# CHANGING PLANKTON COMMUNITIES: CAUSES, EFFECTS AND CONSEQUENCES

EDITED BY : Kristian Spilling, Letizia Tedesco, Riina Klais and Kalle Olli PUBLISHED IN : Frontiers in Marine Science

#### Frontiers Copyright Statement

© Copyright 2007-2019 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 978-2-88963-042-4 DOI 10.3389/978-2-88963-042-4

#### 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

Frontiers in Marine Science 1 September 2019 | Changing Plankton Communities

# CHANGING PLANKTON COMMUNITIES: CAUSES, EFFECTS AND CONSEQUENCES

Topic Editors: Kristian Spilling, Finnish Environment Institute (SYKE), University of Helsinki, Finland and University of Agder, Norway Letizia Tedesco, Finnish Environment Institute (SYKE), Finland Riina Klais, University of Tartu, Estonia Kalle Olli, University of Tartu and Estonian University of Life Sciences, Estonia

Marine ecosystems are changing at an unprecedented rate. In addition to the direct effects of e.g. warming surface temperatures, the environmental changes also cause shifts in plankton communities. Plankton makes up the base of the marine food web and plays a pivotal role in global biogeochemical cycles. Any shifts in the plankton community composition could have drastic consequences for marine ecosystem functioning. This Research Topic focuses on causes, effects and consequences of such shifts in the plankton community structure.

Citation: Spilling, K., Tedesco, L., Klais, R., Olli, K., eds. (2019). Changing Plankton Communities: Causes, Effects and Consequences. Lausanne: Frontiers Media. doi: 10.3389/978-2-88963-042-4

# Table of Contents

#### *05 Editorial: Changing Plankton Communities: Causes, Effects and Consequences*

Kristian Spilling, Letizia Tedesco, Riina Klais and Kalle Olli

#### I. BACTERIOPLANKTON


#### II. PHYTOPLANKTON

*37 Diatom Transcriptional and Physiological Responses to Changes in Iron Bioavailability Across Ocean Provinces*

Natalie R. Cohen, Kelsey A. Ellis, Robert H. Lampe, Heather McNair, Benjamin S. Twining, Maria T. Maldonado, Mark A. Brzezinski, Fedor I. Kuzminov, Kimberlee Thamatrakoln, Claire P. Till, Kenneth W. Bruland, William G. Sunda, Sibel Bargu and Adrian Marchetti

*57 Corrigendum: Diatom Transcriptional and Physiological Responses to Changes in Iron Bioavailability Across Ocean Provinces*

Natalie R. Cohen, Kelsey A. Ellis, Robert H. Lampe, Heather McNair, Benjamin S. Twining, Maria T. Maldonado, Mark A. Brzezinski, Fedor I. Kuzminov, Kimberlee Thamatrakoln, Claire P. Till, Kenneth W. Bruland, William G. Sunda, Sibel Bargu and Adrian Marchetti

*58 Algal Hot Spots in a Changing Arctic Ocean: Sea-Ice Ridges and the Snow-Ice Interface*

Mar Fernández-Méndez, Lasse M. Olsen, Hanna M. Kauko, Amelie Meyer, Anja Rösel, Ioanna Merkouriadi, Christopher J. Mundy, Jens K. Ehn, A. Malin Johansson, Penelope M. Wagner, Åse Ervik, Brian K. Sorrell, Pedro Duarte, Anette Wold, Haakon Hop and Philipp Assmy


#### *109 Drivers of Cyanobacterial Blooms in a Hypertrophic Lagoon* Marco Bartoli, Mindaugas Zilius, Mariano Bresciani, Diana Vaiciute, Irma Vybernaite-Lubiene, Jolita Petkuviene, Gianmarco Giordani, Darius Daunys, Tomas Ruginis, Sara Benelli, Claudia Giardino, Paul A. Bukaveckas, Petras Zemlys, Evelina Griniene, Zita R. Gasiunaite, Jurate Lesutiene, Renata Pilkaitytė and Arturas Baziukas-Razinkovas

#### III. ZOOPLANKTON

*121 Phytoplankton Community Dynamic: A Driver for Ciliate Trophic Strategies*

Lumi Haraguchi, Hans H. Jakobsen, Nina Lundholm and Jacob Carstensen


#### IV. THEORY AND METHODS


Friederike G. Engel, Aleksandra M. Lewandowska, Sarah L. Eggers and Birte Matthiessen

# Editorial: Changing Plankton Communities: Causes, Effects and Consequences

Kristian Spilling1,2,3 \*, Letizia Tedesco<sup>1</sup> , Riina Klais <sup>4</sup> and Kalle Olli 4,5

<sup>1</sup> Finnish Environment Institute (SYKE), Helsinki, Finland, <sup>2</sup> Tvärminne Zoological Station, Faculty of Biological and Environmental Sciences, University of Helsinki, Hanko, Finland, <sup>3</sup> Faculty of Engineering and Science, University of Agder, Kristiansand, Norway, <sup>4</sup> Institute of Ecology and Earth Sciences, University of Tartu, Tartu, Estonia, <sup>5</sup> Institute of Agricultural and Environmental Sciences, Estonian University of Life Sciences, Tartu, Estonia

Keywords: phytoplankton, ciliates, zooplankton, marine bacteria, marine ecosystems, biogeochemical cycles, global change

**Editorial on the Research Topic**

#### **Changing Plankton Communities: Causes, Effects and Consequences**

Marine ecosystems are changing in response to multiple stressors such as global warming, increasing carbon dioxide (CO2) and decreasing oxygen (O2) concentrations and eutrophication of coastal waters, among others. The direct effects of these changes on plankton physiology have been studied for decades; less are known about possible effects these changes might have on the composition of plankton communities, and even less about what effects any such shift in plankton community composition will have on marine ecosystems. The plankton community makes up the base of the marine food web (i.e., primary producers, decomposers, and primary consumers) and plays a pivotal role in global biogeochemical cycles (e.g., Falkowski and Raven, 2013). Any change of the plankton community structure, driven by natural or human induced changes, may consequently have indirect effects on marine ecosystem functioning.

This Research Topic focused on causes, effects and consequences of changing composition of plankton communities. The 12 contributions to this volume include seven original research papers, one method paper, and four reviews; all touching the state-of-the-art in current plankton research, and each from a complementary angle.

Several of the original research papers deal with changing phytoplankton communities, environmental drivers and ecosystem effects. Fernández-Méndez et al. analyzed sea-ice ridges and the snow-ice interface, which are algal hotspots in the Arctic Ocean. Both sea-ice ridges and the snow-ice interface are projected to increase due to thinning of the ice, and Fernández-Méndez et al. described the algal communities, mostly dominated by different diatoms, in these habitats in the Arctic. von Scheibner et al. examined the phytoplankton and bacterioplankton response to short-term warming. Warming increased carbon availability for the bacterial community, but the ratio between bacterial and primary production was still relatively low, suggesting it is not much changed by short-term warming events. Cohen et al. described diatom transcriptional and physiological responses to changes in iron availability in the open Northeast Pacific Ocean and in the California upwelling system. They found species specific differences in gene expression to changes in nutrient availability and taxa specific strategies for coping with Fe stress. Ajani et al. investigated the realized niches of phytoplankton using a long-term data set collected off Eastern Australia. They demonstrated that the ecological niches can be dynamic and that climate change models cannot use fixed niches when forecasting the phytoplankton community composition.

#### Edited and reviewed by:

Angel Borja, Centro Tecnológico Experto en Innovación Marina y Alimentaria (AZTI), Spain

\*Correspondence: Kristian Spilling kristian.spilling@environment.fi

#### Specialty section:

This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science

Received: 22 March 2019 Accepted: 07 May 2019 Published: 28 May 2019

#### Citation:

Spilling K, Tedesco L, Klais R and Olli K (2019) Editorial: Changing Plankton Communities: Causes, Effects and Consequences. Front. Mar. Sci. 6:272. doi: 10.3389/fmars.2019.00272

There are three original research papers on zooplankton dynamics. Lips and Lips investigated the increasing importance of the mixotrophic ciliate Mesodinium rubrum in the Baltic Sea. The abundance of this species was higher in years of earlier warming and the authors suggest that it plays an important role in shaping the inorganic nutrient pools at the start of the summer (Lips and Lips). Haraguchi et al. studied the coupling between phytoplankton and ciliates in Danish waters over 2 years, and demonstrated a close coupling between these communities, suggesting top down control of the phytoplankton community by the ciliates. Karlsson and Winder examined ecosystem effects of two locally adapted populations of the filter feeding copepod Eurytemora affinis that differed in size. They demonstrated that morphologically divergent populations of the same species can perform different ecosystem functions through differences in quantitative and qualitative feeding, and by having different population response to changes in resource supply and the phytoplankton community composition.

