnover are obviously associated with large uncertainties regarding extrapolations across time and space and use of factors for converting thymidine incorporation to carbon production and net bacterial production to carbon demand. However, by applying a high temporal and spatial resolution of sampling and on-site estimation of BGE, this study provides a relatively solid estimation of the annual carbon budget in a high Arctic Fjord compared to previous studies. The fjord is net heterotrophic and in future scenarios with increasing temperatures the relative contribution of riverine DOC is expected to increase, driving the system toward more heterotrophic conditions. However, more measurements of bacterial growth efficiency and factors controlling this are required to provide more solid budget estimates of bacterial carbon consumption on an ecosystem scale.

# AUTHOR CONTRIBUTIONS

MP and SN led the collection and analysis of data and the writing of the paper with equal contribution. MM led the planning of the study, and contributed with sample collection, data analysis and contributing greatly to the writing of the paper. OM led the collection, laboratory work and analysis of the molecular data. SM led the planning and calculations of primary production. AD led the collection and measurements of POC and PON. All other authors contributed to writing the paper. In addition EM, AL, TJ, and MS helped collecting data and CS helped analysing the organic matter samples.

#### FUNDING

This study was funded by research grants from the Danish Ministry of the Environment (DANCEA), the project MicroPolar (RCN 225956) funded by the Norwegian Research Council, Carlsberg Foundation and the Arctic Research Centre at Aarhus University. CS was funded by the Danish Research Council for Independent Research (DFF 1323–00336). The Arctic Science Partnership and the Greenland Ecosystem Monitoring program facilitated this work.

#### ACKNOWLEDGMENTS

Big thanks to Egon Frandsen, Kunuk Lennert, and Ivali Lennert for organization, expertise in the area and technical assistance in the field and to Johan the polar fox for entertainment. We are

#### REFERENCES


grateful to be able to include data from the Greenland Ecosystem Monitoring Programme were provided by the Department of Bioscience, Aarhus University, Denmark in collaboration with Department of Geosciences and Natural Resource Management, Copenhagen University, Denmark.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2017.00176/full#supplementary-material

Figure S1 | Profiles of salinity, temperature (◦C), DOC (µM), DOC:DON and the concentration of nutrients NO<sup>−</sup> 2 <sup>+</sup>NO<sup>−</sup> 3 and Si (µM) at the 4 stations in the periods from July-August (red colors) and September-October (blue colors).

Figure S2 | Canonical correspondence analysis (CCA) for sequencing data of 16S rRNA gene fragments at RNA level (most abundant OTUs >1%). Arrows indicate selected environmental variables: bacterial production, dissolved organic carbon (DOC), percent bioavailable dissolved organic carbon (%BDOC), fluorescence, temperature and salinity. The OTUs selected for further analysis are labeled with their genus classification.

Figure S3 | Time-series measurements from Zackenberg River of temperature ( ◦C), Sediment (mg L−<sup>1</sup> ), LOI (mg L−<sup>1</sup> ) (LOI is organic matter determined by Loss On Ignition when heated at 105◦<sup>C</sup> <sup>≈</sup> particulate organic matter), DOC (µM). Data is collected by Greenland Ecosystem Monitoring, the GEOBASIS Programme.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Paulsen, Nielsen, Müller, Møller, Stedmon, Juul-Pedersen, Markager, Sejr, Delgado Huertas, Larsen and Middelboe. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Upstream Freshwater and Terrestrial Sources Are Differentially Reflected in the Bacterial Community Structure along a Small Arctic River and Its Estuary

Aviaja L. Hauptmann1, 2, <sup>3</sup> , Thor N. Markussen<sup>3</sup> , Marek Stibal 3, 4, Nikoline S. Olsen<sup>3</sup> , Bo Elberling<sup>3</sup> , Jacob Bælum<sup>5</sup> , Thomas Sicheritz-Pontén<sup>2</sup> and Carsten S. Jacobsen3, 6 \*

<sup>1</sup> Center for Biosustainability, Technical University of Denmark, Hoersholm, Denmark, <sup>2</sup> DTU Bioinformatics, Technical University of Denmark, Kgs. Lyngby, Denmark, <sup>3</sup> Center for Permafrost, University of Copenhagen, Copenhagen, Denmark, <sup>4</sup> Department of Ecology, Faculty of Science, Charles University, Prague, Czech Republic, <sup>5</sup> Chr. Hansen A/S, Hoersholm, Denmark, <sup>6</sup> Department of Environmental Science, Aarhus University, Roskilde, Denmark

#### Edited by:

Eva Ortega-Retuerta, Spanish National Research Council, Spain

#### Reviewed by:

Ruben Sommaruga, University of Innsbruck, Austria Clara Ruiz Gonzalez, Institut de Ciències del Mar (CSIC), Spain

> \*Correspondence: Carsten S. Jacobsen csj@envs.au.dk

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Microbiology

Received: 07 June 2016 Accepted: 05 September 2016 Published: 21 September 2016

#### Citation:

Hauptmann AL, Markussen TN, Stibal M, Olsen NS, Elberling B, Bælum J, Sicheritz-Pontén T and Jacobsen CS (2016) Upstream Freshwater and Terrestrial Sources Are Differentially Reflected in the Bacterial Community Structure along a Small Arctic River and Its Estuary. Front. Microbiol. 7:1474. doi: 10.3389/fmicb.2016.01474 Glacier melting and altered precipitation patterns influence Arctic freshwater and coastal ecosystems. Arctic rivers are central to Arctic water ecosystems by linking glacier meltwaters and precipitation with the ocean through transport of particulate matter and microorganisms. However, the impact of different water sources on the microbial communities in Arctic rivers and estuaries remains unknown. In this study we used 16S rRNA gene amplicon sequencing to assess a small river and its estuary on the Disko Island, West Greenland (69◦N). Samples were taken in August when there is maximum precipitation and temperatures are high in the Disko Bay area. We describe the bacterial community through a river into the estuary, including communities originating in a glacier and a proglacial lake. Our results show that water from the glacier and lake transports distinct communities into the river in terms of diversity and community composition. Bacteria of terrestrial origin were among the dominating OTUs in the main river, while the glacier and lake supplied the river with water containing fewer terrestrial organisms. Also, more psychrophilic taxa were found in the community supplied by the lake. At the river mouth, the presence of dominant bacterial taxa from the lake and glacier was unnoticeable, but these taxa increased their abundances again further into the estuary. On average 23% of the estuary community consisted of indicator OTUs from different sites along the river. Environmental variables showed only weak correlations with community composition, suggesting that hydrology largely influences the observed patterns.

Keywords: biodiversity, bacterial community, freshwater network, Greenland, arctic, polar environments

# INTRODUCTION

Arctic river and estuary ecosystems are vulnerable to the ongoing climate change. Increasing temperatures are resulting in negative mass balance of glaciers and increased precipitation, with significant impacts on rivers and estuarine systems (Serreze et al., 2000; Mueller et al., 2003). In addition, Arctic rivers are known to transport significant amounts of organic carbon and biomass from permafrost and glacier ecosystems into the Arctic oceans and are therefore important factors in global climate change models (Kling et al., 1991; Guo et al., 2007; Lawson et al., 2014; Hawkings et al., 2015). Higher river flow associated with the warming climate may result in a more river-dominated community in the estuaries (Fortunato et al., 2013). Once riverine bacteria reach the estuary, they may influence local nutrient cycling through biofilm formation and forming aggregates (flocs; Decho, 2000). Thus, bacterial communities dispersed through Arctic riverine systems may be important for biogeochemical cycling processes in Arctic estuarine and coastal ecosystems.

There are a number of studies on Arctic estuarine ecosystems focused on biodiversity, biological productivity, seasonal variability, food web interactions, and responses to environmental variables (Galand et al., 2006, 2008; Wells et al., 2006; Vallieres et al., 2008; Fortunato et al., 2012, 2013). However, the river communities have usually been assessed as a whole. How different communities added to the river affect the estuarine community has not been addressed to date. While previous studies have shown that increased river flow alters Arctic river and estuary communities in seasonal patterns (Crump et al., 2009; Fortunato et al., 2012, 2013), it is yet unknown how upstream sources of freshwater microbial communities influence these communities.

A few recent studies on freshwater ecosystems at different spatial scales have greatly increased our understanding of the biogeography of riverine networks (Nelson et al., 2009; Crump et al., 2012, 2007; Ruiz-González et al., 2015; Niño-García et al., 2016). These studies have shown that biogeographic patterns of bacterioplankton communities are a result of the interaction between local environmental variables and masseffects. Furthermore, that mass-effects are determined by the hydrology as well as the position along the network (Crump et al., 2007, 2012; Nelson et al., 2009; Ruiz-González et al., 2015; Niño-García et al., 2016).

Water residence time (WRT) has shown to be an important factor for determining the relative influence of hydrology vs. local sorting (Niño-García et al., 2016). There seems to be a greater influence from hydrology and mass-effects in systems with short WRT and a greater influence from local sorting in systems with long WRT (Niño-García et al., 2016). Longer WRT in lakes and larger rivers compared to smaller streams consequently results in less diverse communities due to local sorting (Niño-García et al., 2016). In this way, hydrology and local sorting interact and result in a uni-directional pattern of gradually decreasing diversity from smaller streams to larger rivers and lakes (Niño-García et al., 2016). Furthermore, beyond a WRT of 10 days hydrology has been shown to have no additional impact on the structuring of the microbial community (Niño-García et al., 2016). The importance of WRT for selecting lakespecific phylotypes in a freshwater network was also highlighted in another study (Nelson et al., 2009). The study showed less similarity between the microbial community in the inlet and the outlet of a headwater lake compared to the inlets and the outlets of downstream lakes. This indicates that the first lake selects for a lake-specific community, which is then transported downstream in the network (Nelson et al., 2009). These results also illustrated the importance of the position along the network for understanding the bacterial community structure (Nelson et al., 2009). As the position of water bodies in the system might be a key factor for determining the structure of the microbial community at that particular position, the right spatial resolution is important for understanding the structural changes the microbial community undergoes along a freshwater network.

Together with WRT a terrestrial seed bank for freshwater networks also seem to result in a uni-directional structure of the microbial community (Crump et al., 2012; Ruiz-González et al., 2015). In the catchment of the Toolik Lake, Alaska, a clear pattern of decreasing diversity was shown from soil waters farthest upstream with highest species richness through headwater streams and lastly to lowest richness in lake water (Crump et al., 2012). OTUs originating in soil were numerically dominant throughout a freshwater network in the Eastern boreal region of Québec, Canada, and certain OTUs that were rare in soil were shown to increase in number and become dominant in the downstream freshwater environments (Ruiz-González et al., 2015). These studies indicate that an initial inoculation from soil at the beginning of a freshwater network is followed by a speciessorting process downstream (Crump et al., 2012; Ruiz-González et al., 2015).

On a large spatial scale the uni-directional pattern of decreasing microbial diversity along a river might be explained by the common origin from a highly diverse terrestrial community (Ruiz-González et al., 2015) and by increasing local sorting (Niño-García et al., 2016). However, there might be another pattern on a smaller spatial scale revealed with higher resolution. Higher resolution of samples along a freshwater network might reveal the input of new microbial taxa of different origin along the freshwater network. Input of new taxa along a network could result in a different structuring pattern of the microbial community, which is not uni-directional. The addition of new microbial communities along a freshwater network would be particularly clear in smaller networks where input makes up a larger fraction of the downstream water body. This also implies that in larger networks, the downstream community might mask new communities added along the network. Therefore, river communities on a small spatial scale may show not to have a uni-directional structure, explained by seeding with new microbial communities along the network. Furthermore, it is yet unknown whether the spatial directionality described above extends into saline waters or whether the very different environmental conditions met by the riverine community in the estuary result in a different pattern of the microbial network.

We address the question of how the bacterial communities from the Red River, a small river on the Disko Island, West Greenland (69◦N) are structured at the small spatial scale, by comparing bacterial communities from five sites along the river including input sites from a glacier and a proglacial lake. We investigate whether the structure of the riverine bacterial communities can help explain the structure of the estuary communities by including 23 samples through three transects of the Red River estuary.

Sampling was done in 2013 in August when precipitation events are common and the permafrost active layer thickness is maximum resulting in increased erosion along the river (unpublished data). We hypothesize that the river community is composed of organisms from the surrounding terrestrial environment as well as from upstream freshwater sources, such as glaciers and lakes. Furthermore, due to the relatively short WRT we hypothesize that hydrology rather than local sorting is the dominant factor in shaping the community. We assess to which extent the different communities detected along the river structure the bacterial assemblages in the estuary.

Finally, we test and discuss the potential effects of environmental variables based on multivariate statistical analysis.

#### MATERIALS AND METHODS

#### Sampling

Sampling was carried out in the Red River and its estuary on the Disko Island, West Greenland (69◦N) during August 2013 (**Figure 1**). Around the time of sampling the river flow was 5.7 m3 s −1 . The bedrock consists of iron-rich basalt. As the glacier and the stream erode the bedrock the iron precipitates and gives the marked red color of the river. The river drains directly into the Disko Bay and the freshwater and sediments supplied from the river are mixed with the saline bay waters under varying wave influence. A river plume of high concentrations of sediment is often visible indicating how the supplied sediment is dispersed.