In the method paper by Engel et al., they tested three different ways to manipulate species loss in natural phytoplankton communities. Dilution, filtration, and heat stress was used to remove rare, large and sensitive species, respectively, and this can be used as a method for non-random species manipulation in experiments. The majority of research on species loss has used the approach of random species removal, which may not be a suitable approach for studies of fragile species. The method development and standardization of approaches suggested by Engel et al. are essential for more realistic species loss modeling.

The review papers covered different aspects of plankton dynamics and trait-based approaches. Lindh and Pinhassi presented a comprehensive review of bacterioplankton communities in the Baltic Sea and environmental drivers for community changes based on field and experimental studies. Bartoli et al. reviewed the drivers of cyanobacterial blooms in the Curonian Lagoon (Baltic Sea), where cyanobacteria has benefitted from long term increase in the temperature and reduction in the inorganic N:P ratio. A comparison of the differences between freshwater and marine studies of phytoplankton traits and community assembly is presented by Weithoff and Beisner. Finally, Spilling et al. reviewed and synthesize state-of-the-art knowledge on the observed, long-term increase in dinoflagellate abundance in the Baltic Sea during spring bloom and the consequences

#### REFERENCES


**Conflict of Interest Statement:** RK was previously employed by the University of Tartu but is at present affiliated with the company EcoStat Ltd.

the shift from diatom to dinoflagellate dominance has for biogeochemical cycles.

The topics of the papers published in this Research Topic ranged from heterotrophic bacteria, phytoplankton to zooplankton and covered different marine ecosystems. The potential shift in community composition may have dramatic effects on ecosystem functioning, for example on trophic transfer, and on biogeochemical fluxes through changes in export of organic material, i.e., the biological pump. One of the key challenges for predicting changes to the plankton community is to understand the various functional groups and their niche separation in combination with individual taxa's ability to acclimate, adapt and compete in a changing environment. This trait-based community ecology of plankton has started to gain traction (Litchman and Klausmeier, 2008; Litchman et al., 2013), and is a useful framework to investigate potential effects of environmental change on plankton community structure. In order to disentangle the potential consequences of shifts in plankton communities, more empirical studies of ecological interactions and export are needed. Hence, we consider the research papers in this Research Topic will be a valuable addition to the accumulating empirical evidence of how plankton communities are modulated by natural and human induced changes and the indirect effect this has on marine ecosystems.

#### AUTHOR CONTRIBUTIONS

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

#### FUNDING

This editorial is a result of funding from the Academy of Finland (decisions no 259164, KS and LT), the Estonian Research Council (no 1574P, KO) and the EU BONUS program (grant agreement 2112932-1, LT).

#### ACKNOWLEDGMENTS

We thank all the contributing authors and reviewers to this Research Topic and the editorial team at Frontiers in Marine Sciences.

The remaining 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 © 2019 Spilling, Tedesco, Klais and Olli. 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(s) 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.

# Sensitivity of Bacterioplankton to Environmental Disturbance: A Review of Baltic Sea Field Studies and Experiments

Markus V. Lindh1,2 and Jarone Pinhassi <sup>1</sup> \*

*<sup>1</sup> Centre for Ecology and Evolution in Microbial Model Systems—EEMiS, Linnaeus University, Kalmar, Sweden, <sup>2</sup> Swedish Meteorological and Hydrological Institute, Gothenburg, Sweden*

#### Edited by:

*Kristian Spilling, Finnish Environment Institute (SYKE), Finland*

#### Reviewed by:

*Hermanni Kaartokallio, Finnish Environment Institute (SYKE), Finland Stilianos Fodelianakis, King Abdullah University of Science and Technology, Saudi Arabia*

> \*Correspondence: *Jarone Pinhassi jarone.pinhassi@lnu.se*

#### Specialty section:

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

Received: *03 May 2018* Accepted: *19 September 2018* Published: *10 October 2018*

#### Citation:

*Lindh MV and Pinhassi J (2018) Sensitivity of Bacterioplankton to Environmental Disturbance: A Review of Baltic Sea Field Studies and Experiments. Front. Mar. Sci. 5:361. doi: 10.3389/fmars.2018.00361* Bacterioplankton communities regulate energy and matter fluxes fundamental to all aquatic life. The Baltic Sea offers an outstanding ecosystem for interpreting causes and consequences of bacterioplankton community composition shifts resulting from environmental disturbance. Yet, a systematic synthesis of the composition of Baltic Sea bacterioplankton and their responses to natural or human-induced environmental perturbations is lacking. We review current research on Baltic Sea bacterioplankton dynamics *in situ* (48 articles) and in laboratory experiments (38 articles) carried out at a variety of spatiotemporal scales. *In situ* studies indicate that the salinity gradient sets the boundaries for bacterioplankton composition, whereas, regional environmental conditions at a within-basin scale, including the level of hypoxia and phytoplankton succession stages, may significantly tune the composition of bacterial communities. Also the experiments show that Baltic Sea bacteria are highly responsive to environmental conditions, with general influences of e.g. salinity, temperature and nutrients. Importantly, nine out of ten experiments that measured both bacterial community composition and some metabolic activities showed empirical support for the *sensitivity* scenario of bacteria—i.e., that environmental disturbance caused concomitant change in both community composition and community functioning. The lack of studies empirically testing the *resilience* scenario, i.e., experimental studies that incorporate the long-term temporal dimension, precludes conclusions about the potential prevalence of resilience of Baltic Sea bacterioplankton. We also outline outstanding questions emphasizing promising applications in incorporating bacterioplankton community dynamics into biogeochemical and food-web models and the lack of knowledge for deep-sea assemblages, particularly bacterioplankton structure-function relationships. This review emphasizes that bacterioplankton communities rapidly respond to natural and predicted human-induced environmental disturbance by altering their composition and metabolic activity. Unless bacterioplankton are resilient, such changes could have severe consequences for the regulation of microbial ecosystem services.

Keywords: bacterial diversity, archaea, 16S rRNA, metabolic activity, ecosystem functioning, climate change

# BACTERIOPLANKTON COMMUNITY AND FUNCTIONAL DYNAMICS

Bacterioplankton communities have a remarkable capability in responding to environmental disturbances, such as changes in temperature, salinity and nutrients (Allison and Martiny, 2008; Logares et al., 2013; Cram et al., 2015; Salazar et al., 2016). Rapid responses are made possible thanks to relatively short generation times, and involve both adjustments in metabolic activity and restructuring of community composition (**Box 1**; Allison and Martiny, 2008; Brettar et al., 2016). There is now ample evidence that bacterioplankton activity and community composition play central roles in regulating biogeochemical cycles of elements, with particular focus on carbon in the form of dissolved organic matter (DOM) and dissolved organic carbon (DOC; Azam and Malfatti, 2007; Falkowski et al., 2008).

Given the importance of bacterioplankton-driven cycling of carbon, a fundamental question in microbial ecology is whether shifts in bacterioplankton community composition resulting from changes in environmental conditions also lead to changes in ecosystem functioning (Loreau, 2000, 2004; Langenheder et al., 2010; Comte and Del Giorgio, 2011; Beier et al., 2017). Bacterioplankton communities could, in theory, respond to environmental disturbances according to three scenarios termed sensitivity, resistance, and resilience [**Box 2**; sensu Allison and Martiny (2008); Shade et al. (2012)]. Each of these scenarios have distinctly different implications regarding how we interpret the consequences of environmental disturbances on the linkages between shifts in community composition and changes in community functioning. Ecosystem stability is linked with a community's response to disturbance. The stability depends on the sensitivity, resistance (insensitivity to disturbance), and the resilience (ability to return to original condition after disturbance) of the ecosystem (Shade et al., 2012). A disturbance is an event that can influence the community in pulses (discrete short-term events) or presses (long-term or continuous events). Shade et al. (2012) carried out a literature review to investigate bacterioplankton community responses to such disturbances in a variety of habitats. The authors identified that bacterioplankton communities are, in general, sensitive to disturbances but highlighted that empirical tests of resilience, i.e., studies that explicitly and comprehensively included the temporal dimension, were lacking.

# THE BALTIC SEA AND PREDICTED CLIMATE CHANGE

The shallow brackish Baltic Sea system, with an average depth of ∼54 m, varies in both hydrology and physicochemical features; it consists of a 2,000 km long salinity gradient ranging from truly marine to freshwater conditions through several basins of which three are major (the Baltic Sea Proper, the Bothnian Sea, and the Bothnian Bay; Omstedt et al., 2014). In addition to strong shifts in salinity, Baltic Sea basins are affected by different magnitudes of river discharges, transferring freshwater, and terrestrial DOM, socalled allochthonous DOM (i.e., DOM transported into the sea from terrestrial sources; Kritzberg et al., 2004), to coastal waters, with seasonal variation (Omstedt et al., 2014; Rowe et al., 2018). The Baltic Sea is periodically affected by seasonal phytoplankton blooms, including typical diatom/dinoflagellate spring blooms and massive blooms of cyanobacteria in summer (Legrand et al., 2015), producing so-called autochthonous DOM (i.e., DOM produced in situ by e.g., phytoplankton; Kritzberg et al., 2004).

Projections of human-induced climate change in aquatic environments, with global temperature increases of 1.4–5.8◦C and a global atmospheric CO<sup>2</sup> increase of 400 atm, resulting in lower pH by ∼0.4 units until 2,100, implicate that ecosystem changes of unparalleled extent will occur (Brettar et al., 2016). In the Baltic Sea region the Swedish meteorological and hydrological institute (SMHI) projects increased precipitation by up to 48% until 2,100, leading to lower salinities and increased output of allochthonous matter from river discharge (Meier, 2006). Moreover, cyanobacterial blooms are increasing in magnitude due to human-driven climate change and the hydrography of this semi-enclosed system (Omstedt et al., 2014; Legrand et al., 2015). Eutrophication brings excess nutrients to the microbial food web, stimulating phytoplankton blooms that in turn influence the bacteria, resulting in increased community respiration rates through the degradation of phytoplankton DOM that sinks and ultimately causes bottom water and sediment hypoxia (Tamelander et al., 2017). Decreasing oxygen levels (<2 mg ml−<sup>1</sup> O2) at the seafloor and in sediments of the Baltic Sea have been documented in the last 100 years and decreases are predicted to intensify in the future (Carstensen et al., 2014). Taken together, future selective forces in the marine environment will include, among others, increased sea surface temperatures, lower pH, eutrophication, hypoxia, increased allochthonous carbon inputs, decreased salinity, and massive cyanobacterial blooms (Andersson et al., 2015). Thus, the Baltic Sea offers a unique study system to investigate, in depth, both the causes and consequences of shifts in bacterioplankton community composition and functioning responding to environmental perturbations.

#### OVERVIEW OF BALTIC SEA BACTERIOPLANKTON STUDIES

We identified a total of 86 articles carrying out field studies and/or experiments focusing on bacterioplankton community composition in the Baltic Sea, as summarized in **Table 1** and **Figure 1**. Studies performed in the Skagerrak and Kattegatt seas were examined (**Supplementary Table 1**), but were not included in the main analyses. Among the Baltic Sea studies, 48 were categorized as research performed in situ and 38 were experimental studies at different scales (micro- or mesocosms; **Table 1**). Twenty-six articles included samples from deepwaters/sediments. Among all articles, one third (n = 29) included some form of activity measurement and among the in situ studies less than one fifth (n = 9) measured metabolic activity. A majority of the articles focus on spatial distributions for in situ data (n = 27; 56% of all in situ studies) and on the effect of nutrient additions for experimental data (n = 22; 58% of all experimental studies; **Table 1**). The conducted studies have been performed throughout the Baltic Sea and include empirical evidence of shifts

#### BOX 1 | Determining bacterioplankton community composition.

Marine bacteria and archaea, collectively known as *bacterioplankton*, are a fundamental part of the planktonic food-web and imperative for ecosystem functioning (Falkowski et al., 2008). *Bacterioplankton community composition* refers to the taxonomic identity of organisms and their frequency distribution in the ecosystem. However, phenotypic identification of bacterioplankton is problematic since only a small fraction of all bacterioplankton are easily cultivable and these typically do not mirror the complete bacterioplankton diversity (Pedros-Alio, 2006; Hagström et al., 2017). Microbial ecologists therefore use culture-independent genetic identification techniques to differentiate bacterioplankton taxa (see e.g., Hugerth et al., 2015; Sunagawa et al., 2015; Beier et al., 2017; Celepli et al., 2017). In general, identification of individual bacterioplankton taxa is done by sequencing of the 16S rRNA gene, ITS (Internal Transcribed Spacer) or similar genetic fragments, processing raw *sequence reads* using bioinformatic methods (see e.g., Edgar, 2013; Callahan et al., 2015) and clustering these into specific phylotypes or *operational taxonomic units* (OTUs), followed by taxon delineation at a specific sequence identity threshold (typically 97–99% sequence identity; Hugerth and Andersson, 2017). Taxonomic annotation of these OTUs is then performed by matching a type- or centroid sequence of the OTU cluster against a database with taxonomic information (see e.g., Quast et al., 2013).

Technical advances in the last decade in the form of high-throughput sequencing have increased the sequencing resolution and thereby the detection levels of individual OTUs by several orders of magnitude at rapidly decreasing costs (Poisot et al., 2013; Hugerth and Andersson, 2017). As a result, the field of microbial ecology has advanced from primarily describing major bacterioplankton groups or a few dominant populations to resolving the distribution of thousands of populations over different temporal and spatial scales. This makes bacterioplankton attractive also for monitoring and thus important for stakeholders and marine management.

#### BOX 2 | Bacterioplankton community responses to environmental disturbance

Below follows a brief description of three theoretical responses that bacterioplankton assemblages may undertake in the light of natural or human-induced changes in environmental conditions, *sensu* Allison and Martiny (2008), (Shade et al., 2012). The figure is laid so that it emphasizes the temporal dimension on the x-axis in the *resilience* scenario.

Sensitivity—when community composition is altered by environmental disturbance and the resulting community functioning (*f*2) is typically different from the initial functioning (*f*1).

Resistance—when community composition is not altered by environmental disturbance and community functioning (*f*2) may be different from, or the same as the initial *f*1.

Resilience—when community composition is initially altered by environmental disturbance but returns to the original composition and *f*2 is typically temporally different from the initial *f*1 but community functioning over time "returns to" *f*3 similar to the initial *f*1.

in bacterioplankton community composition in different basins (**Figure 1**). It is noteworthy that detailed temporal studies are typically limited to two sites in the Baltic Sea Proper (i.e., the Landsort Deep in the central Baltic Sea and at the Linnaeus Microbial Observatory off the coast of the island Öland in the Baltic Sea Proper) and that many deep-water/sediment samples have primarily been obtained from only two locations in the central Baltic Sea Proper (the Landsort and Gotland deeps in the central Baltic Sea; n = 10; 40% of all deep-water/sediment studies; **Figure 1**).

#### TEMPORAL AND SPATIAL VARIABILITY IN BACTERIOPLANKTON

Extended temporal studies are scarce in the Baltic Sea, but there are four published reports of bacterioplankton dynamics extending over ≥1 year at semi- to high-resolution (Pinhassi and Hagström, 2000; Riemann et al., 2008; Andersson et al., 2010; Lindh et al., 2015c). They all show pronounced shifts in bacterioplankton community composition with fairly distinct spring, summer and autumn communities. In general, Bacteroidetes dominate in spring during the diatom and dinoflagellate bloom. In summer, particular populations of e.g., Verrucomicrobia increase in abundance, associated with pronounced blooms of picocyanobacterial populations and filamentous Cyanobacteria. In autumn, actinobacterial populations proliferate, potentially following changes in temperature, mixing of the water column, and/or autumn phytoplankton blooms. Changes in temperature over the yearly cycle typically explain around 0.45 (Pearson r) of total variation in community composition (Andersson et al., 2010; Lindh et al., 2015c).

There have been three major transects studies of bacterioplankton community composition across the entire salinity gradient of the Baltic Sea, showing that changes in salinity is the major driver of the distribution of bacterial populations (OTUs) at large spatial scales in the Baltic Sea (http://www.noaa.gov).

using the package "marmap" (Pante and Simon-Bouhet, 2013) in R V.3.4.4 (R Core Development Team, 2017) and the bathymetry data was retrieved from NOAA

(Herlemann et al., 2011, 2016; Dupont et al., 2014; Larsson et al., 2014; Celepli et al., 2017). Notably, salinity explained almost all of the observed variation in bacterioplankton community composition with r values ranging from 0.78 to 0.97. In contrast, in a transect across one particular basin, the Baltic Sea Proper, Bunse et al. (2016a) found that, instead of salinity, the advancing phytoplankton spring bloom had a substantial influence in structuring the bacterial community (i.e., the distribution of specific OTUs). The authors emphasized how interactions between bacterioplankton and phytoplankton populations may influence carbon cycling to higher trophic levels thus extending the importance of bacterioplankton dynamics for ecosystem services. In addition, there have been several transects performed at high-resolution spatial scales of which two locations are particularly well-studied; the western Gotland Sea (Bertos-Fortis et al., 2016; Lindh et al., 2016, 2017) and the Gulf of Finland (Laas et al., 2014, 2015, 2016). For example, Bertos-Fortis et al. (2016) show that Cyanobacteria have plastic responses to changes in environmental conditions where opportunistic picocyanobacteria and N2-fixing cyanobacteria may be able to utilize nutrients from filamentous Cyanobacteria during the summer bloom (Bertos-Fortis et al., 2016). An approach using metacommunity theory (Leibold et al., 2004) found that local environmental conditions rather than dispersal-driven assembly is the main mechanism for structuring bacterioplankton assemblages in the Baltic Sea (Lindh et al., 2016). Similarly, bacterioplankton communities in areas such as the Pacific Ocean and East China Sea are also mainly shaped by environmental factors and to a less extent by spatial factors (Yeh et al., 2015; Lindh et al., 2018). Moreover, deep-water bacterioplankton communities in the Gulf

of Finland are structured by seasonally anoxic conditions where redox-specialized bacterioplankton populations proliferate (Laas et al., 2015). Surface seawater communities in that study, on the other hand, were largely assembled following variation in phytoplankton succession. Collectively, these studies indicate that whereas the salinity gradient sets the boundaries for bacterioplankton composition, regional environmental conditions at a within-basin scale, including the level of hypoxia and phytoplankton succession stages, may significantly tune the composition of bacterioplankton.