Five locations were sampled in the river with three replicates at each site (**Figure 1**). The top sample (R1) being just upstream of an outlet from an adjacent proglacial lake and the second sample (R2) at the outlet from the proglacial lake. The third sample (R3) being at another outlet to the river supplying water directly from the glacier and the fourth sample (R4) 100 m downstream of R3. Sample R4 was collected at the eastern bank on the opposite site of the upstream outlet from the glacier stream (R3), while all other river samples were collected at the center of the river. The last and fifth sample (R5) was collected close to the river mouth. The distance of each sample to the river mouth is supplied in **Table 1**. In the bay, sampling was done along three transects perpendicular to the coast (**Figure 1**). Each transect consisted of four sampling locations at distances of 100, 300, 700, and 1100 m from the river mouth. At each distance two samples were collected, one surface sample at 0.5 or 1 m from the surface and one deep sample 1 m from the bottom. At water depths above 20 m, the deepest water sample was collected at 20 m depth. In transect 1, 100 m into the estuary (E100) the deep sample is missing so that there are only two replicates (transect 2 and 3) of E100 samples. At E700 one sample, which should have been sampled at 20 m, was sampled at 1 m depth, so that there are 4 replicates of surface samples and 2 replicates of deep samples for E700. Water was sampled by grab sampling using sterile 50 ml syringes (Sarstedt, Germany) either collecting water directly from the river or collected from a 5 L Niskin water sampler (KC Denmark, Denmark) that had been filled at the sampled depth. The 50 ml water samples collected in the syringes were forced through SterivexTM filters (Merck Millipore, MA, USA) and the filters were afterwards partly dried by forcing air from the syringes through the filters. The filters were frozen and kept at −20◦C until analysis.

Temperature, turbidity, and oxygen saturation were measured at all sites using a YSI 6600-V2 CTD sensor with attached probes (YSI, OH, USA). In the bay, the size (in equivalent spherical diameter, ESD), total area and total number of particles were measured in 6 mL water using a laser sheet camera system, the Pcam (Markussen et al., 2016). Individual water samples were taken at all locations and transferred to new 100 ml polyethylene bottles, frozen as quickly as possible and shipped to Copenhagen for further analysis. The total nitrogen (TN) and dissolved organic carbon (DOC) were determined on a Shimadzu TOC-V total organic carbon analyzer (Shimadzu, Japan) with a TNM-1 total nitrogen measuring unit and pH was measured using a Radiometer Analytical SAC90 autosampler (Hach, CO, USA). DOC measurements were based on triplicate measurements. A standard curve using 1000 ppm sodium hydrogen phthalate with concentrations ranging from 0 to 5 ppm were made and a 100 ppm certified Total Organic Carbon (TOC) standard (SCP Science, QC, Canada) was diluted to 1 ppm for use as reference.

#### DNA Extraction and Sequencing

DNA was extracted from the SterivexTM filters using the PowerWater© SterivexTM DNA extraction kit (MO BIO Laboratories, CA, USA), using the protocol provided by the manufacturer. The extracted DNA was stored at −80◦C until library preparation.

The nucleic acid concentrations of all samples were assessed by spectrophotometer (Nanodrop <sup>R</sup> ND-1000, Saveen Werner, SE) to be within the range of 3–5 ng µl −1 . DNA was then amplified in triplicate using universal prokaryotic primers targeting the variable region V4 of the 16S rRNA gene (Caporaso et al., 2011), forward primer 515F (GTGCCAGCMGCCGCGGTAA) and reverse primer 806R (GGACTACHVGGGTWTCTAAT), using the HiFi polymerase (PCR-Biosystem, UK). The primers were supplied with 12 distinct barcode sequences of 4–6 bases each and combined as differential sets, thus labeling the samples with individual differently tagged sequences. All PCR runs included triplicate positive (E. coli) and negative (dd H2O) controls. The resulting PCR products (350 bp) were quality controlled by quantification of concentrations using the Qubit <sup>R</sup> 2.0 dsDNA HS Assay Kit (Life Technologies, CA, USA) and visual inspection of band size following gel electrophoresis. The amplified DNA was then purified using the HighPrepTM PCR size selective carboxyl coted magnetic beads (Magbio, MD, USA). The resulting DNA (average concentration 19.3 ngµl −1 ) was then pooled and ligation of adaptors was performed according to manufacturer's instructions following the Low Sample (LS) Protocol (TruSeq DNA PCR-Free Sample Preparation Guide, Illumina, CA, USA) with minor modifications. Overhangs on the 3′ ends were removed and 5′ ends filled in by end repair, performed as described in the protocol on 1µg DNA. Size selection was replaced with a clean-up step with magnetic bead based chemistry (HighPrepTM PCR, CleanNA). A volume of 100µl from the end repair reaction was purified according to manufacturer's instructions and subsequently eluted in 20µl

FIGURE 1 | Sample sites, Disko Island, West Greenland, 69◦N (Worldview, 2013).


Distance, Distance to river mouth; Oxygen sat, Oxygen saturation; Particle MD, Particle Mean Diameter; Particle TA, Particle Total Area; Particle N, Particle Number. Particle data was measured in a water volume of 6 mL, vol, measured water volume. Transect XX was sampled outside the three transects.

molecular biology grade water (MO BIO Laboratories, CA, USA). Following this step, the 3′ ends were adenylated (adding an "A" nucleotide) to prevent them from ligating to one another during the ligation reaction. Then adaptors with a "T" overhang were ligated onto the DNA fragments of the two assemblages as described in the protocol. The assemblage was ligated with the AD012 index adaptor (CTTGTA), and then subjected to a clean-up step with purification beads provided in the kit. Quality control of the ligated amplicons (∼400 bp) was performed by PCR amplification using primers targeting the index adaptor followed by gel electrophoresis and visual inspection. Finally, the amplicon assemblage was diluted to a concentration of 3.3 ngµl −1 and sequenced with MiSeq 250PE (Illumina), adding 30% PhiX DNA. Demultiplexed merged reads are deposited in the NCBI Sequence Read Archive (SRA) database under SRA accession SRP076603.

#### Computational Analyses

The sequencing data was quality checked using FastQC (Patel and Jain, 2012) and read pairs were merged with the paired-end read merger PEAR (Zhang et al., 2014). Only properly merged reads were used for downstream analysis. Merged reads were processed using Qiime version 1.8.0 (Caporaso et al., 2010a). Demultiplexing with split\_libraries\_fastq.py was performed with quality filtering at phred threshold ≥ 20. Chimeric sequences were removed from demultiplexed data with USEARCH uchime reference-based chimera removal using the Greengenes database from May 2013 as reference (Edgar et al., 2011). Chimera check removed 12.3% of sequences. Operational taxonomic units (OTUs) were subsequently picked based on 97% identity using de novo OTU picking, which also includes taxonomy assignment using PyNAST alignment against the Greengenes core set of 16S rRNA sequences (Caporaso et al., 2010b). Sequences only represented once in the dataset were removed, which reduced the dataset with 13.7%.

Shannon indices (Shannon, 1948), Chao1 richness (Chao, 1984) and rarefaction plots were computed using alpha\_rarefaction.py. Chloroplast sequences were removed and samples were rarefied to the shallowest sample depth of 12,180 sequences per sample with R version 3.1.0 (R Development Core Team, 2008) and R package Vegan (Oksanen et al., 2015). BIOENV analyses were used to assess how well the community structure was explained by environmental variables using non-factorial metadata (**Table 1**; Clarke and Ainsworth, 1993). For BIOENV analysis the Vegan package was used to create distance matrices of environmental data (Euclidean distances) and community composition (Bray-Curtis distances), which were then compared through Spearman's rank coefficients. DOC, TN, temperature, salinity, oxygen saturation, and turbidity were log transformed prior to analysis. Depth was not included for BIOENV analyses including only river samples as depth was constant and particle data was not included for any analyses including river samples, as the data was not available. LabDSV package in R was used for non-metric multidimensional scaling (NMDS) and indicator species analysis. Indicator species are here denoted indicator OTUs and are defined as OTUs having a higher abundance at one site compared to other sites with indicator values d ≥ 0.3 at a significance level of p ≤ 0.05. Indicator values are a product of relative abundance of an OTU in samples from one site (between 0 and 1) and the relative average abundance of that OTU across all sites (Dufrene and Legendre, 1997). The used indicator species and indicator OTU concept in this study are not equal to the Indicator Species concept representing species that are markers for certain environmental variables in an ecosystem. NMDS analyses were conducted using Bray-Curtis distance matrices. NMDS stress values are included in **Figure 3**.

# RESULTS

#### River System Characteristics

Dissolved organic carbon (DOC) and total nitrogen (TN) concentrations in the river were in the same range as those in the estuary (**Table 1**). River site R2 by the lake outlet had the highest DOC and TN concentrations of all samples in the river and estuary. pH values in the river were slightly lower than in the estuary except for R2, which had a higher pH comparable to the estuary samples. Temperature ranged from 1.7◦C at 20 m depth 700 m into the estuary to 12.6◦C at river site R2. Temperatures were generally lower in the deep water samples from the estuary compared to the surface samples. Salinity in the river samples was 0.01 PSU for all samples except for R2 where it was 0.023 PSU. The higher salinity in the water from the lake can be explained by the accumulation of ions in the lake due to longer WRT in the lake compared to the river allowing for evaporation of water from the lake. The longer WRT may also explain the higher temperature at site R2. In the estuary, salinity was consistently lower in the shallow water samples compared to deep water samples at the same distance from the river mouth. This was expected from the lower density of the freshwater from the river being mixed into the estuary. Turbidity across all samples, excluding river site R2, ranged from 8.0 to 19.8 NTU, while it was remarkably lower at R2 (1.2 NTU). Camera data from the estuary showed that particle mean diameter was generally higher at shallow depths compared to deep water samples.

#### Alpha Diversity

Illumina sequencing of variable region V4 of the 16S SSU rRNA gene from a total of 38 samples resulted in 462,840 individual sequences after rarefaction to 12,180 sequences per sample, which were binned into 63,624 unique OTUs (97% sequence identity). The number of observed OTUs was not exhausted at this level of rarefaction (**Supplementary Figure 1**). Shannon indices for the river samples ranged from 5.6 to 10.8 and Chao1 richness in the river ranged from 1408 to 19,117 OTUs per sample (**Figure 2**). The alpha diversity of the bacterial community represented by both Shannon indices and Chao1 richness decreased at R2 and R3, the sites at which the lake and the glacier stream drains into the river (**Figures 1**, **2**). The alpha diversity rose again at R4, ca. 100 m from the glacial input site. At the river mouth (R5), the diversity increased again and reached a similar level to the first river site (R1) upstream of the glacier and lake input sites.

Shannon indices for the estuary samples ranged from 5.4 to 10.6 while Chao1 richness varied from 2589 to 19,021 OTUs (**Figure 2**). The diversity and richness were higher in the estuary than in the glacier stream and lake input samples, and slightly lower than in the remaining river samples. There was no apparent pattern in the difference in diversity and richness attributed to different depths of the estuary, sample sites or the distance to the river mouth.

# Community Composition Analysis

The samples from the first site of the river (R1), upstream of the lake and glacier stream outlets to the river, clustered with samples from the bottom site of the river (R5; **Figure 3**). These two sites also shared a high number of indicator OTUs (**Figures 4A,E**) and showed similar diversity and richness (**Figure 2**). The river site by the proglacial lake outlet (R2) clustered with the river site at the glacier stream outlet (R3). These sites, R2 and R3, also had lower diversity than the other river sites (**Figure 2**).

In the estuary, the bacterial communities clustered according to sample site for the two sites that were farthest into the estuary (E700 and E1100). The samples from the sites closest to the river mouth (E100 and E300) were dispersed across NMDS 1 and 2 (**Figure 3**). The samples did not cluster according to sample depth. Samples from the sample sites closest to the river mouth (E100 and E300) clustered more closely with river samples than the samples farthest from the river mouth (E700 and E1100).

# Environmental Controls

BIOENV analysis showed that the total community as well as the non-indicator and indicator OTUs in the river correlated significantly with turbidity at p ≤ 0.05. The strongest correlation was found between the river non-indicator OTUs and turbidity with a Spearman's rank correlation coefficient of 0.586.

BIOENV analysis of the estuary community showed no significant correlations with environmental variables (**Table 2**).

#### Indicator Taxa Analysis

The number of indicator OTUs in the river ranged from 158 at the second-to-last site of the river (R4) to 678 at the input site from the proglacial lake (R2), which also had the highest percentage of top indicator OTUs (Indicator Value = 1; **Table 3**).

There was a high number of shared indicator OTUs between the top and bottom of the river (**Figures 4A,E**). Two hundred and eight indicator OTUs from R1 were found at R5 while only 14, 31, and 28 indicator OTUs from R1 were found at R2, R3 and R4 respectively. Taxonomic composition of indicator OTUs at Class level showed similar fractions of Flavobacteria and Gammaproteobacteria across river samples. A greater fraction of Actinobacteria were found in R2-R4 while very few Acidobacteria were found in these samples compared to R1 and R5 where also a higher fraction of Unknown were found (**Figure 5**). A number of indicator OTU sequences at the uppermost river site (R1) showed similarity to members of Rhizobiales isolated from plant roots and soil (Lee et al., 2005) as well as to strict anaerobes such as Caldilinea, Anaerolineaceae (Yamada et al., 2006), and Desulfobacteraceae (Garrity et al., 2006; **Figure 5**).

Indicator OTUs identified at the outlet from the lake and glacier stream were found in low numbers at the other river sites (**Figures 4B,C**). The lake outlet site (R2) had the highest number of indicator OTUs and percentage of top indicator OTUs (**Table 3**). A number of taxa known to be psychrophilic, such as Moritella (Urakawa et al., 1998), Polaribacter (Gosink et al., 1998), Oleispira (Yakimov et al., 2003), Crocinitomix (Bowman et al., 2003), and Psychromonas (Mountfort et al., 1998) were found among the best matches for the indicator OTUs from the lake outlet, unlike at the other river sites.

The distribution of estuary indicator OTUs showed a different pattern than the river indicator OTUs (**Figures 4F–I**). The number of indicator OTUs in the estuary was generally lower than at the river sites. An exception to this was the outermost estuary sample (E1100), which had a number of indicator OTUs comparable to the river sites (**Table 3**). No top indicator OTUs were found in any of the estuary samples, meaning that no OTUs from the estuary were unique to any of the sample sites. The indicator OTUs for each sample site in the estuary were found only in low numbers at the other sites both in the river and the estuary (**Figure 4**) and the taxonomic composition at Class level was less similar among the estuary samples than among the river samples (**Figure 5**).

On average, the bacterial communities in the estuary were made up of 23% river indicator OTU sequences (**Figure 6**). There was an overall decreasing contribution of river indicator OTUs in the estuary sites with 26–27% river indicator OTUs closest to the river mouth at E100 sites, 17–25% at E300 sites and 8– 10% at E1100 sites. E700 sites were exceptions with 22–52% of the community being river indicator OTUs (**Figure 6**). Closer to the river mouth at sites E100 and E300 there was a larger fraction of the indicator OTUs from the top of the river (R1) and river mouth (R5), except for the deep sample at E100, where the distribution of river indicator OTUs was similar to the E700 estuary sites (**Figure 6**). In E700 both in the deep and surface

samples R2 (lake outlet) indicator OTUs were more abundant than in the other estuary samples, and were more abundant than indicator OTUs from any other river sites (**Figure 6**). At the estuary sites farthest from the river mouth (E1100) the samples had the highest fraction of non-river indicator OTU sequences (**Figure 6**).

# DISCUSSION

# Alpha Diversity

Shannon indices for the river samples at the lake outlet (R2) and glacier stream outlet (R3) were comparable to a recent study of 87 small streams and rivers in the La Côte-Nord region of Québec, Canada, where OTUs were clustered with the same method as in the present study (Ruiz-González et al., 2015). The remaining river samples had slightly higher diversity than found in previous studies (Galand et al., 2006, 2008; Crump et al., 2009). The difference from less recent studies is likely due to the difference in the technologies applied and the resulting lower number of sequences in the previous studies. Together with sequencing technologies, which have changed dramatically in the last decade, OTU clustering has shown to have a great impact on the detected alpha diversity (Sinclair et al., 2015). Therefore, the comparison of alpha diversity among studies should be interpreted with care. Our results were obtained with the use of the Qiime pipeline (Caporaso et al., 2010a), which has shown to create a larger number of OTUs when compared to other popular clustering methods (Sinclair et al., 2015). Consequently, we might detect a higher diversity because of the clustering method used.

In an extensive study of freshwater networks, small streams were shown to have higher Shannon indices than larger rivers (Ruiz-González et al., 2015). This was attributed to the common terrestrial origin of the microbial community resulting in an initially high diversity in the small streams originating from the surrounding soil (Ruiz-González et al., 2015). This is in contrast with our results showing less diversity in the glacier stream compared to the main river (**Figure 2**). The lower diversity might be due to the origin of this stream in a glacier rather than subsurface groundwater and surface runoff as described in the above-mentioned study (Ruiz-González et al., 2015). This notion is supported by the fact that the diversity of the glacier stream outlet (R3) is comparable to those recently described for proglacial lakes (Peter and Sommaruga, 2016). Also, the indicator OTUs from the glacier stream were similar to taxa commonly found in freshwater and marine environments (details not shown). These results emphasize the importance of high spatial resolution for assessing the origin of the metacommunity in a complex freshwater network. In this study, the diversity along the network does not follow a uni-directional pattern (**Figure 2**). Our results illustrate that the origin and structuring of the microbial community might be very different from one network to another. How glaciers and glacier streams affect the metacommunity of freshwater networks is a highly relevant topic yet to be investigated.

The Chao1 richness in the river samples was higher than in the previous papers focused on large Arctic rivers (Galand et al., 2006, 2008) but comparable to that described in a recent paper using the same sequencing platform (Niño-García et al., 2016).

The drop in alpha diversity at the input sites from the lake and the glacier (R2 and R3, **Figure 2**) shows that the lake and the glacier stream input less diverse bacterial communities into the main river. Lower diversity in lakes compared with the connected rivers has been attributed to longer WRT in lakes (Crump et al., 2012; Ruiz-González et al., 2015; Niño-García et al., 2016). This is especially pronounced in small streams and rivers, where WRT is too short to allow for local sorting of the bacterial community (Crump et al., 2012; Ruiz-González et al., 2015; Niño-García et al., 2016). Downstream of the input sites, the alpha diversity rises again and by the river mouth reaches a level similar to the first river site (R1) upstream of the input sites (R2 and R3). This shows that the volume of water from the lake and glacier outlets does not dilute the downstream river community. Importantly, it also suggests that the less diverse communities from the lake and the glacier stream are concealed downstream of the input sites by the higher diversity of the main river.

Shannon indices in the estuary samples were higher than those previously described for Arctic estuaries (Galand et al., 2006, 2008). Previous studies of large Arctic rivers show that bacterial diversity and abundance decrease from rivers to estuaries probably due to upstream input from terrestrial sources (Meon and Amon, 2004; Galand et al., 2006, 2008). Our results from a small river support this conclusion by showing a slight decrease in diversity from the main river sites (R1 and R5) to the estuary (**Figure 2**). The diversity in the estuary sites closer to the river and in the shallow samples could be expected to be higher than more distant and deep estuary samples due to a higher concentration of the river bacterial community, which is not evident from our results (**Figures 2**, **6**). This indicates that although the community structure from the river to the estuary aligns with previous results by showing a directional decrease in diversity, this directional structure cannot be detected further down the network, in the estuary transects. The lack of pattern in diversity and richness attributed to different depths of the estuary samples or the distance to the river mouth could partly be attributed to an insufficient resolution in sample depth. The low resolution might not allow for the detection of a clear plume

#### TABLE 2 | BIOENV analysis of the total, river (n = 15) and estuary (n = 23) communities and the indicator OTUs and non-indicator OTUs.


TABLE 3 | Number of Indicator OTUs and Top Indicator OTUs across sample sites.


\* Including only indicator OTUs with Indicator Value ≥ 0.3 and P-value ≤ 0.05. \*\* Indicator Value = 1.

and different depth zones. While distinct bacterial communities have been found to be associated with the plume and different oceanic zones in an estuary, these results are from sampled oceanic zones several kilometers farther into the ocean than our samples (Fortunato et al., 2012). Estuary samples were previously discussed as harboring a mix of bacterial communities from the river and the coastal ocean with no distinct autochthonous estuary-community (Fortunato et al., 2012), which resembles our results. The high variability in diversity found among the estuary samples suggests a highly heterogeneous community, also expected in such a region where waters of very different chemistry and origin meet. The large variability in diversity and richness seen at the different estuarine sites (**Figure 2**) could be explained by the sampling of the different water masses, since sequences from the three different transects were pooled together for each estuarine site and depth. The variability of diversity as well as environmental variables seem to lessen farther into the estuary, which could be expected as homogeneity increases as a greater fraction of the estuary is made up of marine waters (**Figure 2** and **Supplementary Table 1**).

#### Community Composition Analysis

Samples from the first site of the river (R1), upstream of the lake and glacier stream outlets, clustered with samples from the bottom site of the river (R5) while the river site by the proglacial lake outlet (R2) clustered with the river site at the glacier stream outlet (R3) as shown by the NMDS plot (**Figure 3**). R1 and R5 also shared a high number of indicator OTUs (**Figures 4A,E**) as well as similar diversity and richness (**Figure 2**). In contrast, R2 and R3 had lower diversity than the other river sites. The NMDS plots, indicator OTU analysis and alpha diversity results imply that waters sourced from the lake and the glacier stream carry different bacterial communities than that of the main river. The larger volume of the main river community then probably masks the lake and glacier stream communities, thus resulting in the close similarity between the sites R1 and R5. The isolation of the R4 samples from other river samples in the NMDS plots may be explained by the difference in sampling at this site, which was closer to the river bank compared to the other river samples. Another explanation might be the imperfect mixing of water from the upstream lake and glacier outlets with that of the main river at this site. The latter seems to be the best explanation since the comparably low number of indicator OTUs found at R4 suggests that this site contains a mixture of the upstream communities rather than a distinct community from the sampling site (**Table 3**). This agrees with results from a study on an Arctic tundra catchment, showing that streams leaving lakes have decreasing similarity to the lake microbial community as a function of distance (Crump et al., 2007).

The dispersal of estuary samples on the NMDS plots was in accordance with the diversity measures, which were similar within individual sampling sites independent of sample depth (**Figures 2**, **3**). Samples from the sites closest to the river mouth (E100 and E300) clustered more closely with river samples than the samples farthest from the river mouth (E700 and E1100), consistent with a gradual mixing of the river community with a marine community within the estuary environment. Remarkably, R4 river samples clustered more closely with estuary samples than with the other river samples. Indicator OTUs from R4 are present throughout the estuary transects (**Figure 6**) and the taxonomic composition of samples from R4 has the largest resemblance to that of estuary site E1100 (**Figure 5**), which might explain the NMDS results (**Figure 3**). Furthermore, the bacterial community at this site seems to be a mixture of the different river communities as suggested by the indicator OTU results (**Table 3**). Therefore, the clustering of R4 samples with estuary samples might also reflect the resemblance to the estuary, in which the river communities are also mixed (**Figures 4F**–I, **6**).

Samples from E100 and E300 were more widely dispersed across the NMDS plots than the samples from farther into the estuary, indicating greater heterogeneity of the bacterial communities. This is not unexpected from a region of mixing of largely different water bodies both in terms of physical and chemical variables as well as origin. The NMDS plot did not indicate that the bacterial communities were stratified according to sample depth. The low resolution of samples through the water column might be part of the explanation. However, the results

might also indicate a high degree of mixing through the water column of the Red River estuary at the time of sampling. As previously discussed, this may also be explained by the proximity to the river of the estuary samples in this study compared to other studies, where bacterial communities in the estuary were shown to be stratified according to depth (Fortunato et al., 2012).

#### Environmental Controls

The BIOENV analysis did not show strong correlations between environmental variables and community composition but it did highlight turbidity as a community-shaping factor in the river (**Table 2**). The bacterial community in a freshwater network fed by glaciers has recently been shown to be structured along the turbidity gradient (Peter and Sommaruga, 2016). The BIOENV results support the idea that the bacterial community in the Red River freshwater network is partly sourced from the glacier. The lower turbidity at R2 (1.2 NTU) compared to an average of 16.3 NTU (SD = 1.2) at the other river sites is noteworthy since proglacial lakes are known to have high turbidity (Peter and Sommaruga, 2016). While turbidity of the proglacial lake outlet (R2) is higher than that shown for a non-glacier fed lake in the Austrian Central Alps, it is remarkably low compared to other glacier-fed lakes (Peter and Sommaruga, 2016). This might indicate that the proglacial lake is losing hydrological connectivity to the glacier (Peter and Sommaruga, 2016). It should be taken into consideration that the samples are not taken from the actual lake but several 100m downstream (**Figure 1**).

The Red River is small in size compared to large rivers previously described such as the Mackenzie River (Galand et al., 2008; Garneau et al., 2009) and the Columbia River (Fortunato et al., 2013). For comparison, the average water flow from August to November in the Columbia River was 2988 m<sup>3</sup> s −1 (Fortunato et al., 2013), while the river flow in the Red River around sampling time was 5.7 m<sup>3</sup> s −1 . An estimated time from top sampling site R1 to the river mouth at R5 is 40 min for the moving water body where the samples are taken. It has been shown that at sites with shorter WRT than 10 days the bacterial community composition was predominantly structured by hydrology (Niño-García et al., 2016). Accordingly, we hypothesized that hydrology would be dominant in shaping the bacterial community in the relatively small Red River with short WRT. Consequently, we did not expect strong correlations between community composition and environmental variables in the river samples. Our study represents a single catchment with short WRT and the results of the BIOENV analysis agrees with previous results by showing weak correlations between the bacterial community and environmental variables (Niño-García et al., 2016).

Salinity has previously been highlighted as a communityshaping factor in estuaries and rivers. For example, the abundance of Alphaproteobacteria, Betaproteobacteria, and Actinobacteria correlated strongly with salinity in the Delaware estuary where a strong negative correlation between Betaproteobacteria and Actinobacteria was shown together with a positive correlation between salinity and Alphaproteobacteria (Kirchman et al., 2005). Salinity, together with temperature, explained 45% of the variation in the community composition in a study of the Mackenzie Shelf (Garneau et al., 2009). Salinity was not identified as a significant factor in the BIOENV analysis. The lack of correlation to salinity in our study is also evident from the NMDS analysis (**Figure 3**), where R4 river samples cluster with estuary samples despite the large difference in salinity between these environments (**Table 1**). These results suggest that there may be environmental or hydrological factors other than salinity that explain the observed patterns in taxonomic composition in the study site.