# EXPERIMENTAL MANIPULATION OF ENVIRONMENTAL CONDITIONS

#### Salinity

The in situ studies described above dictate that grand scale spatial distributions of particular OTUs in the Baltic Sea are driven by changes in salinity. In accordance, manipulated changes in salinity in micro- and mesocosm experiments also show that salinity can significantly shape bacterioplankton community composition (Langenheder et al., 2003; Kaartokallio et al., 2005; Sjöstedt et al., 2012b; Lindh et al., 2015a). For example, shifts in bacterioplankton community structure occurred in response to salinity changes following the experimental melting of sea ice (Kaartokallio et al., 2005). Compositional shifts occurred in response to a change in salinity and DOM conditions using chemostats, showing how rare or inactive taxa already present in the "seed bank," proliferated over time (Sjöstedt et al., 2012b). Moreover, a transplant experiment showed how transfer of bacterioplankton assemblages from ambient to changed salinity (brackish compared to near freshwater) and DOM conditions

#### TABLE 1 | Summary of Baltic Sea bacterioplankton studies in which data on community composition is available (*n* = 83).


#### Lindh and Pinhassi Sensitivity of Baltic Sea Bacterioplankton

#### TABLE 1 | Continued


#### TABLE 1 | Continued



*Yellow symbols denote in situ transects and temporally sampled stations and blue symbols denote all stations that have been sampled for use in experimental studies with empirical evidence for the effect of natural or human-induced changes in environmental conditions on bacterioplankton community dynamics. Filled circles and triangles denote samples obtained from surface waters and from deep-water/sediments, respectively. The table is sorted according to (i) Field or experimental studies, (ii) include activity measurements, and (iii) author name.*

*¶ A water source from the "Baltic Sea" denotes multiple locations across different basins.*

\**Activity measurements encompass uptake rates such as* <sup>3</sup>*H-Leucine/Thymidine and enzymatic activities.*

TABLE 1 | Continued

(allochthonous compared to autochthonous) has a major impact on composition and metabolic activity (Lindh et al., 2015a). Still, how such changes in salinity affect bacterioplankton community functioning, with potential implications for ecosystem services, remains largely unknown.

#### Temperature and pH

Seasonal variation in temperature significantly affects bacterioplankton community dynamics in the Baltic Sea (Andersson et al., 2010; Lindh et al., 2015c). Empirical evidence from micro- and mesocosm experiments highlights temperature as a major driver of compositional shifts in bacterioplankton (Muren et al., 2005; Sommer et al., 2007; Hoppe et al., 2008; von Scheibner et al., 2014; Vaquer-Sunyer et al., 2015). Typically, OTUs affiliated with Gammaproteobacteria such as Glaciecola spp. proliferate at higher temperatures (von Scheibner et al., 2017). Gammaproteobacteria also increased in relative abundance in warming experiments of bacterioplankton amended with dissolved organic nitrogen sources and incubated in higher temperatures (Vaquer-Sunyer et al., 2015). Here it is also worthwhile to mention that cell size is influenced by increasing temperatures (Sjöstedt et al., 2012a), as also observed in other marine ecosystems (Morán et al., 2015).

Evidence for the effect of ocean acidification on Baltic Sea bacterioplankton community dynamics are few (Bergen et al., 2016). Still, there are some indications of synergistic effects between increased temperature and pCO<sup>2</sup> levels (Lindh et al., 2013). This is in line with analyses from other systems suggesting that ocean acidification could affect bacterioplankton growth and physiology both directly and indirectly, e.g., by affecting DOM release and composition originating from higher trophic levels (Vega Thurber et al., 2009; Joint et al., 2011; Bunse et al., 2016b; Sala et al., 2016). Therefore, increased sea surface temperatures together with acidification due to climate change will likely affect the dynamics of particular bacterioplankton populations either alone or synergistically with potential amplification effects.

## Nutrient Inputs

The quality and composition of DOM is partly dependent on its origin, which can be either allochthonous or autochthonous (Kritzberg et al., 2004; Nagata, 2008). Since bacteria are the main contributors to the biological transformation of marine DOM, much attention has been put into uncovering the role of bacterial community composition in this biogeochemical process. In the Baltic Sea, particular opportunistic bacterioplankton populations are capable of successfully utilizing elevated concentrations of labile, low-molecular weight (LMW) compounds (Gomez-Consarnau et al., 2012). Moreover, multiple studies report important effects of allochthonous DOM on Baltic Sea bacterioplankton responses in community composition (Kisand et al., 2002; Grubisic et al., 2012; Herlemann et al., 2014). An important and intriguing future challenge is to determine the relationship between DOM composition and bacterioplankton community composition and functioning. Curiously, we found few studies reporting on the direct effects of inorganic nutrient enrichments on Baltic Sea Bacterioplankton communities (but see Tammert et al., 2012). This is perhaps surprising given the importance nitrogen and phosphorus plays in eutrophication of the Baltic Sea (Conley et al., 2009). Simple experiments with bacterioplankton assemblages and nitrogen and phosphorus additions will be a promising avenue of future research to couple changes in inorganic nutrients and DOM to shifts in bacterioplankton community structure and its relationship with Baltic Sea ecosystem function.

## Opportunistic Gammaproteobacteria—Bottle Effects or True Dynamics?

OTUs affiliated with Gammaproteobacteria often increase in relative abundance in micro- and mesocosm experiments. Such dynamics are sometimes attributed to the so-called "bottle effect" in which confinement of water causes shifts in bacterioplankton community composition and physiological rates (Fuchs et al., 2000; Massana et al., 2001; Baltar et al., 2012). Such effects are typically detected by rapidly increasing proportions of copiotrophic gammaproteobacterial populations and metabolic activity (see e.g., Gomez-Consarnau et al., 2012; Sjöstedt et al., 2012b). However, we note that particular gammaproteobacterial OTUs also peak in relative abundance in situ (Andersson et al., 2010; Lindh et al., 2015c). This pattern does not only occur in the Baltic Sea but also in the North Sea and Mediterranean Sea (Fodelianakis et al., 2014; Alonso-Saez et al., 2015; Teeling et al., 2016). These findings suggest that Gammaproteobacteria could respond faster to environmental disturbances aided by their copiotrophic lifestyle and thus potentially become dominant more frequently due to increased incidence of perturbation in situ. It remains to be determined whether such changes would significantly alter microbial-driven carbon cycling or if these Gammaproteobacteria perhaps carry functional redundancy (see e.g., Pedler et al., 2014).

#### BACTERIOPLANKTON COMMUNITY FUNCTIONING IS CORRELATED WITH COMMUNITY DYNAMICS BUT DATA IS SCARCE

In this section we provide an overview of experimental studies that statistically tested the linkage between changes in bacterioplankton community composition and metabolic activity (i.e., functioning) upon environmental forcing (disturbance; **Table 2**). A few in situ studies have shown simultaneous shifts in bacterioplankton community composition and metabolic activity (Pinhassi et al., 1997; Riemann et al., 2008; Glaubitz et al., 2009; Holmfeldt et al., 2009; Piwosz et al., 2013; Lindh et al., 2015c; Rieck et al., 2015; **Table 1**), but as these measurements can be influenced by and/or auto-correlated to a number of undetermined environmental variables they were not included in the current analysis. A majority of experimental studies performed (non-quantitative) cluster-like analyses based on betadiversity estimates that indicated shifts in composition (see e.g., Lindh et al., 2013; von Scheibner et al., 2014). Notably, most of the studies using cluster statistics methods show that functioning to some degree changes upon shift in composition. A number TABLE 2 | Summary of Baltic Sea bacterioplankton studies in which experimental data with empirical testing of the effect of natural or anthropogenically induced disturbances on total bacterioplankton community composition and functioning have been determined concomitantly (*n* = 10).