BIOENV analysis of the estuary community showed no significant correlations with environmental variables (**Table 2**). Also no correlation was found between the bacterial community and spatial variables including distance from the river mouth and depth. The results from the BIOENV analyses indicate that the bacterial community in the estuary is not dispersed according to environmental variables or stratified according to distinct water bodies of riverine or oceanic origin, supporting the results from the diversity assessments (**Figure 2**) as well as the NMDS plots (**Figure 3**). Our samples represent a very small fraction of the total estuary; a higher resolution of samples in the estuary might result in more conclusive results.

#### Indicator Taxa Analysis

Indicator OTUs identified at the input sites from the lake and glacier stream were found in low numbers at the other river sites and, therefore, seem to be specific to their respective sources (**Figures 4B,C**). Notably, of the 678 indicator OTUs from the lake outlet (R2, **Table 3**), 570 OTUs were not found in the upstream river site (R1) and seem to originate from the proglacial lake. The lake outlet site (R2) had a particularly high number of indicator OTUs and percentage of top indicator OTUs (**Table 3**). Water bodies with longer WRT have been shown to harbor a less diverse and more differentiated community explained by local sorting of the microbial community (Niño-García et al., 2016). Results from the indicator OTU analysis and diversity of the lake outlet site (R2) show that the lake with a longer WRT has a less diverse and more specialized community compared with the river. A number of taxa known to be psychrophilic, such as Moritella (Urakawa et al., 1998), Polaribacter (Gosink et al., 1998), Oleispira (Yakimov et al., 2003), Crocinitomix (Bowman et al., 2003), and Psychromonas (Mountfort et al., 1998), were found among the best matches for the indicator OTUs from the lake outlet, unlike at the other river sites.

The short WRT in the Red River network should accordingly result in a low degree of differentiation, which is confirmed by the low number of top indicator OTUs (i.e., OTUs unique for a particular site), which average <2% in present study (**Table 3**). Another study of freshwater networks highlighted an average of 11% unique OTUs between different ecosystems as representing a low number (Ruiz-González et al., 2015). These results differ from our study in that they considered many different lakes while our results are obtained from one lake only (Ruiz-González et al., 2015).

Samples from the lake and glacier outlets (R2 and R3) as well as the sample site just after the glacier outlet (R4) had very few Acidobacterial classes compared to the top and bottom site of the river (**Figure 5**). Acidobacterial classes were shown to be most common in soil compared to the adjacent freshwater network and Acidobacteria in rivers seem to be sourced from the surrounding terrestrial environment (Ruiz-González et al., 2015). The more differentiated lake community thus seems to harbor a lower fraction of organisms from the surrounding soil community compared to the main river. This might partly be explained by the local sorting of the bacterial community in the lake. It could potentially also to be explained by the presence of taxa with different origin than the main river community, as indicated by the large number of indicator OTUs, which are not present in the upstream site. The lower turbidity at the lake outlet site (R2) might also indicate that there is less input of soil to the lake than to the main river, which causes less mass dispersal effect from the surrounding terrestrial environment. The lower turbidity might, however, also be explained by less suspended particles in the lake because of increased sedimentation due to the longer WRT.

A great number of previous studies of riverine microbial communities have suggested and shown that the river communities are influenced by input of microorganisms from surrounding soil environments (Crump and Baross, 2000; Galand et al., 2006, 2008; Crump et al., 2012, 2007; Ruiz-González et al., 2015; Niño-García et al., 2016). Our results support this by showing that potentially soil-related taxa make up a significantly large fraction of the bacterial community making them part of the indicator OTUs of the main riverine bacterial community (**Figure 5**). Notably, the results also indicate that, along the river, distinct communities may not have the same degree of influence from the terrestrial surroundings. August is a month of high precipitation and increased erosion around the Red River, which would result in a relatively high influence from the surrounding soil community. The influence from soil may be less pronounced in other months as water flow and erosion levels change.

The dominance of soil microbes in freshwater networks has been established in several recent studies, highlighting soil as the origin of the network metacommunities (Ruiz-González et al., 2015; Niño-García et al., 2016). A gradual differentiation of a stream from an upstream lake as a function of distance has been attributed to the origin of the freshwater communities from a terrestrial metacommunity (Crump et al., 2012, 2007). Our results suggest that glaciers may also supply part of the metacommunity resulting in a different structuring pattern of the network. In our case the structuring pattern was not unidirectional throughout the network but rather showed local changes as different bacterial communities were added to the river. This is illustrated in the diversity results (**Figure 2**) as well as the NMDS plots of the community composition (**Figure 3**). These results together with the indicator OTU analysis highlight the importance of additional sources of the metacommunity such as glaciers.

The indicator OTUs from the lake-sourced water (R2) can be found in the second highest abundance in the estuary site 700 m into the estuary (E700) (**Figures 4B**, **6**). The taxonomic composition of E700 differs from the other estuary sites and this site contains a high fraction of Alphaproteobacteria (71%) and a relatively small fraction of Gammaproteobacteria (3%) (**Figure 5**). Of the 678 R2 indicator OTUs 161 are found at the estuary site 700 m into the estuary, where they make up 20% of the sequences at E700 with a higher fraction in the deep samples compared to the surface samples (**Figures 4B**, **6**). This resembles results from the Columbia River, where the estuary samples were comprised of just over 20% riverine community (Fortunato et al., 2012). The distribution of R2 indicator OTUs suggests that although the organisms from the lake do not form a large enough fraction of the community to be notable along the downstream river, they are transported into the estuary where they form a larger fraction of the community. Our results align with the "landscape reservoir" concept proposed for the Toolik lake, Alaska, where rare organisms from the upslope landscape influence downslope bacterial diversity and become dominant in environments with favorable conditions (Crump et al., 2012).

The taxonomy of the nine indicator OTUs from R2 found in high numbers (>100 sequences) in the E700 samples were mostly related to organisms isolated from oceanic environments such as Marinomonas (Van Landschoot and De Ley, 1983), Oleispira (Yakimov et al., 2003), Pseudoalteromonas (Bowman, 2007), Polaribacter (Gosink et al., 1998), and Sulfitobacter (Sorokin, 1995). Of related non-marine organisms were Glaciecola, which was first described as a Gammaproteobacterium isolated from Antarctic sea ice (Shivaji and Reddy, 2014) and Rhodobacteraceae known from aquatic environments (Pujalte et al., 2014). The fact that indicator OTUs from the proglacial lake outlet to the river are similar to known marine organisms suggests that these organisms are commonly found in marine environments and that they are not originally known from terrestrial environments. Since it is unlikely that organisms are transported from the estuary to the proglacial lake over 2 km upstream, these organisms in the estuary more likely originate from the upstream freshwater network. Possibly, they become such common organisms in the estuarine and marine environments, that these are the environments from which they have become known. It is well established that bacterial communities found in freshwater networks can be traced from upstream positions in the network (Crump et al., 2012; Ruiz-González et al., 2015; Niño-García et al., 2016). We show that in the Red River estuary the river community can be found in the estuary with an overall decreasing fraction from the river mouth toward the ocean (**Figure 6**). Interestingly, communities that are not notable throughout the river are transported to the estuary where they seem to become an equally large fraction of the estuary as the main river community (**Figures 4B**, **6**). Distinct communities from the river seem to influence the estuary to different extend, so that communities from certain parts of the river make up notably larger fractions of the estuary at some sites (**Figures 4**, **6**).

"Seed bank" is a term proposed for the fraction of dormant organisms that may be resuscitated when met with different environmental conditions through e.g., dispersal to other ecosystems (Lennon and Jones, 2011). The concept of seed banks was recently extended to freshwater networks where organisms originating in a soil community were proposed as the seed bank for boreal freshwater networks (Ruiz-González et al., 2015). For freshwater networks it was discussed that shallower sequencing depth might lead to the erroneous conclusion that freshwater communities do not derive from a shared pool of terrestrial microbes (Ruiz-González et al., 2015). This could lead to an incomplete understanding of the mechanisms of assembly and the actual linkages and dispersal of microbes between connected ecosystems (Ruiz-González et al., 2015). We show that sampling resolution not only in terms of sequencing depth but also resolution along the network may result in overlooking distinct microbial communities and how these are distributed and linked to the downstream estuary. Our results indicate that not only the terrestrial surroundings but also upstream glaciers may act as seed banks for freshwater networks. While the uni-directional structure in freshwater networks might be a consequence of the numerical dominance of terrestrial OTUs as shown previously (Crump et al., 2012; Ruiz-González et al., 2015) our results suggest that this does not necessarily imply that the bacterial community in a freshwater network has a common origin from microbes from soil. A higher resolution along the river might reveal distinct bacterial communities of different origin and with different composition, which are introduced downstream in the network. These distinct communities, which might be concealed by the numerically dominant terrestrial community along the river, are able to act as seed banks for downstream environments. The different composition of inputs along the river affects the structure of the community, which is not necessarily unidirectional for all freshwater networks as shown in the present study.

Therefore, sampling with the right resolution, both in terms of sequencing depth and the distance between samples along the network, is crucial for understanding the source of microbial communities found in the estuary. This is especially true at times with high precipitation and erosion. Our study shows that with the right resolution, microbial communities can be valuable in understanding transport pathways of meltwater and matter from source to oceans in that they can serve as both tracers as well as indicators of origin in their adaptation to the environment.

The indicator OTUs for each sample site in the estuary were found only in low numbers at the other sites both in the river and the estuary. This pattern is in contrast with the results from the Columbia River, USA, where indicator OTUs from the upper water-column of the estuary (<56 m depth) showed generalist taxa qualities by having high relative abundance and occurrence in a high number of samples outside their indicator environment when compared to the river environment (Fortunato et al., 2013). Ocean-specific taxa are not expected to be found upstream in the river, which is also apparent from the distribution of estuary indicator OTUs (**Figures 4F–I**). The low number of indicator OTUs and lack of top-indicators in the estuary show that the different sites in the estuary do not hold distinct communities. These results are in accordance with the results of the BIOENV analysis and NMDS plots, which suggest that the bacterial communities are not dispersed according to environmental or spatial variables, as well as the highly variable diversity measures in the estuary. As previously discussed, samples in the present study are sampled relatively close to the river mouth and a more distinct stratification of the bacterial communities might become visible farther into the estuary.

We expected the estuary sites to contain a mixture of the communities found in the river and the ocean, with more environmental variability closer to the river mouth due to the mixing of river- and sea-water. This is supported by our data, which show greater variance in environmental data closest to the river mouth (**Supplementary Table 1**) as well as river indicator OTUs from more of the river sites closer to the river mouth at E100 and E300 (**Figure 6**). Mixing of river and ocean water may result in an allochthonously dominated community shaped by hydrology rather than by environmental selection, also indicated by the NMDS plot (**Figure 3**) and BIOENV analysis (**Table 2**). This is supported by the fact that the outermost sample site of the estuary (E1100) had a higher number of indicator OTUs (**Table 3**) as variability is expected to decrease with increasing distance from the river mouth and the most distant estuary site is expected to contain a higher number of ocean indicator OTUs. This was supported by the taxonomy of the indicator OTUs that were all similar to marine-related taxa at E1100 (details not shown). This site also had a lower fraction of river indicator OTUs compared to the estuary sites closer to the river mouth (**Figures 4**, **6**).

#### CONCLUSIONS

The bacterial community in the Red River, a small river on the Disko Island, West Greenland, is sourced partly from the surrounding terrestrial environment but also receives distinct microbial communities from a proglacial lake and a glacier stream that harbor lower diversity and different composition than the main river. These input communities are less influenced by terrestrial sources than the main river and the proglacial lake input has a higher fraction of OTUs resembling psychrophilic taxa. The combined community in the river is then mixed with oceanic waters in the estuary, where the indicator OTUs of the river communities made up on average 23% of the estuary community at different sites. While the indicator OTUs from the lake and glacier outlets are not notable in the downstream river they make up large fractions of the community at some sites in the estuary. The bacterial community of the river showed a weak correlation to turbidity while the estuarine bacterial community showed no correlation to environmental or spatial variables. Our results illustrate the added value of examining bacterial communities to better understand and trace the transport of meltwaters from their source to the oceans. Lastly the results show that sampling resolution along the river is crucial for understanding the source of different bacterial communities in a river and estuary system.

#### AUTHOR CONTRIBUTIONS

CJ and TM designed the study and sampled, NO performed sample preparations and DNA sequencing, AH performed bioinformatical analyses with contributions from JB and TS and statistical analyses with contributions from MS, AH wrote the manuscript with contributions from MS, TM, NO, BE,

#### REFERENCES


and CJ. All authors discussed the results and reviewed the manuscript.

#### FUNDING

This work was supported by the Center for Permafrost (CENPERM) Center no 100 from the Danish National Research Foundation (DNRF100) as well as the Novo Nordisk Foundation Center for Biosustainability.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01474

Supplementary Figure 1 | Rarefaction curve showing number of observed species against sequences per sample.

Supplementary Table 1 | Statistics of environmental data for estuary samples.