*Statistical tests of significant effects and indicated p-values are given for bacterioplankton functioning followed by community composition separated by "|". All studies consisted of samples obtained from the surface and not from the deep-water or sediments. None of the studies measured shifts in bacterioplankton community composition and functioning over a longer period of time, therefore we can only determine initial sensitivity and not resilience.*

\**Only studies with total community and with statistical testing of observed effects were included in this qualitative analysis.*

of studies focused on changes in the relative or absolute abundance of particular taxa or groups of bacterioplankton such as Sulfurimonas GD-1, Thiomicrospira denitrificans, and Glaciecola spp. and also measured metabolic activity of these populations (Brettar et al., 2006; Glaubitz et al., 2014; von Scheibner et al., 2017). Importantly, shifts in abundance of these particular populations were typically coupled to pronounced changes in metabolic activity when exposed to environmental disturbance.

Relatively few studies have carried out statistical assessments with quantitative data for changes in both the overall bacterioplankton community composition and bacterial functioning (e.g., using multivariate statistics such as Mantel's test, PERMANOVA, or similar). We analyze in detail these studies to reach a general understanding of current quantitative data. We note a lack of deep-water or sediments studies that in a strict sense target such quantitative assessments. Consequently, only studies with surface water samples were included in this analysis (**Table 2**). Nine out of ten studies with measurements of total bacterioplankton community composition and metabolic activity show empirical evidence of concomitant shifts in community composition and functioning. One study, by Herlemann et al. (2014), found no significant effects on bacterioplankton community dynamics responding to riverine DOM additions, including limited growth, nor links between composition and functioning. Thus, a majority of studies highlight bacterioplankton assemblages responding to disturbances following the sensitivity scenario (**Table 2**; **Box 2**).

The studies showing statistical support for concomitant shifts in community composition and functioning essentially investigated the effect of nutrient amendments on bacterioplankton community composition dynamics and metabolic activity (Grubisic et al., 2012; Tammert et al., 2012; Degerman et al., 2013; Lindh et al., 2015a; Vaquer-Sunyer et al., 2015, 2016; Bergen et al., 2016; Traving et al., 2017; Camarena-Gómez et al., 2018). The Bergen et al. (2016) study emphasized how temperature significantly alters community composition and heterotrophic bacterial production but that pCO2 had a more limited effect. The authors noted how these changes in bacterioplankton dynamics could be indicative of changed carbon fluxes in a future ecosystem where heterotrophy may become more important compared to today. Dinasquet et al. (2013) showed that bioavailable DOC triggered a shift in community composition and changed the abundance of specific extracellular enzymes suggesting that bacterioplankton communities can respond, following the sensitivity scenario, to varying supplies of DOC ultimately affecting heterotrophic respiration. Vaquer-Sunyer et al. (2015) also noted that nutrient inputs of dissolved organic nitrogen (DON) and changed temperature may lead to a more heterotrophic system. In a different study by Vaquer-Sunyer et al. (2016), they showed an effect of wastewater treatment plant effluent inputs on bacterioplankton dynamics, with an increase in the relative abundance of Cyanobacteria responding to the influx of wastewater with concomitant changes in metabolic activity. A recent study by Camarena-Gómez et al. (2018) investigated the relationship between phytoplankton bloom dynamics (different phytoplankton species) and bacterioplankton community function and taxonomic structure. Their work emphasized how bacterial metabolic activity and community composition were affected by changes in the phytoplankton community—in particular the divergent bacterial responses to phytoplankton dominated by diatoms compared to dinoflagellates. Amendment of a diatom dominated community resulted in a significantly changed bacterial heterotrophic production and shift in bacterioplankton community structure compared controls, while amendment of a dinoflagellate community had overall smaller impact on both community function and composition Camarena-Gómez et al., 2018). It is noteworthy how the addition of riverine DOM could result in changed ecosystem function from shifts in bacterioplankton community dynamics toward a more heterotrophic system (Degerman et al., 2013; Figueroa et al., 2016). Also in the Gulf of Finland enrichment with labile DOC changed bacterial production and led to shifts in community composition, promoting filamentous bacteria - the authors indicate that this DOC source may impact diversity, food-web structure and biogeochemical processing of carbon (Tammert et al., 2012). Only three specific studies have directly determined the correlation between shifts in community composition and metabolic activity responding to environmental perturbations (Lindh et al., 2015a; Vaquer-Sunyer et al., 2015, 2016). These studies indicated that variation in bacterioplankton community functioning is often significantly explained by changes in metabolic activity explaining up to around 0.65 of the variance (Pearson r).

Although environmental disturbances can influence bulk bacterioplankton community composition, the replacement of sensitive taxa will occur at the individual OTU level. For example, although Bergen et al. (2016) implicated limited effects on bulk community dynamics resulting from increased pCO2, there was a substantial influence on particular populations with statistical support from linear discriminant effect size (LEfSe). Traving et al. (2017) performed statistical tests (generalized linear models) to determine the coupling between specific OTUs and amendment of riverine DOM where several OTUs were significantly correlated with changed environmental conditions. Although these shifts in the relative abundance of particular OTUs were not coupled directly to metabolic activity there were concomitant significant changes in protease activity. Taken together, the nine studies emphasize the sensitivity scenario of bacterioplankton communities in Baltic Sea, where particular populations respond to different sets of environmental forcing. The causal loop diagram in **Figure 2** summarizes the significant effects of changes in environmental conditions for community composition and functioning included in this review. Overall, the studies in this section highlight how changes in environmental forcing may significantly impact pelagic remineralization of organic carbon and result in a drastic reduction in the flow of organic matter through the microbial loop.

#### OUTSTANDING QUESTIONS

Although the Baltic Sea bacterial communities are fairly well studied, research to integrate bacterioplankton dynamics and its consequences in food-web and biogeochemical models is much needed. We infer that such integration is necessary to allow predictions and counter measures of detrimental future climate change effects on microbial food webs and their ecosystem services.

in carbon cycling due to responses among bacterioplankton populations to natural and anthropogenic disturbances. Bacterioplankton community composition and functioning have been abbreviated to BCC and BCF, respectively. The figure is redrawn and modified from (Azam and Malfatti, 2007). Continuous and dashed arrows denote nutrient fluxes and changes in environmental variables with empirical evidence of significant effects on bacterioplankton community composition covered in this review, respectively. Changes in bacterioplankton community functioning, such as bacterial heterotrophic production and respiration, may occur due to increases in the availability of dissolved organic matter (DOM), resulting from, for example, increased freshwater runoff and phytoplankton blooms.

Recent work on bacterial populations and their activity at the seafloor of the Baltic Sea has shown a plethora of interesting dynamics and lifestyles including, an active community that was previously deemed unlikely (Glaubitz et al., 2009, 2014; Berg et al., 2013, 2015; Broman et al., 2017a,b; Reyes et al., 2017; Zinke et al., 2017). Detailed studies with concomitant measurements of bacterioplankton community composition and function are required in deep-waters and sediments of the Baltic Sea to determine the role of bacteria in key biogeochemical cycles (e.g., of C, N, P, and S) in general, and in relation to hypoxia in particular.

Seasonal in situ studies of bacterioplankton community dynamics have emphasized how communities return to a similar composition and functioning in specific seasons year after year (Fuhrman et al., 2006; Andersson et al., 2010; Lindh et al., 2015c). This could indicate that bacterioplankton communities are in fact resilient to recurrent pulse and press disturbances, but could also indicate that the environmental conditions over seasons are "repeated" and that the same bacterial populations are repeatedly selected for. Overall, there is a recognized need to incorporate the long-term temporal dimension when quantifying resistance and resilience of microbial communities (Shade et al., 2012; Fodelianakis et al., 2017; Liu et al., 2018). Only by including the temporal dimension can microbial ecologists fully understand the consequences of ocean change for bacterioplankton community dynamics, and we urgently need more information on changes occurring over time, preferably long-term studies. In particular, for experimental studies, there is a lack of data that empirically test the scenarios of resistance and resilience for Baltic Sea bacterioplankton communities. Nevertheless, in one study using a transplant and re-transplant experimental design (Lindh et al., 2015a), we showed how bacterioplankton community dynamics remain altered even after returning to the same environmental conditions as the original community experienced. One of the most important questions to address in future studies in the Baltic Sea will therefore be to empirically test whether bacterioplankton communities remain sensitive to an environmental disturbance or if the community can return to a similar original community structure and function following the resilience scenario. Presently, we observe that a majority of the Baltic Sea studies point toward the sensitivity scenario, and that it is necessary to await future research to make definite conclusions regarding the extent of resilience. Thus, within the time frame of 2 weeks, the studies indicate that sensitivity is more developed than resistance. Nevertheless, natural temporal changes in environmental conditions for time frames of more than a month will alter the composition and functioning of bacterioplankton communities, likely interfering with interpretations of the scenarios of sensitivity, resistance and resilience.

## SUMMARY

A majority of the studies included in this review show that bacterioplankton community composition is sensitive to changes in environmental conditions, in particular salinity and nutrient inputs, with empirical support for simultaneous changes also in community functioning. As noted also in other aquatic environments, pronounced shifts in composition and bacterial heterotrophic production occur following spring diatom and dinoflagellate blooms and summer blooms of filamentous Cyanobacteria. The sensitivity scenario and its impact on the microbial loop is highlighted in **Figure 3A**. Salinity singles out as a principal determinant of the distribution of bacterioplankton populations across basin-wide spatial scales in the Baltic Sea. Nevertheless, at regional (within-basin) scales, changes in nutrient availability and both autochthonous and allochthonous DOM influence the bacterioplankton community composition. Hence, this review provides a roadmap as to what ecological consequences climate change could have on the Baltic Sea ecosystem if bacterioplankton dynamics changes. Yet, although bacterioplankton communities are initially sensitive we know very little about the potential of resilience due to the lack of empirical studies testing this scenario (**Figure 3B**).

For quantifying the level of sensitivity or resilience of bacterioplankton communities to natural and human-driven changes in environmental conditions it is essential to investigate particular bacterioplankton populations and their responses to alternate physicochemical conditions and specific (in)organic nutrient sources, e.g., different DOM and DOC compounds.

FIGURE 3 | Conceptual diagram of two possible scenarios of Baltic Sea bacterioplankton community dynamics, i.e., *sensitivity* (A), and *resilience* (B). In both the *sensitivity* and *resilience* scenario an environmental disturbance leads to the replacement of individual bacterioplankton populations and changed community functioning with potential impact on the microbial loop and carbon cycling. The scenarios differ in that over time the replacement of individual bacterioplankton populations and changed community functioning remains in (A), but returns to similar individual bacterioplankton populations and community functioning as the original community in (B). Scenarios with empirical evidence of significant effects on bacterioplankton community composition and functioning in the present review is the impact of dissolved organic nitrogen (DON) additions and increased temperatures Bergen et al. (2016), Vaquer-Sunyer et al. (2015). However, for Baltic Sea data, the lack of long-term studies precludes conclusions regarding whether the *sensitivity* or *resilience* scenario are the most prominent—currently studies have been designed primarily to determine the extent of the initial *sensitivity* (gray filled areas).

The sensitivity scenario has important implications for the Baltic Sea ecosystem: it dictates that environmental disturbances and climate change are likely to affect processes like the microbialdriven carbon pump and/or the general microbial regulation of energy and matter fluxes. We infer that knowledge of the linkages between bacterioplankton composition and bacterial cycling of carbon and other elements under varying environmental forcing is critically important for developing action plans to sustain microbial ecosystem services invaluable for all marine life.

# AUTHOR CONTRIBUTIONS

ML and JP both participated in the study design, analysis, and writing of the article.

# FUNDING

The research was supported by the Swedish Research Council FORMAS strong research programme EcoChange, the Linnaeus University Center for Ecology and Evolution in Microbial model Systems (EEMiS), and by grants from the Swedish Research Council VR. This work resulted from the BONUS BLUEPRINT project which was supported by BONUS (Art 185), funded jointly by the EU and FORMAS.

# ACKNOWLEDGMENTS

This article was partly inspired by work done in the Ph.D. thesis by Lindh (2014), in particular the section on synthesis on bacterioplankton communities responding to natural and anthropogenically induced environmental disturbances. We also thank Carina Bunse, Mireia Bertos-Fortis, Hanna Farnelid, Catherine Legrand, Christofer Karlsson, Saraladevi Muthusamy, Federico Baltar, and Sandra Martinez-García for enthusing discussions that inspired writing this article.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars. 2018.00361/full#supplementary-material

#### REFERENCES


community composition and associated metabolism. Microbiome 5:96. doi: 10.1186/s40168-017-0311-5


production in the subarctic northern baltic sea. Microb. Ecol. 71, 789–801. doi: 10.1007/s00248-015-0714-4


emission scenarios. Climate Dyn. 27, 39–68. doi: 10.1007/s00382-006- 0124-x


gene libraries and the impact of temperature and NaCl concentration in experimental enrichment cultures. FEMS Microbiol. Ecol. 78, 428–438. doi: 10.1111/j.1574-6941.2011.01175.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 © 2018 Lindh and Pinhassi. 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(s) 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.

# Phyto- and Bacterioplankton During Early Spring Conditions in the Baltic Sea and Response to Short-Term Experimental Warming

Markus von Scheibner <sup>1</sup> , Daniel P. R. Herlemann1,2, Aleksandra M. Lewandowska<sup>3</sup> and Klaus Jürgens <sup>1</sup> \*

<sup>1</sup> Leibniz Institute for Baltic Sea Research Warnemünde, Rostock, Germany, <sup>2</sup> Center of Limnology, Estonian University of Life Sciences, Tartu, Estonia, <sup>3</sup> Tvarminne Zoological Station, University of Helsinki, Hanko, Finland

Edited by: Riina Klais,

# University of Tartu, Estonia

Reviewed by: Raquel Vaquer-Sunyer, Instituto Mediterráneo de Estudios Avanzados (IMEDEA), Spain Jelena Godrijan, Bigelow Laboratory For Ocean Sciences, United States

\*Correspondence:

Klaus Jürgens klaus.juergens@io-warnemuende.de

#### Specialty section:

This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science

Received: 10 January 2018 Accepted: 14 June 2018 Published: 13 July 2018

#### Citation:

von Scheibner M, Herlemann DPR, Lewandowska AM and Jürgens K (2018) Phyto- and Bacterioplankton During Early Spring Conditions in the Baltic Sea and Response to Short-Term Experimental Warming. Front. Mar. Sci. 5:231. doi: 10.3389/fmars.2018.00231 Predicted increases in sea surface temperatures are expected to shift the balance between autotrophic production and the heterotrophic degradation of organic matter toward a more heterotrophic system. For early phytoplankton spring blooms at low water temperature the impact of rising temperatures has been mainly investigated in mesocosm experiments, while field observations are scarce. During a Baltic Sea research cruise we examined early spring bloom conditions, characterized by low temperatures (0–3◦C), and performed on-board warming experiments to compare the responses of phyto- and bacterioplankton production to an increase in temperature. In the northern Baltic Sea, the low phytoplankton biomass indicated pre-bloom conditions. In the southern Baltic Sea, a diatom-dominated phytoplankton bloom with increased primary production (PP) occurred. Associated with this bloom were increases in bacterial production (BP) and bacterial abundance as well as shifts in bacterial community composition toward an increased proportion of Gammaproteobacteria and Bacteroidetes. However, the low BP/PP ratios (average: 1.2 ± 0.14%) indicated weak coupling between the bacterial and phytoplankton communities. Short-term warming (6 h, 1+6 ◦C) significantly enhanced PP (mean Q<sup>10</sup> 1.4) and especially BP (mean Q<sup>10</sup> 2.3). Hence, the higher water temperature increased both carbon flow into the bacterial community and bacterial processing of organic matter, thereby confirming previous experimental studies. By contrast, BP/PP ratios remained relatively low after warming (average: 1.7 ± 0.5%), unlike in previous mesocosm experiments performed at comparable temperatures and with similar plankton communities. Overall, these results imply that bacterial activities are suppressed during early phytoplankton blooms at low temperatures in the Baltic Sea and are not substantially altered by short-term warming events.

Keywords: phytoplankton spring bloom, bacteria, primary production, bacterial production, temperature, global warming, Baltic Sea, bacterial community composition

# INTRODUCTION

Under various greenhouse gas emission scenarios, ocean surface temperatures are predicted to increase 2–5◦C by the end of this century (IPCC, 2013). Rising temperatures directly and indirectly impact pelagic organisms and aquatic food webs, leading to changes in the structure and functioning of marine ecosystems (Boyd and Doney, 2002; Sarmiento et al., 2004). Phytoplankton account for ∼50% of global net primary production (PP) and are the main energy source for aquatic ecosystems (Field et al., 1998). The major consumers of phytoplankton-derived organic matter are heterotrophic bacteria present in the upper water layers of aquatic ecosystems. The coupling of phytoplankton dissolved organic carbon (DOC) production with DOC consumption by heterotrophic prokaryotes (mostly bacteria) plays a central role in the biogeochemistry of pelagic food webs (Azam, 1998; Ducklow, 2000). The phytoplankton-bacteria relationship and the coupling between the two components is generally analyzed by comparing primary production to bacterial production (BP) rates (Hoppe et al., 2002; Morán et al., 2002, 2013) or, more comprehensively, to bacterial carbon demand (BCD), thereby also including bacterial respiration (del Giorgio et al., 1997; Rivkin and Legendre, 2001). The degree of coupling between autotrophic producers and heterotrophic decomposers in planktonic systems has a strong impact on the fate of organic matter and its partitioning into different pathways, such as microbial utilization, transfer to higher trophic levels, or accumulation and export (Wohlers et al., 2009). The DOC production by phytoplankton also shapes the succession of bacterial taxa and their specific functions (Sarmento and Gasol, 2012; Teeling et al., 2012), and the amount and the composition of the released DOM strongly depend on phytoplankton species and the physiological status of this cell (Nagata, 2000; Thornton, 2014).