Garrity, G., Brenner, D. J., Staley, J. T., Krieg, N. R., Boone, D. R., Vos, P. D., et al. (2006). Bergey's Manual <sup>R</sup> of Systematic Bacteriology: Volume Two: The Proteobacteria. Berlin: Springer Science & Business Media.


the northern high-latitude environment. Clim. Change 46, 159–207. doi: 10.1023/A:1005504031923


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Hauptmann, Markussen, Stibal, Olsen, Elberling, Bælum, Sicheritz-Pontén and Jacobsen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Biogeochemical and Microbial Variation across 5500 km of Antarctic Surface Sediment Implicates Organic Matter as a Driver of Benthic Community Structure

Deric R. Learman<sup>1</sup> \*, Michael W. Henson<sup>2</sup> , J. Cameron Thrash<sup>2</sup> , Ben Temperton<sup>3</sup> , Pamela M. Brannock <sup>4</sup> , Scott R. Santos <sup>4</sup> , Andrew R. Mahon<sup>1</sup> and Kenneth M. Halanych<sup>4</sup>

*<sup>1</sup> Department of Biology, Institute for Great Lakes Research, Central Michigan University, Mt. Pleasant, MI, USA, <sup>2</sup> Department of Biological Sciences, Louisiana State University, Baton Rouge, LA, USA, <sup>3</sup> Department of Biosciences, University of Exeter, Exeter, UK, <sup>4</sup> Department of Biological Sciences, Auburn University, Auburn, AL, USA*

#### Edited by:

*Julie Dinasquet, Scripps Institution of Oceanography, USA and Observatoire Oceanologique de Banyuls, France*

#### Reviewed by:

*Jean-Christophe Auguet, Centre National de la Recherche Scientifique, France Stephanie Ann Carr, Colorado School of Mines, USA*

> \*Correspondence: *Deric R. Learman deric.learman@cmich.edu*

#### Specialty section:

*This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Microbiology*

Received: *24 November 2015* Accepted: *22 February 2016* Published: *23 March 2016*

#### Citation:

*Learman DR, Henson MW, Thrash JC, Temperton B, Brannock PM, Santos SR, Mahon AR and Halanych KM (2016) Biogeochemical and Microbial Variation across 5500 km of Antarctic Surface Sediment Implicates Organic Matter as a Driver of Benthic Community Structure. Front. Microbiol. 7:284. doi: 10.3389/fmicb.2016.00284* Western Antarctica, one of the fastest warming locations on Earth, is a unique environment that is underexplored with regards to biodiversity. Although pelagic microbial communities in the Southern Ocean and coastal Antarctic waters have been well-studied, there are fewer investigations of benthic communities and most have a focused geographic range. We sampled surface sediment from 24 sites across a 5500 km region of Western Antarctica (covering the Ross Sea to the Weddell Sea) to examine relationships between microbial communities and sediment geochemistry. Sequencing of the 16S and 18S rRNA genes showed microbial communities in sediments from the Antarctic Peninsula (AP) and Western Antarctica (WA), including the Ross, Amundsen, and Bellingshausen Seas, could be distinguished by correlations with organic matter concentrations and stable isotope fractionation (total organic carbon; TOC, total nitrogen; TN, and <sup>13</sup> δ C). Overall, samples from the AP were higher in nutrient content (TOC, TN, and NH <sup>+</sup> <sup>4</sup> ) and communities in these samples had higher relative abundances of operational taxonomic units (OTUs) classified as the diatom, *Chaetoceros,* a marine cercozoan, and four OTUs classified as *Flammeovirgaceae* or *Flavobacteria*. As these OTUs were strongly correlated with TOC, the data suggests the diatoms could be a source of organic matter and the *Bacteroidetes* and cercozoan are grazers that consume the organic matter. Additionally, samples from WA have lower nutrients and were dominated by *Thaumarchaeota*, which could be related to their known ability to thrive as lithotrophs. This study documents the largest analysis of benthic microbial communities to date in the Southern Ocean, representing almost half the continental shoreline of Antarctica, and documents trophic interactions and coupling of pelagic and benthic communities. Our results indicate potential modifications in carbon sequestration processes related to change in community composition, identifying a prospective mechanism that links climate change to carbon availability.

Keywords: benthic communities, Antarctica, aquatic microbiology, biogeochemistry, microbial ecology

# INTRODUCTION

Changing climate in Antarctica has potential to initiate a domino effect that could impact ecosystems from continental ice sheets to the seafloor. Recent research has shown the West Antarctic Ice Sheet (WAIS) is one of the fastest warming locations on Earth (Bromwich et al., 2013; Hillenbrand et al., 2013). Consequently, the WAIS has been losing ice mass (Rignot et al., 2011; Pritchard et al., 2012; Shepherd et al., 2012; Depoorter et al., 2013), which not only has the potential to cause changes in sea-level (Bindschadler, 2006), but also to influence water temperature, salinity (Arneborg et al., 2012 and references therein) and nutrient cycling. The melt season is known to stimulate primary productivity (Smith and Gordon, 1997; Arrigo et al., 1998; Ducklow et al., 2006; Smith et al., 2007), which then increases the amount of organic matter sourced from the water column to sediments (Billett et al., 1983; Wefer et al., 1988; Honjo et al., 2000; Kennedy et al., 2002; Ducklow et al., 2006; Gillies et al., 2012). Further, increasing melt from the continental ice sheet would introduce more terrestrial carbon into marine and benthic environments, altering the quantity and type of carbon. Thus, a greater understanding of Antarctic ecosystems is essential to predict how changing climate will influence organic fluxes between benthic and pelagic communities.

Microbial marine communities play an important role in mediating nutrient cycling to both the water column (reviewed in Arrigo, 2005) and the deep biosphere (reviewed in Edwards et al., 2012) and diversity in these sediments is sensitive to environmental changes. Microbial sediment communities are important as they have been shown to play a major role in carbon cycling (Mayor et al., 2012), in addition to other biogeochemical cycles, such as sulfur, nitrogen, and phosphorus (Azam and Malfatti, 2007; Jorgensen and Boetius, 2007; Falkowski et al., 2008; Edwards et al., 2012). Further, marine sediment communities can be impacted by various chemical and physical parameters (Austen et al., 2002; Schauer et al., 2010; Zinger et al., 2011; Bienhold et al., 2012; Durbin and Teske, 2012; Liu et al., 2015; Nguyen and Landfald, 2015). In Antarctica, recent studies of planktonic marine microbial communities have shown beta diversity is correlated to both physical oceanographic (Wilkins et al., 2013) and chemical variation (Luria et al., 2014; Signori et al., 2014). Sediments from the Ross Sea (Carr et al., 2013) and the Drake Passage at the Antarctic Polar Front (Ruff et al., 2014) have shown organic matter can increase estimates of microbial abundance based on phospholipids and DNA sequencing, respectively. Another study examining sediments from the Southern Ocean (Jamieson et al., 2013) has shown organic matter did not impact species richness. Since climate change in Antarctica will alter the flow of organic matter and nutrients to benthic sediments, understanding how these changes will impact diversity is essential to predicting change in ecosystem functions.

In the present study, we examine the relationships between organic matter, nutrient content, and sediment microbial diversity. While several studies have examined microbial diversity in Antarctic sediments (e.g., Bowman and McCuaig, 2003; Bowman et al., 2003; Powell et al., 2003; Baldi et al., 2010; Carr et al., 2013, 2015; Jamieson et al., 2013; Ruff et al., 2014), each has used various sequencing methods, collected different types of geochemical parameters, and some have had a focused geographic scope. In contrast, we collected 24 sediment samples over a 5500 km transect of Western Antarctica that spans the Ross to the Weddell Seas. Small subunit (SSU) rRNA genes from all three domains of life were sequenced via Illumina MiSeq and correlated with geochemical and nutrient data. The total dataset greatly expands our existing understanding of benthic Antarctic sediments communities, and demonstrates important correlations between organic matter and sediment microbial diversity.

### MATERIALS AND EXPERIMENTAL METHODS

#### Sampling Details

Surface sediment samples were collected from the continental shelf of Antarctica during two research cruises. The first cruise (Dec. 2013–Feb. 2014, RVIB Nathaniel B. Palmer) sampled Western Antarctica (WA), which includes the Amundsen Sea, Bellingshausen Sea, and Ross Sea, using a multicorer (**Figure 1**). The second cruise (Nov.–Dec. 2014, ASRV Laurence M. Gould) sampled the Antarctic Peninsula (AP) using a box corer. Samples were collected on the Antarctic shelf at depths ranging from 223 to 820 m. The top 3 cm of sediments were transferred into sampling tubes and stored frozen (−80◦C). Samples were shipped frozen to Central Michigan University (CMU) within 3 months of collection. More details about sampling locations are found in Table S1.

# Sediment Chemical Analysis

Sediment samples were homogenized and shipped to EcoCore Analytical Services at Colorado State University for stable isotope ( <sup>13</sup>C and <sup>15</sup>N) and percent nitrogen and organic carbon analyses. Macronutrient (total organic carbon; TOC, total nitrogen; TN, NO<sup>3</sup> <sup>−</sup>, NH<sup>4</sup> <sup>+</sup>, S, P) analyses were conducted along with collection of available trace nutrient (e.g., metals) data at the Soil, Water, and Plant Testing Laboratory at Colorado State University (Table S1). The available fraction was defined as the metals that were extracted via Mehlich 3 acid digestions (Mehlich, 1984), and thus does not include an insoluble mineral component.

Correlations between chemical parameters were statistically examined via a Spearman's rank-order correlation in SPSS Statistics 22 (IBM, 2013). If a parameter had high correlation (R > 0.9 and significant at 0.01 level, two-tailed t-test) with another variable (%N and %TOC), then one parameter was removed from the dataset for downstream analysis (%N was removed). Also, data that were reported in percent or a ratio were transformed using an arcsine square root. Resulting nutrient data were then examined for broad trends via principal components analysis (PCA) in the statistical software PAST (Hammer et al., 2001).

# Microbial Taxonomic Analysis

DNA was extracted using a PowerSoil DNA extraction kit (MoBio) following the manufacturer's protocol. Approximately

4–8 extractions were completed on each sediment sample and pooled and concentrated with a DNA Clean and Concentrator kit (Zymo) due to low yields from some of the samples. DNA was then quantified using the Qubit2.0 Fluorometer (Life Technologies) and stored at −20◦C. Bacterial and archaeal sequences were generated from the V4 region of 16S rRNA gene using the primer set 515f and 806r (Caporaso et al., 2012) while eukaryotic sequences were obtained with the V4 region of the 18S rRNA gene using the primer set 1391r and EukBR (Amaral-Zettler et al., 2009). Resulting amplicons were sequenced on an Illumina MiSeq as paired-end (PE) reads of 250 bp at Michigan State University's (MSU) Research Technology Support Facility (RTSF) Genomics Core. 16S and 18S rRNA gene amplicons were analyzed with Mothur v.1.33.3 (Kozich et al., 2013) using the Silva v119 database (Quast et al., 2013). Briefly, SSU 16S rRNA gene sequences were assembled into contigs and discarded if the contig had any ambiguous base pairs, possessed repeats greater than 8 bp, or were greater than 253 bp in length. Contigs were aligned using the Silva rRNA v119 database, checked for chimeras using UCHIME (Edgar et al., 2011), and classified using the Greengenes rRNA May 2013 database (DeSantis et al., 2003, 2006). Contigs classifying to chloroplast, eukaryotes, mitochondria, or "unknown" affinities were removed from the data and the remaining contigs were clustered into operational taxonomic units (OTUs) using a 0.03 dissimilarity threshold (OTU0.03).

Due to sequencing error, 18S rRNA reverse sequences were shorter than expected, causing little to no overlap of the paired end contigs. Therefore, only forward read sequences were used in downstream analyses (mean sequence length = 250 bp). Those classified as "unknown," Archaea, or Bacteria were removed from the data and remaining contigs were clustered into OTUs using a 0.03 dissimilarity threshold (OTU0.03).

Data analyses of OTUs was done using the R statistical environment v3.2.1 (R Development Core Team, 2008), within the package PhyloSeq (McMurdie and Holmes, 2013). For estimating alpha-diversity, the filtered OTUs were used to calculate species richness using the "estimate\_richness" command within PhyloSeq, which plots Simpson, Chao1, and Shannon diversity (McMurdie and Holmes, 2013, and references therein). After alpha diversity calculations were completed, potentially erroneous rare OTUs, those without at least a total of two sequences in two or more samples, were discarded. The amplicon reads were normalized using the package DESeq2 (Love et al., 2014) following the general procedure for normalization using a variance stabilizing transformation (see Supplemental Materials for all R code used). DESeq2 normalized reads were used for all downstream analyses. For 16S rRNA sequences, beta-diversity between samples was examined using Bray-Curtis distances and ordinated using non-metric multidimensional scaling (NMDS). For 18S rRNA sequences, the relative abundance counts were converted to a presence/absence matrix (due to the potential for eukaryotic organisms to be pluricellular) and beta diversity was calculated by generating Jaccard indices. Analysis of similarity (ANOSIM) was used to test the significance of differences between groups of samples (e.g., WP vs. AP) of the NMDS analyses. Correlation between measured geochemical and macronutrient (defined as TOC, TN, NO<sup>3</sup> <sup>−</sup>, NH<sup>4</sup> <sup>+</sup>, S, Fe, P) data and the beta-diversity data was investigated in R with the envfit function (Oksanen et al., 2013). In addition, environmental variables (pH, TOC, TN, NO<sup>3</sup> −, NH<sup>4</sup> <sup>+</sup>, S, P, Fe, Si, δ <sup>15</sup>N, and δ <sup>13</sup>Corg) were tested for significance when compared to the axes of NDMS plots compared to betadiversity variables to examine patterns between geochemical data and beta-diversity plots. Further, correlations between geochemical parameters and relative abundance of OTUs were statistically evaluated with Spearman's rank-order correlation (SPSS).