In principle, all biological processes are modulated by temperature but the observed effects on metabolism in marine plankton are generally stronger for heterotrophic than for autotrophic organisms (Pomeroy and Deibel, 1986; Morán et al., 2006). The metabolic theory of ecology (Brown et al., 2004) predicts that respiration increases at higher rate than photosynthesis with increasing temperature, due to the lower activation energy of autotrophs (Harris et al., 2006). Moreover, phytoplankton are most often limited by light or nutrient levels (Tilzer et al., 1986), which diminishes the temperature sensitivity of growth (Edwards et al., 2016). The effects of temperature are reflected in reduced bacterial growth during phytoplankton spring blooms at low temperatures (e.g., polar regions), which may temporarily uncouple heterotrophic DOC consumption from autotrophic organic matter production (Pomeroy and Deibel, 1986; Kirchman et al., 2009). Conversely, an increase in water temperature has the potential to intensify the degree of phytoplankton-bacterioplankton coupling by stimulating bacterial growth and substrate consumption more than for the phytoplankton. Therefore, an increase in temperature potentially shifts the balance of autotrophic production and heterotrophic consumption toward the latter (Hoppe et al., 2002; López-Urrutia et al., 2006; Morán et al., 2006; O'Connor et al., 2009; Degerman et al., 2013).

Over the last decade, the impact of temperature changes on phyto-bacterioplankton coupling and the consequences for the marine carbon cycle have mainly been investigated in mesocosm studies that included experimental warming (e.g., Morán et al., 2006; O'Connor et al., 2009; Lindh et al., 2012). For example, the effect of sea surface warming on food web dynamics and pelagic carbon flow patterns has been investigated in several indoor-mesocosm experiments using natural spring plankton communities from the Baltic Sea (Sommer et al., 2012; Wohlers-Zöllner et al., 2012). Among other results, a temperature increase was repeatedly shown to strongly stimulate bacterial abundance, bacterial production (BP), and bacterial respiration, resulting in an increased processing of phytoplankton-derived organic matter by heterotrophic bacteria and a higher carbon flow into the microbial food web (Hoppe et al., 2008; Wohlers et al., 2009; von Scheibner et al., 2014). The results of other experiments, performed both in the Baltic Sea (Müren et al., 2005; Eriksson Wiklund et al., 2009; Degerman et al., 2013; Vaquer-Sunyer et al., 2015) and in other marine areas (e.g., Keller et al., 1999; O'Connor et al., 2009), also revealed that an increase in temperature increases planktonic respiration and intensifies the coupling of primary producers and heterotrophic consumers. Overall, warming-induced increases in heterotrophic activities resulted in a higher net consumption of DOC and subsequently in a reduced net consumption of dissolved inorganic carbon (DIC), thus constituting a positive feedback response to global warming (Wohlers et al., 2009).

However, current knowledge of the underlying mechanisms by which surface water warming influences food web dynamics and phyto-bacterioplankton coupling is still limited, and it is not clear whether the results from mesocosm studies can be extrapolated to in situ conditions. This is due to a paucity of field studies investigating the coupling of phyto- and bacterioplankton production under the in situ conditions of the early spring bloom, when water temperatures are low and the phytoplankton development depends on a first and generally weak stratification of the water column.

The Baltic Sea is a brackish, semi-enclosed shelf sea, with pronounced phytoplankton blooms in spring and autumn. In the southern Baltic, the phytoplankton spring bloom normally occurs between late February and early April whereas in northern regions it often begins later and extends until May, depending on the intensity of the surface irradiance, water stratification, and ice cover (Spilling and Markager, 2008; Wasmund et al., 2008; Klais et al., 2013). The spring blooms are typically dominated by diatoms (e.g., Chaetoceros spp. and Skeletonema costatum) although in some parts of the Baltic Sea (e.g., the central Baltic Sea) cold-water dinoflagellates may be dominant, especially after warmer winters (Wasmund et al., 2008, 2011; Klais et al., 2011, 2013). A strong increase in annual mean surface temperature is predicted for the Baltic Sea, with pronounced winter warming by up to 6◦C expected by the end of this century (HELCOM, 2013).

The aim of this study was to investigate the phytobacterioplankton coupling during early spring bloom conditions at low water temperatures in the Baltic Sea, and to examine their response to short-term warming. We hypothesized that at these low temperatures the activity and production of planktonic bacteria would be suppressed compared to phytoplankton. For this purpose, we collected water samples from different stations in the northern and southern Baltic Sea, assessed phyto- and bacterioplankton composition and measured primary and bacterial production. Additionally, we performed shipboard incubations, where PP and BP levels in response to an increase in temperature were measured. Some of the results confirmed those of previous mesocosm warming studies, but striking differences with respect to the strength of phyto-bacterioplankton coupling became also obvious.

#### MATERIALS AND METHODS

#### Sampling

Water samples were taken at eight stations (stations 1–8), extending from the Gulf of Finland to the southern Baltic Sea, during a research cruise of the R/V Alkor between March 4 and 11, 2009 (**Figure 1**). Surface water samples were collected directly after sunrise using a rosette comprising 24 10-L bottles equipped with a conductivity/temperature/depth (CTD) sensor (SeaBird 911) and sensors for fluorescence. The concentrations of inorganic nutrients were determined as described by Grasshoff et al. (1999). To collect microbial biomass for DNA extraction, water samples of 1–1.5 L were filtered onto 0.2-µm polycarbonate filters (without pre-filtration) and stored frozen at −80◦C. DNA was extracted as described in Weinbauer et al. (2002). For phytoplankton, 250-mL samples were fixed with Lugol's iodine and a subsample was later counted using an inverted microscope (Utermöhl, 1958). Phytoplankton species identification was performed in agreement with the HELCOM COMBINE protocol and the Checklist of Baltic Sea Phytoplankton Species (Baltic Sea Environment Proceedings No. 95, Helsinki Commission). Phytoplankton cell volumes were calculated after an approximation to geometric standards and converted to phytoplankton biomass (µg C L−<sup>1</sup> ) according to HELCOM recommendations (Olenina et al., 2006).

Heterotrophic and autotrophic picoplankton were analyzed by flow cytometry using a FACScalibur (Becton & Dickinson) with a constant flow rate (35 µL min−<sup>1</sup> ) and yellow-green latex beads (0.5µm, Polysciences), which served as an internal standard. Duplicates of unfiltered 4-mL samples were fixed with 400 µL of 1% paraformaldehyde and 0.05% glutaraldehyde (final concentration), shock frozen in liquid nitrogen, and stored at −20◦C. Smaller autotrophic cells, including Synechococcus as well as pico- und nanoeukaryotic phytoplankton (<5 µm), were distinguished by size and fluorescence (chlorophyll-a and phycoerythrin). Heterotrophic cells were analyzed after staining with 2.5µM (final concentration) SYBR Green (Molecular Probes). Bacteria were detected by their characteristic position in a plot of side scatter (SSC) vs. green fluorescence (FL1) and were further divided into high nucleic acid (HNA) and low nucleic acid (LNA) bacterial cells as described by Gasol and del Giorgio (2000).

# Primary Production and Bacterial Secondary Production

map was plotted using Ocean Data View (Schlitzer, 2011).

The first CTD in the morning (7:30 a.m. for all stations except stations 1, 5, and 8, 10:30 a.m.) was used to collect surface water (1–2.5 m) for the experimental incubations and the PP and BP measurements at in situ and experimentally increased (+6 ◦C) temperatures. PP was measured using the [14C]-bicarbonate incorporation method of Gragas (1975), with 200 µL of [14C] bicarbonate (10 µCi/mL) per 250-mL sample and three different light intensities (100, 75, and 50% of in situ light irradiation) to simulate the first few meters of the surface water layer. Triplicate samples of each light intensity and of one sample subjected to dark conditions (covered with aluminum foil) were incubated in a closed transparent incubator for 6 h on the deck of the ship at two different temperatures (10 ◦C and 1+6 ◦C) (Figure S1). The temperature was carefully adjusted to 1+6 ◦C using a thermostatic water bath according to the measured ambient water temperature (**Table 1**). The reactions were terminated by immediately filtering the samples through cellulose-nitrate filters (0.2µm) and then exposing them to HCl fumes for 10 min before they were fixed with Lumagel scintillation cocktail (Packard). The PP values were used to calculate the surface water production rates (µg C m−3). Due to technical problems in gathering data of daily rates of photosynthetically active radiation (PAR), daily production rates were roughly estimated by doubling the values for the half-day incubations (6 h) (Wasmund et al., 2001).