Raw rRNA reads (16S and 18S) have been submitted to the European Nucleotide Archive (study accession number: PRJEB11496 and PRJEB11497, respectively). A table with OTU relative abundances can be found in the Supplementary Materials.

#### RESULTS AND DISCUSSION

#### Relating Organic Matter (OM) and Nutrients to Benthic Diversity

This study shows community structure in benthic sediments is correlated to nutrient content and also suggests a possible coupling between pelagic and benthic communities. We collected 24 benthic samples spanning a 5500 km region of Antarctica that includes Western Antarctica (WA) and the Antarctic Peninsula (AP; **Figure 1**). Sediment macronutrients (TOC, TN, NO<sup>3</sup> −, NH<sup>4</sup> <sup>+</sup>, S, Fe, P) were differentiated based on geographic region, with the general trend that samples from AP had relatively higher TOC, TN, and NH<sup>4</sup> <sup>+</sup> compared to WA (Figure S1). Stable carbon and nitrogen isotopes were collected from sediments to provide insight into the provenance of the organic matter. The δ <sup>13</sup>Corg-values found in the sediments ranged from −27.5 to −22.2‰ (average −24.6‰) and the δ <sup>15</sup>N-values range from 1.2 to 4.2‰ (combined average 3.0‰) (Figure S2). Other studies have shown phytoplankton and phytodetritus have δ <sup>15</sup>N-values from 3.2 to 7.9 and δ <sup>13</sup>Corg-values from −14.94 to −33.93 (Meyers, 1994; Cloern et al., 2002; Mincks et al., 2008). As the ranges of these data is large, it is difficult to determine exact sources, however, the data collected here does suggest phytoplankton as a possible source of organic material to sediments. Samples from WA had, on average, significantly lower δ <sup>13</sup>Corg-values than those from the AP (−25.5 and −23.4, respectively, t-test two-tailed P < 0.0001), suggesting WA samples had relatively more recalcitrant carbon. Overall, sediment geochemistry suggests organic matter predominantly sourced by phytoplankton deposition and degradation; however, samples from the AP had relatively higher quantities of macronutrients and more labile carbon. Since the AP is generally warmer than WA (Barnes et al., 2006), the possibility of a higher melting process could bring more nutrients into the water, which is one explanation for this variation. In addition, warmer temperatures could favor microbial activities and the remineralization of nutrients.

The 16S rRNA amplicon dataset included a total of 11,380 OTUs following filtering (initially 10,894,711 raw sequencing reads) and the 18S rRNA gene data generated a total of 4691 OTUs (initially 8,161,768 raw reads). Both 16S and 18S rRNA gene beta diversity ordinations showed communities strongly segregated according to the WA and AP collection regions and sampling time (**Figure 2A**, ANOSIM R = 0.993, P = 0.001 and **Figure 2B** ANOSIM R = 0.989, P = 0.001), driven by differences in macronutrients and organic matter (**Figures 2A,B**). The partitioning of diversity based on geographic region is also similar to the variation seen with the macronutrient data. Jamieson et al. (2013) examined the impact of chlorophyll content on Antarctica sediment communities and found few differences in bacterial diversity between sites with different chlorophyll content. However, the %N and TOC were similar between sampling sites, leading the authors to hypothesize that the minor variations found between them may be related to organic matter quality in addition to quantity. Quality of carbon might be a stronger driver in the present study as the samples from AP have relatively more liable carbon.

Prokaryotic communities from WA had on average >1.4x higher richness (Chao1 standard error range 13,075.6–26, 868.7) than those from the AP (Chao1.se range 9051.0–25,498.8, Figure S3A), potentially due to a large number of low abundance (less than two reads) OTUs in the WA samples. Previous 16S rRNA gene studies have calculated Chao1-values between 360 and 899 in the Southern Ocean (Jamieson et al., 2013), 1166 in surface sediment from the Ross Sea (Carr et al., 2013), and 2217 (OTUs calculated at 98% identity) in the Polar Front region (Ruff et al., 2014). Though some variation in richness estimations is seen, this can be attributed to the advent of improved next generation sequencing technologies (i.e., Illumina MiSeq) that has greatly increased the number of sequences garnered from amplicon studies and therefore increased the number of observed OTUs.

The driving force behind diversity and nutrient differences found in WA and AP is likely to be related to multiple, intertwined parameters. Taken together, geochemical and microbial diversity data here suggests a combination of the state (e.g., liable vs. recalcitrant) of organic matter and relative nutrient

concentration influences Antarctic sediment communities. One possible explanation for the differences observed between WA and AP is that they were collected in two different austral summers. Both regions were sampled in the Astral Summer, during which particulate organic matter deposition begins and increases in the following months (Smith et al., 2008, 2012). However, organic matter deposition in Western Antarctica and the Antarctic Peninsula can be highly variable (Ducklow et al., 2006; Smith et al., 2006, 2008, 2012; Fragoso and Smith, 2012) and related to numerous factors such as temperature, sedimentation, ice cover, currents, and phytoplankton blooms (Arrigo et al., 1998; Ducklow et al., 2006; Smith et al., 2006).

#### Shared Community Members across a 5500 km Portion of Western Antarctica

Proteobacteria dominated the 16S rRNA sequenced communities of all sites, with major contributions (∼15.5% of the sequenced community) from Crenarchaeota, primarily Nitrosopumulis-type Thaumarchaeota (these are classified as Phylum Crenarchaeota by the Greengenes database). Other OTUs were represented by organisms from the phyla Bacteroidetes, Verrucomicrobia, Planctomycetes, Actinobacteria, Acidobacteria, and Gemmatimonadetes (**Figures 3A**, **4**), and minor contributions (phyla with <0.52% relative abundance) from dozens of others (Figure S4). Relative abundances of Thaumarchaeota, Planctomycetes, Acidobacteria, and Gemmatimonadetes decreases along the WA to AP transit, whereas Bacteriodetes and Actinobacteria showed the opposite trend (**Figure 3A**, statistical differences calculated with a t-test, two-tailed P < 0.0001 for all mentioned other than Actinobacteria, P = 0.0010). Notable class level distinctions showed a relative abundance of Gamma- and Delta-proteobacteria across all samples, and an increase in Cytophagia of Bacteriodetes in the AP (**Figure 4**). Overall, the relative abundant phyla identified in these sediments have been seen in other studies on Antarctic sediments. Specifically, Ruff et al. (2014) documented sediment that were relatively abundant with Proteobacteria and various other studies (Bowman and McCuaig, 2003; Powell et al., 2003; Baldi et al., 2010) all found Gamma- and Delta-proteobacteria to be relatively dominant community members. In addition, the taxa documented in this study have also been identified as relatively abundant in marine sediments in general (Zinger et al., 2011). Notably, the 16S primer set used here has recently been shown to overestimate Gammaproteobacteria and under represent Alphaproteobacteria (see Parada et al., 2015), which could impact these data.

In general, the sequenced 16S rRNA communities of all benthic sediments had similar dominant OTUs. Both WA and AP shared three of the top five most abundant OTUs. OTU0000001, classified to the family of Piscirickettsiaceae, was the most dominate OTU in samples from AP (Figure S5A) and its relative abundance negatively correlated with δ <sup>13</sup>Corg, NH<sup>4</sup> <sup>+</sup>, and Si (Table S2). Bacteria within the family Piscirickettsiaceae have been identified in planktonic communities (Giovannelli et al., 2012; Wang et al., 2015) and Cycloclasticus pugetii (found within the family Piscirickettsiaceae) has been shown to degrade complex carbon (Geiselbrecht et al., 1998; Kasai et al., 2002; Wang et al., 2008). OTU0000002, classified to the genus Nitrosopumilus, was extremely abundant (Figure S5A) and its relative abundance was statistically correlated with δ <sup>13</sup>Corg and Si (Table S2). All known isolates of Nitrosopumilus are lithotrophic archaea (Konneke et al., 2005, 2014; Walker et al., 2010; Qin et al., 2014), and these have been documented as dominant community members in Antarctic sediments in the Weddell Sea (Gillan and Danis, 2007), as well as in a variety of pelagic environments (Konneke et al., 2005; Labrenz et al., 2010; Baker et al., 2012). OTU0000004 (Figure S5A) is a member of the OM60/NOR5 clade and its relative abundance correlated with pH, δ <sup>13</sup>Corg, NH<sup>4</sup> <sup>+</sup>, and Si (Table S2). OM60/NOR5 clade members are known inhabitants of surface water and coastal sediments (Connon and Giovannoni, 2002; Yan et al., 2009; Spring et al., 2013; Sharma et al., 2014). A recent study has also implicated members of this clade as key in degrading phytoplankton-derived organic matter in sediments of the Drake Passage at the Antarctic Polar Front (Ruff et al., 2014). The correlation of the Piscirickettsiaceae and OM60/NOR5 clade OTU to carbon isotope values and carbon

quality might be related to the ability of each organism to metabolize carbon.

Overwhelmingly, the dominant organisms were members of the SAR (Stramenopiles, Alveolates, Rhizaria) supergroup (Burki et al., 2007), with some exceptions in WA (**Figure 3B**). Members of SAR have been found in marine environments, including the Antarctic water column (Luria et al., 2014) and sediments (Habura et al., 2004; Pawlowski et al., 2011). Other notable groups documented in the sediments were Opisthokonts and Haptophytes. Alpha diversity calculations show richness (Chao1) and evenness (Shannon and Simpson) values as being higher in WA then AP (Figure S3B).

# Nutrient Quantity and Quality Drive Relative Abundance of Dominant Taxa

In general, the relative abundance of Thaumarchaeota was higher in samples from WA (**Figure 4**), which contained fewer organics and nutrients when compared to samples from AP (Figure S1). In addition, relative abundance of the class Thaumarchaeota and an abundant Thaumarchaeota OTU (OTU000003, most abundant OTU in WA) were both inversely correlated (0.01 level, two tailed t-test) with pH, δ <sup>13</sup>Corg, NH<sup>4</sup> <sup>+</sup>, and Si (Table S2). Representatives from Thaumarchaeota have been previously reported in Antarctic marine environments (Delong et al., 1994; Murray et al., 1998; Alonso-Saez et al., 2012; Signori et al., 2014; Hernandez et al., 2015), sediments (Gillan and Danis, 2007), and soils (Ayton et al., 2010; Richter et al., 2014). Known Thaumarchaeota are lithotrophs capable of ammonium oxidation and some are known to degrade proteins (Leininger et al., 2006; Wuchter et al., 2006; Alonso-Saez et al., 2012; Luo et al., 2014). The overall low nutrient content in the sediments from WA and the statistical relationship between Thaumarchaeota, low TOC, and slightly more refractory carbon (e.g., lower δ <sup>13</sup>Corg) suggests these are conditions that allow these organisms to flourish.

The relative abundance of the phylum Bacteroidetes correlated with TOC, Si, NH<sup>4</sup> <sup>+</sup>, and <sup>δ</sup>13Corg (Table S2). Of the five most abundant OTUs classifieds as Bacteroidetes, four were positively correlated with TOC, Si, and δ13Corg (Table S2) and the other was negatively correlated with pH, TOC, Si, and δ13Corg (Table S2). Bacteria from the Bacteroidetes phylum are very diverse and have been found in numerous environments from freshwater, soils, and oceans, and are expected to play a role in degrading complex carbon (reviewed in Kirchman, 2002; Gupta, 2004; Thomas et al., 2011).

Strikingly, the bulk of the measured SAR abundance across the AP could be accounted for by a single OTU (OTU0000013), designated SAR "unclassified" by GreenGenes taxonomy (Figure S5B). NCBI BLAST of the representative sequence for OTU0000013 returned hits with a 98% identity to the phylum Cercozoan. A marine cercozoan, Cryothecomonas, has been identified as a heterotroph that can feed on diatoms (Thomsen et al., 1991; Thaler and Lovejoy, 2012). Similarly, OTU0000010, designated as a Stramenopile, was only found in the AP. Examination of the representative OTU sequence for OTU000010 returned a 100% identity score to organisms classifying as Chaetoceros, the largest genus of diatoms. Conversely, an OTU (OTU0000030), classified as Stramenopiles, Ochrophyta, had the highest mean relative abundance in WA samples (Figure S5B), although its distribution was highly variable. In contrast to bacterial and archaeal data, relative abundance of broad phylogenetic eukaryotic categories was not significantly correlated with the nutrient data, however OTU0000013 (Cercozoan) and OTU0000010 (Chaetoceros), were correlated (0.01 level two tailed t-test) with pH, TOC, Si, and δ13Corg (Table S2).

Although, it is only speculation, the relationship between Bacteroidetes and SAR OTUs with TOC and Si could suggest a trophic relationship. Our data demonstrated an increased Learman et al. Benthic Community Structure of Western Antarctica

proportion of TOC, TN, NH<sup>4</sup> <sup>+</sup>, Si, and labile carbon in the AP relative to the WA samples, suggesting a recent diatom bloom and increased deposition of these components to the sea floor. Relative abundance of the phylum Bacteroidetes and four OTUs that fall under this phylum were also strongly correlated to TOC (Table S2). One of the OTUs that positively correlated with TOC was classified to the family Flavobacteriaceae. Members of this family have been linked to organic matter degradation in Antarctic sediments (Bowman and McCuaig, 2003; Baldi et al., 2010) and the Southern Ocean, with their relative abundance positively correlated with chlorophyll (Abell and Bowman, 2005) and algal biomass (Ruff et al., 2014), and genomic and metagenomic evidence for a role in processing organic matter from algae (Bauer et al., 2006; Fernandez-Gomez et al., 2013; Williams et al., 2013). Thus, our isotopic and SSU sequence results are consistent with the hypothesis that AP sediment organics were sourced by diatom sedimentation with subsequent degradation by Bacteroidetes and Cercozoan taxa. While this is only one possible explanation, sampling may have captured the remnants of a phytoplankton bloom and the subsequent trophic interaction.

#### CONCLUSIONS

This study offers a unique look at a spatially diverse sample set covering 5500 km of Antarctic surface sediment. Analyses revealed a diverse benthic microbial community that was highly variable throughout the two cruises and geographic regions. Though the cruises were conducted during two different field seasons, one of the possible drivers of the highly variable communities could be quality and quantity of organic matter (TOC and δ <sup>13</sup>C). WA was characterized by relatively more recalcitrant carbon and had a larger influence of archaea, specifically Thaumarchaeota. Additionally, AP was characterized by relatively higher organics and had a large presence of sequences corresponding to diatoms (e.g., Chaetoceros) and taxa from the phyla Bacteroidetes and Cercozoan, which have been known to be associated with degradation of the corresponding organics from sinking particles and fecal pellets from blooms and their associated grazers. In addition, similarities were found throughout the entire sample set as three of the top five OTUs documented in the 16S rRNA sequenced communities were shared: OTU0000001 (Piscirickettsiaceae), OTU0000002 (Nitrosopumilus), and OTU0000004 (OM60/NOR5 clade).

Future variability in ice coverage, light, temperature, and food web structure could have a profound influence on the amount

#### REFERENCES


of organic carbon reaching the bottom, thus influencing the benthic community structure and their associated functions. With the continued warming of the WAIS, the amount of melt water entering coastal water around the ice shelf is predicted to increase, which impacts both coastal and open ocean water composition (Dierssen et al., 2002; Rignot et al., 2011; Pritchard et al., 2012; Shepherd et al., 2012; Depoorter et al., 2013). This increase in melt water could lead to the increase in phytoplankton blooms in these areas (Smith and Gordon, 1997; Arrigo et al., 1998; Ducklow et al., 2006; Smith et al., 2007) and, therefore, could increase organic matter transport to the sea floor. These increases of organic matter may ultimately influence communities that were once composed predominately of lithotrophic organisms, as observed in WA samples, to ones often associated with degradation of increasing organic matter, as observed in AP samples. Thus, changes to these communities in the form of their taxonomic members and resulting impacts on global nutrient cycling must continue to be studied.

#### AUTHOR CONTRIBUTIONS

DL, MH, JT, BT, and PB analyzed data and prepared figures and tables. AM, PB, and KH collected samples. All authors contributed to writing the paper.

### FUNDING

Funds through NSF Antarctic Program: AM (CMU: Award Number 1043670), KH, and SS (AU Award Number: 1043745) and from Central Michigan University Faculty Research and Creative Endeavors (FRCE) Committee and College of Science and Technology.

#### ACKNOWLEDGMENTS

We thank the crew of the RVIB Nathaniel B. Palmer and ASRV Laurence M. Gould. This is Auburn University Marine Biology Program contribution #140 and Molette Lab contribution #48. This is contribution #66 of the CMU Institute for Great Lakes Research.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.00284


isolates. Appl. Environ. Microbiol. 68, 3878–3885. doi: 10.1128/AEM.68.8.3878- 3885.2002


IBM (2013). IBM SPSS Statistics for Windows, 22 Edn. Armonk, NY: IBM Corp.


Cycloclasticus sp. P1. Environ. Microbiol. 10, 1948–1963. doi: 10.1111/j.1462- 2920.2008.01611.x


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Learman, Henson, Thrash, Temperton, Brannock, Santos, Mahon and Halanych. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# High Prevalence of Gammaproteobacteria in the Sediments of Admiralty Bay and North Bransfield Basin, Northwestern Antarctic Peninsula

Diego C. Franco<sup>1</sup> , Camila N. Signori<sup>1</sup> , Rubens T. D. Duarte<sup>2</sup> , Cristina R. Nakayama<sup>3</sup> , Lúcia S. Campos<sup>4</sup> and Vivian H. Pellizari<sup>1</sup> \*

<sup>1</sup> Departamento de Oceanografia Biológica, Instituto Oceanográfico, Universidade de São Paulo, São Paulo, Brazil, <sup>2</sup> Centro de Ciências Biológicas, Universidade Federal de Santa Catarina, Florianópolis, Brazil, <sup>3</sup> Departamento de Ciências Ambientais, Instituto de Ciências Ambientais, Químicas e Farmacêuticas, Universidade Federal de São Paulo, Diadema, Brazil, <sup>4</sup> Departamento de Zoologia, Instituto de Biologia, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil

#### Edited by:

Julie Dinasquet, University of California, San Diego, USA

#### Reviewed by:

Deric R. Learman, Central Michigan University, USA Sachia Jo Traving, University of Copenhagen, Denmark

> \*Correspondence: Vivian H. Pellizari vivianp@usp.br

#### Specialty section:

This article was submitted to Aquatic Microbiology, a section of the journal Frontiers in Microbiology

Received: 17 October 2016 Accepted: 20 January 2017 Published: 02 February 2017

#### Citation:

Franco DC, Signori CN, Duarte RTD, Nakayama CR, Campos LS and Pellizari VH (2017) High Prevalence of Gammaproteobacteria in the Sediments of Admiralty Bay and North Bransfield Basin, Northwestern Antarctic Peninsula. Front. Microbiol. 8:153. doi: 10.3389/fmicb.2017.00153 Microorganisms dominate most Antarctic marine ecosystems, in terms of biomass and taxonomic diversity, and play crucial role in ecosystem functioning due to their high metabolic plasticity. Admiralty Bay is the largest bay on King George Island (South Shetland Islands, Antarctic Peninsula) and a combination of hydrooceanographic characteristics (bathymetry, sea ice and glacier melting, seasonal entrance of water masses, turbidity, vertical fluxes) create conditions favoring organic carbon deposition on the seafloor and microbial activities. We sampled surface sediments from 15 sites across Admiralty Bay (100–502 m total depth) and the adjacent North Bransfield Basin (693–1147 m), and used the amplicon 454-sequencing of 16S rRNA gene tags to compare the bacterial composition, diversity, and microbial community structure across environmental parameters (sediment grain size, pigments and organic nutrients) between the two areas. Marine sediments had a high abundance of heterotrophic Gammaproteobacteria (92.4% and 83.8% inside and outside the bay, respectively), followed by Alphaproteobacteria (2.5 and 5.5%), Firmicutes (1.5 and 1.6%), Bacteroidetes (1.1 and 1.7%), Deltaproteobacteria (0.8 and 2.5%) and Actinobacteria (0.7 and 1.3%). Differences in alpha-diversity and bacterial community structure were found between the two areas, reflecting the physical and chemical differences in the sediments, and the organic matter input.

Keywords: marine sediments, microbial diversity, bacterial community structure, Antarctica, polar microbiology

# INTRODUCTION

Marine microbial communities in sediments play a critical role in ecosystems functioning and are the main drivers of biogeochemical cycling of carbon, nitrogen and sulfur. They constitute a huge biomass portion of the Earth, are highly taxonomic diverse, and responsible for the majority of metabolic activity in the ocean including the polar regions (Parkes et al., 2000; Azam and Malfatti, 2007; Fuhrman, 2012). Various physical and chemical parameters can affect the marine sediment communities, in particular the organic matter. Studies conducted in the sediments of the

Ross Sea (Carr et al., 2013) and the Drake Passage at the Antarctic Polar Front (Ruff et al., 2014) showed that the organic matter can increase the microbial abundance based on phospholipids and DNA sequencing, respectively. The organic matter can also influence communities composed by lithotrophic and heterotrophic microbes, associated with degradation of organics from sinking particles and fecal pellets in the Western Antarctica (Learman et al., 2016).

As consequences of climate change and rapid warming registered in the Western Antarctic Peninsula, the sea ice declines and glacial inputs increase. These conditions can favor phytoplankton and seaweed blooms (Meredith and King, 2005; Clarke et al., 2007; Montes-Hugo et al., 2009; Pearson et al., 2015), which then increase the quantity of organic matter transported from surface layers to the ocean floor (ex. Ducklow et al., 2006; Gillies et al., 2012). The quantity and type of carbon could also be altered due to the introduction of more terrestrial carbon into marine and benthic environments in polar systems (Learman et al., 2016).

Admiralty Bay is the largest bay on King George Island (South Shetlands Archipelago), located within the maritime Antarctic region. It was designated an Antarctic Specially Managed Area no. 1 (ASMA 1) and contains the Antarctic Specially Protected Area 128 (ASPA 128, former SSSI N◦ 8). Its hydrology is complex, as it receives different contributions from water masses originating in the Bransfield Strait, and also from ice melt within the bay (Szafranski and Lipski, 1982 ´ ). Depending on the bathymetry, regional water circulation, winds and seasonal regime, waters from the Bransfield Strait that penetrate the bay originate from the adjacent warm and low saline Bellingshausen Sea (normally in the summer) or the cold and saline waters of Weddell Sea (in the winter) (Gordon and Nowlin, 1978; Tokarczyk, 1987; Huneke et al., 2016). The fjord-like shape of Admiralty Bay reflects freshwater input due to a strong glacial influence and high water column turbidity caused by suspension of soft sediments (Pichlmaier et al., 2004). All this organic matter falls to the seafloor, providing organic carbon to the reservoir, which is partially consumed by the cold-adapted microorganisms from the sediments, leading to low oxygen conditions (Orcutt et al., 2011).

To understand how the contribution of the organic matter from different sources can shape the microbial community in the Admiralty Bay and surrounding areas, in this study the bacterial diversity and community structure of the sediment samples from Admiralty Bay (AB) and adjacent areas of North Bransfield Basin (NBB) were compared. The results contribute to a better understanding of the sequenced microbial community structure in this polar area, and show differences in richness and community structure between the two sites across physicalchemical characteristics of the sediments.

#### MATERIALS AND METHODS

#### Sampling Strategy

Surface sediment samples were collected along a bathymetric gradient in AB – King George Island (stations 1–9, depths ranging from 100 to 502 m) and NBB – Bransfield Strait (stations 10–15, depths 693–1,147 m), located in the Northwestern Antarctic Peninsula (**Figure 1**). Sampling was conducted by the Brazilian Navy vessel NApOc Ary Rongel during the austral summer, December 2008. In total, 15 samples of marine sediments were collected using a Mini Box Corer (MBC) at different depths. The top 5 cm of sediments were transferred to sterile Whirl-Pack sample bags (Nasco, WI, USA) and stored frozen onboard (−20◦C). Samples were shipped to the University of São Paulo (USP) after 4 months of sampling.

#### Environmental Parameters

Grain size was determined by laser diffraction (SALD 3101, Shimadzu, Japan), following the Wentworth scale (Suguio, 1973), yielding the mean and standard deviation of particle size (Folk and Ward, 1957), and classified into categories (clay, sand and silt). Concentrations of chlorophyll and phaeopigments were estimated according to Lorenzen (1967) and Plante-Cuny (1978), adapted for sediments (Gheller, 2014). Percentages of total organic carbon (TOC), total nitrogen, organic matter and carbonates were obtained using the CHNSO elemental analyzer (Elemental Combustion System 4010, Costech Analytical Technologies, USA).

#### DNA Extraction, 16S rRNA Gene Amplification and Sequencing

Genomic DNA was extracted from 0.25 g of surface sediment in quadruplicate using a PowerSoil DNA Kit (MoBio, Carlsbad, CA, USA), according to the manufacturer's instructions. Microbial 16S rRNA gene fragments were amplified using a set of primers designed by adding a 10-nucleotide barcode to the forward primer, 519F, (5<sup>0</sup> -CAGCMGCCGCGGTAATWC-3<sup>0</sup> ) and reverse primer 1068R (5<sup>0</sup> -CTGACGRCRGCCATGC-3<sup>0</sup> ) (Wang and Qian, 2009). The amplification reaction was carried out using the Accuprime pfx SuperMix (Thermo Scientific, USA) according to the manufacturer. PCR was performed with a thermal cycler (Thermo Scientific, USA) under the following conditions: 95◦C for 5 min, 26 cycles of 95◦C for 15 s, 59◦C for 30 s and 68◦C for 1 min. The PCR products were purified by using a DNA clean & concentrator kit (Zymo Research, USA). The amplicons from each sample were mixed at equimolar concentrations and then sequenced using GSFLX titanium instruments and reagents (Roche 454, Life Sciences, USA) at the Center for Advanced Technologies in Genomics (University of São Paulo, Brazil). All sequence data have been deposited in the National Center for Biotechnology Information Sequence Read Archives (SRA) under BioProject ID PRJNA335729.

#### Sequencing Data Analyses

Raw sequence reads were filtered to eliminate the effect of the random sequencing using the Mothur 454 SOP (Schloss et al., 2011). The primer and barcodes of each read were removed and trimmed. Sequences shorter than 150 nucleotides with ambiguous bases or homopolymer regions were excluded. Sequences were clustered into operational taxonomic units (OTUs) by setting a 97% similarity. OTUs

occurring once (singletons) were removed from dataset. Qualityfiltered sequences were classified using the RDP Naïve Bayesian Classifier (Wang et al., 2007). For each sample, alpha-diversity indexes (Simpson diversity index and abundance-based coverage estimators - ACE index) were calculated using Mothur (version 1.35.1), and differences in alpha-diversity estimates between groups of samples were tested using Student t-test in R (version 3.3.2). The amplicon reads were normalized using the package DESeq2 (Love et al., 2014) following the general procedure for normalization using a variance stabilization transformation. DESeq2 normalized reads were used for all downstream analyses. The number of shared OTUs between samples were visualized using ggplot2 package in R (Wickham, 2009). Beta-diversity between samples was examined using Bray-Curtis dissimilarity matrix and ordinated by non-metric multidimensional scaling (nMDS) in R, with fitting of the environmental parameters applying the envfit function from the vegan package (Oksanen et al., 2013). To test the significance of differences between groups of samples (AB vs. NBB), analysis of similarity (adonis) was used.

#### RESULTS

#### Environmental Parameters

Sediments in the area of study were mainly composed of silt and clay (representing 70.4 to 100.0% of sediment composition). In 12 of 15 samples, granulometry was dominated by silt, with the remaining three dominated by clay. The silt fraction in sediments varied from 39.1 to 64.9%, followed by clay (23.1–59.3%) and sand (0–29.5%) (**Table 1**). Stations located in the NBB had higher sand contents (6.6–20.6%) when compared to AB samples (0 – 10.1%), with the exception of station 5 (29.5%), located in front of Lange glacier.

The sediment chlorophyll concentrations varied from zero (stations 14 and 15) to 8.25 mg.m−<sup>2</sup> (station 9), and phaeopigment concentrations from 41.68 (station 1) to 84.53 mg.m−<sup>2</sup> (station 3), indicating a predominance of degraded organic matter and small quantities of recent organic matter in the sediments. Stations in AB showed higher concentrations of chlorophyll than in the NBB, indicating the presence of higher quantities of recent organic matter to be degraded in the bay. Station 13 presented the highest organic matter concentration (6.0%) and C/N ratio (8.6).

#### Community Composition

In this study, we used massively parallel signature sequencing technologies to obtain a total of 117,267 sequences (range 1000– 26178 reads per sample) from 15 sediment samples from depths varying from 100 to 1.147 m. At the phylum level, all OTUs could be classified and belonged to 22 formally described bacterial phyla and 18 candidate phyla.


The abundance analysis showed that eight phyla accounted for more than 97% of the total amplicons: Proteobacteria (89%), Firmicutes (1.5%), Bacteroidetes (1.4%), Actinobacteria (0.9%), Chloroflexi (0.7%), Planctomycetes (0.4%), Verrucomicrobia (0.3%) and Acidobacteria (0.1%) (**Figure 2**).

Considering the total of 40 classes identified, only 10 were found in all samples and accounted for 99.4% of the total tags. Based on the relative abundance, Gammaproteobacteria was the top dominant class in both sites, varying from 87.1 to 95.7% in AB and 70.5 to 91.4% in NBB. Alphaproteobacteria was the second dominant class, accounting for 0.9–5.9% in AB and 2.4–9.6% in NBB. Firmicutes (0.4–3.0% in AB and 0.4–4.0% in NBB) and Bacteroidetes (0.3–2.2% in AB and 0.6–3.4% in NBB) were also present in all sediment samples at similar rages. Other groups were more abundant in NBB in comparison with AB, such as Deltaproteobacteria (2.5 × 0.7%), Actinobacteria (1.3 × 0.7%) and Cyanobacteria (1.4 × 0.2%).

At the genus level, Psychrobacter showed high abundance in all 15 samples, varying from 79.3% (station 15) to 95.4% (station 4). Other nine abundant genera were Psychromonas (0–6.1%), Gillisia (0–4.1%), Loktanella (0–3.9%), Paenisporosarcina (0–2.4%), Bacillariophyta (0–1.8%), Carnobacterium (0– 1.4%), Planococcus (0–1.1%), Filomicrobium (0–0.8%) and Blastopirellula (0–0.7%).

#### Alpha Diversity

In general, alpha diversity values were higher in NBB when compared to AB. The number of observed OTUs per sample ranged from 36 (station 1) to 417 (station 10) (**Table 2**). The ACE index used for richness varied from 119.10 (station 1) to 1794.19 (station 11). Simpson's diversity index varied from 0.40 (station 15) to 0.84 (station 9). Stations 11, 12, 15 and 10 (descending order), located in the NBB, showed the highest richness, and stations 15 and 10 presented the highest diversity. For ACE index, it was verified a significant difference between the two sites (t-test = 2.95, p = 0.02).

#### Microbial Community Structure and the Influence of Environmental Parameters

When analyzing the shared OTUs, it was found that more OTUs were shared within samples inside the bay versus samples outside the bay (**Figure 3**), except for station 9. Stations 2–8 shared more OTUs among each other, with sharing percentages varying from 48.3 to 92.8%. Station 9, placed at the entrance of the bay, shared 7–20 OTUs with stations 1–8, and 41–53 OTUs with stations 10–15, corresponding to 11.7–33.3% and 68.3–88.3%, respectively, and showing as a transitional environment. It was also evident that stations 10–15 shared more OTUs between each other (42.6–88.3%) than with stations inside the bay (3.8–33.3%).

To determine the distribution of the bacterial community (beta-diversity), the relative abundance of the different phyla and proteobacterial classes were analyzed in relation to the sampling locations and the sediment physical-chemical parameters using nMDS. It was revealed a clear distinction between samples collected inside AB and in the NBB, where stations 10 and 15 were more distant, which was statistically supported by adonis

fmicb-08-00153 February 1, 2017 Time: 18:38 # 4

(%); Total N, Total Nitrogen (%); C/N, Total Carbon to Nitrogen Ratio.

FIGURE 2 | Taxonomic composition and relative abundance of 16S rRNA sequences based on bacterial phyla and proteobacterial classes. Samples 1–9 belong to Admiralty Bay, and 10–15 are from North Bransfield Basin.



<sup>a</sup>Results of Student t-test, testing whether samples means of Admiralty Bay vs. North Bransfield Basin are different. P-value < 0.05 is significant.

analysis (r <sup>2</sup> = 0.20, p = 0.002). Only station 9, which is a transitional sampling point located at the entrance of the bay, was more related to the second group. Gammaproteobacteria was the dominant taxon for all samples, but others influenced station 10 (Firmicutes, Actinobacteria, Alphaproteobacteria, and Deltaproteobacteria) and station 15 (Deltaproteobacteria and others).

The environmental factors associated with the sediment characteristics influenced the samples dissimilarity and the taxonomic composition. Sediments composed by clay, as well as relatively higher concentrations of phaeopigments were prevalent inside the bay and positively correlated with nMDS axis 1 (r = 0.62), whereas sediments constituted by sand and silt, and higher concentrations of total organic carbon and C/N ratio were predominant outside the bay (NBB), and negatively correlated with nMDS axis 1 (r = −0.98 for both parameters) (**Table 1**, **Figure 4**).

# DISCUSSION

fmicb-08-00153 February 1, 2017 Time: 18:38 # 7

The present study compared the bacterial diversity and community structure found in marine sediments of AB and NBB. Although similar taxonomic groups were identified, mostly related to heterotrophic metabolism, their relative abundances, as well as alpha-diversity values and bacterial community structure have differed between the two sampling regions, possibly reflecting the distinct physical-chemical characteristics of the sediments (e.g., grain size, organic matter, chlorophyll, carbonate percentages, carbon and nitrogen concentrations). Additionally, hydro-oceanographic conditions such as bathymetry, regional water circulation and winds regime, continental inputs from glaciers melting, contribution of primary producers, seasonal entrance of water masses, and turbidity, might have favored the organic matter supply in the area. Indeed, a combination of physical and chemical parameters can influence the marine sediment community structure (e.g., Austen et al., 2002; Schauer et al., 2010; Zinger et al., 2011; Bienhold et al., 2012; Nguyen and Landfald, 2015; Learman et al., 2016).

The most abundant phylum (Proteobacteria) was represented by the classes Gamma-, Alpha- and Deltaproteobacteria in different proportions between the samples. This supports the results from previous studies showing that this phylum has a wide phylogenetic and phenotypic diversity in several marine and benthic environments (Sogin et al., 2006; Danovaro et al., 2010; Williams et al., 2010; Zinger et al., 2011; Lyra et al., 2013; Sun et al., 2013). Similarly, Proteobacteria has been reported from sediments of Antarctic Polar Front (Ruff et al., 2014), Ross Sea (Baldi et al., 2010) and Western Antarctic Peninsula (Learman et al., 2016).

The report of Psychrobacter as the dominant genus of Gammaproteobacteria in Antarctic sediments has not been previously reported. These heterotrophic versatile microorganisms include piezophilic and halophilic species (Kawasaki et al., 2002; Nogi et al., 2002), found in shallow and deep-sea sediments, in the water column, as part of the fish and krill microbiomes, and in association with brown macroalgae (Bozal et al., 2003; Lee et al., 2006; Dang et al., 2008; Teske et al., 2011; Tropeano et al., 2012). Most species of this genus can adapt to cold conditions, such as polar permafrost and ice, and are capable of reproducing at temperatures ranging from −10◦C to 40◦C (Rodrigues et al., 2009). They often produce low temperature-adaptive lipases (Bozal et al., 2003; Yumoto et al., 2003) and play essential roles in fat decomposition reactions (Li et al., 2013). The polar environment therefore constitutes an ecological niche for Psychrobacter strains (Bozal et al., 2003).

Alphaproteobacterial sequences affiliated with Loktanella, Litorimicrobium and Hoefla, which play a common role in the nitrogen cycle, were present in all samples, although in higher relative abundances outside the bay. They are able to reduce nitrate to nitrite, degrade aromatic compounds, and oxidize sulfur, ammonia, carbon monoxide, iron and manganese (Van Trappen et al., 2004a; Ivanova et al., 2005; Jin et al., 2011; Jung et al., 2013).

Flavobacteria, generally a major clade of Bacteroidetes in marine environments, was the third most abundant class. The clade was mainly represented by the genus Gillisia, specifically in samples 15 (4.1%) and 13 (3.4%), both located in NBB. Several strains have previously been isolated from marine and polar environments, such as Antarctic lakes (Van Trappen et al., 2004b). These heterotrophs are especially important, as they break down complex organic matter using exoenzymes to degrade algal cells and algal-detrital particles (Kirchman, 2002; Gómez-Pereira et al., 2012; Teeling et al., 2012). In fact, the presence of such distinct taxa in particular at stations placed outside the bay, indicated by alpha-diversity and community composition, helps to explain the differences in bacterial community between the two study areas.

The microbial community composition found in AB and NBB showed the prevalence of microorganisms related to heterotrophic metabolism. High relative abundance of heterotrophs can be explained by the high input of organic matter from different sources. The phytodetritus in deep sediments are allochthonous and can be derived from microphytobenthos, abundant in shallow areas of AB, or deposition of phytoplankton and macroalgae fragments (Gheller, 2014). Macroalgae can cover about 30% of the seabed surface of AB, and produce ca. 74,000 tons of wet biomass (Zielinski, 1990; Nedzarek and Rakusa-Suszczewski, 2004), resulting in substrates that favor the presence of these heterotrophs.

Phytoplankton blooms can also contribute to the transfer of organic matter to the seafloor. They normally occur in the study area during the austral summer due to the advection of continental input of ice melts (Nedzarek, 2008) and coastal upwelling (Brandini and Rebello, 1994; Schloss et al., 2002). This rapidly transported phytoplankton-derived organic carbon will be recycled by heterotrophic microbes fueling a diverse microbial community in deep sediments, as previously shown by Ruff et al. (2014). Moreover, the system of currents in the study area, influenced by the warmer and less saline waters from Bellingshausen Sea and the cold and saline waters from Weddell, may also indirectly influence microbial composition. The currents within the bay are more intense when compared to those in the inlets (Gordon and Nowlin, 1978; Huneke et al., 2016), thus enabling the transport, deposition and homogenization of the organic matter that drives the heterotrophic bacterial community.

# CONCLUSION

The next generation DNA sequencing data of 15 samples of seafloor sediments provides the first results characterizing the sediment microbial communities of the Northwestern Antarctic Peninsula, in a transect from AB to the NBB. Our study revealed high prevalence of heterotrophic gammaproteobacterial phylotypes, and differences in bacterial diversity and community structure between the two sites. A combination of conditions that favor the organic matter input, like regional water circulation and winds regime, bathymetry, continental influence from glaciers melting, inputs from primary producers of the euphotic zone or continental areas, water masses carrying nutrients, besides the grain size and sediment characteristics (organic matter, carbon and nitrogen concentrations), may contribute to shape the marine sediment communities in Antarctica.

#### AUTHOR CONTRIBUTIONS

fmicb-08-00153 February 1, 2017 Time: 18:38 # 8

DF, RD, and CS analyzed data and prepared figures and tables. All authors contributed to writing the paper.

#### REFERENCES


#### ACKNOWLEDGMENTS

We thank the Brazilian Antarctic Program (PROANTAR). We also thank Dr. Paula F. Gheller and Dr. Thaïs Corbisier for sharing the environmental data. This work was supported by the Brazilian National Council for Scientific and Technological Development – CNPq (MABIREH/IPY/CAML Project n. 520293/2006-1). DF was supported by the CAPES-Master's fellowship.


western Antarctic Peninsula. Science 323, 1470–1473. doi: 10.1126/science.116 4533


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Franco, Signori, Duarte, Nakayama, Campos and Pellizari. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.