BP was measured based on [3H]-leucine (306 mCi mmol−<sup>1</sup> ) incorporation as described by Simon and Azam (1989). Triplicate


TABLE 1 | Overview of the baseline parameters for all sampled stations in the Baltic Sea.

10-mL aliquots of the unfiltered samples and of one blank were incubated in dark with [3H]-leucine (100 nM final concentration) at two different temperatures (10 ◦C and 1+6 ◦C) for 2 h before the reactions were stopped by the incorporation of formaldehyde (1% final concentration). The blank consisted of a sample in which formaldehyde was added before the addition of [3H] leucine. The same thermostatic water bath, as for the PP, was used to adjust the temperature to 1+6 ◦C. All samples were filtered onto 0.2-µm polycarbonate filters (Millipore) and rinsed with 10 mL of cold 5% trichloroacetic acid. The filters were dissolved in 4 mL of scintillation cocktail (Lumagel Plus) and the incorporated label subsequently counted in a scintillation counter (Packard). BP was calculated, assuming a leucine to carbon conversion factor of 1.5 kgC mol−<sup>1</sup> leucine (Kirchman, 2001). Bacterial carbon demand (BCD) was estimated based on BP and an estimate of bacterial respiration, using the models proposed by Rivkin and Legendre (2001) as well as by del Giorgio and Cole (1998), **Table 2**.

Activation energy for metabolic rates (PP and BP) was derived from the Arrhenius equation by

$$\ln(\mathbf{k}\_2/\mathbf{k}\_1) = \mathbf{E}\_\mathbf{a} / \mathbf{R}^\* (1/\mathbf{T}\_1 - 1/\mathbf{T}\_2) \tag{1}$$

where k<sup>1</sup> and k<sup>2</sup> are the metabolic rates at in situ and elevated temperature, respectively, T<sup>1</sup> and T<sup>2</sup> are the corresponding in situ and elevated incubation temperatures in Kelvin (K) and R is the gas constant (8.314472 mol−<sup>1</sup> K −1 ). The Q<sup>10</sup> (the relative change in a metabolic rate expected for a 10 K temperature increase) was calculated by using the equation of Raven and Geider (1988):

$$\mathbf{Q}\_{10} = \mathbf{e}^{10\text{Ea/RT2}} \tag{2}$$

where Ea is the activation energy, R is the gas constant and T is the mean temperature in Kelvin.

#### Analysis of Bacterial Community Composition

DNA was amplified using the bacterial 16S rRNA gene primers Bakt\_341F (CCTACGGGNGGCWGCAG) and Bakt\_805R (GACTACHVGGGTATCTAATCC) (Herlemann et al., 2011) and sequenced using pyrosequencing technology. For data analysis, the resulting sequences were assembled using QIIME 1.9.1 (Caporaso et al., 2010) and the "joins paired-end Illumina reads" function with default settings to merge forward and reverse sequences with an overlap of at least 30 bp. Sequences without overlap were discharged. After converting fastq to fasta using the "convert\_fastaqual\_fastq" function, the resulting sequences were evaluated using the SILVA NGS pipeline (Klindworth et al., 2013) with default settings. This automated pipeline aligns the reads to a curated database using the SINA aligner (Pruesse et al., 2012), in which problematic reads such as PCR artifacts (including potential chimeras) and non-ribosomal reads are filtered out. The reads were quality filtered with the following settings: reads <50 aligned nucleotides and reads with >2% ambiguities, >2% homopolymers, or low alignment quality. After alignment, the sequences were dereplicated by clustering according to their 98% sequence identity with each other (pairwise distance and single linkage clustering) using CD-HIT (Li and Godzik, 2006). The longest read in each cluster was BLAST searched against SILVA SSU Ref 128 for the classification of sequences. The resulting classification of the reference sequence of a cluster was mapped to all members of the respective cluster as well as to their replicates. Similar classifications (approximately resembling genus level) were merged to operational taxonomic units (OTUs). Closest related sequences in the SILVA SSU Ref 128 database to the sequence with most abundant reads are shown for the dominant OTUs in Table S1. Sequences having an average BLAST alignment coverage and alignment identity of <93% were considered as unclassified and assigned to the group "no relative." OTU counts were rarefied to 1000 reads per sample using the single\_rarefraction.py script implemented in Qiime version 2.0 (Caporaso et al., 2010). The 16S rRNA gene sequences were part of a previous study (Herlemann et al., 2016). The raw sequencing data were deposited at the Short Sequence Archive under accession number PRJEB14590.

#### Statistical Analyses

PP and BP data sets were statistically analyzed using the paired t-test to determine the significance of the temperature effects at the in situ (10 ◦C) and elevated (1+6 ◦C) temperatures. The significance level was set at p < 0.05 (IBM SPSS Statistics 20). To compare the dominant taxa between stations, the relative abundances of the >75% most abundant OTUs (>40 reads) Q10

◦C)

 (

 PP BP

16

◦C)

were visualized in a heatmap. Explicet (Robertson et al., 2013) was used to estimate richness and Shannon diversity for the eight stations. Bacterial OTUs that significantly differed in their relative abundances between stations were identified using a linear discriminant analysis effect size (LEfSe) analysis (Segata et al., 2011) with the default settings, except the "One against all" strategy, for multi-class analysis.

#### RESULTS

#### In Situ Conditions of the Baltic Sea

During the cruise in early March, the in situ water temperatures were between 0.1◦ and 3.2◦C (**Figure 1**, **Table 1**). Within the first part of the transect (hereafter called "northern stations"), extending from the Gulf of Finland (station 1) to those in the central Baltic Sea (stations 3–4), there was no indication of an ongoing phytoplankton spring bloom. Phytoplankton biomass (5.1–39.2 µg C L−<sup>1</sup> ) (**Figure 2A**, **Table 1**) and production (**Figure 3A**) levels in the surface waters were very low, with no sign of nutrient depletion (**Table 1**). The phytoplankton community was dominated by the dinoflagellate Scrippsiella sp., which at station 1, for example, contributed ∼95% to total phytoplankton biomass (**Figure 2A**, Table S2). Several large ciliates (mainly Lohmaniella sp.) were also identified. By contrast, at stations in the southern Baltic Sea (5–8), developing phytoplankton blooms were evidenced by biomass and PP levels up to 10 times higher than at the northern stations (PP: 180– 330 vs. 2–20 µg C L−<sup>1</sup> day−<sup>1</sup> , respectively; **Figures 2A**, **3A**) and by the depletion of nutrient concentrations, particularly nitrate (**Table 1**). The phytoplankton communities at the southern stations were dominated by the diatom Skeletonema costatum (59–93% of phytoplankton biomass), with minor contributions by the dinoflagellate Gymnodinium ssp. and other diatoms (Chaetoceros spp., Thalassiosira rotula) (**Figure 2A**). Overall, 26 different phytoplankton species within nine phyla (e.g., Dinophyta or Heterokontophyta) were detected over the whole transect (Table S2).

Heterotrophic bacterial abundance was also higher at the southern (0.64–0.90 × 10<sup>6</sup> cells mL−<sup>1</sup> ) than at the northern (0.43–0.59 × 10<sup>6</sup> cells mL−<sup>1</sup> ) stations but the increase was relatively modest (**Figure 2B**). HNA bacteria accounted for 0.13– 0.16 × 10<sup>6</sup> cells mL−<sup>1</sup> at the northern and 0.24–0.35 × 10<sup>6</sup> cells mL−<sup>1</sup> at the southern stations (average 30 ± 4% and 39 ± 3% of total prokaryotes, respectively) (**Figure 2B**). The difference in BP was much stronger, as the rates measured at the southern stations were 8-fold higher (**Figure 3B**). Picocyanobacteria (mainly Synechococcus), picoeukaryotes, and nanoeukaryotes, enumerated by flow cytometry, did not show a similarly clear pattern. Their abundance was highest at station 5 and lowest at station 2 (**Figure 2B**).

#### Temperature Effect on Bacteria-Phytoplankton Coupling

The experimental temperature rise of △+6 ◦C resulted in a significant increase in PP at all stations (paired t-test, p = 0.008, n = 8) (**Figure 3A**). This warming-dependent increase in PP was recorded for incubations at 100 and 75% light intensity, while


Frontiers in Marine Science | www.frontiersin.org

1 2 3 4 5 6 7 8

Bacterial carbon demand (BCD) was estimated according to the models by del Giorgio and Cole (1998) (BCD1) and Rivkin and Legendre (2001) (BCD2). BCD2 was used for the calculation of BCD/PP ratios.

303.2

± 55.0

 413.8

± 66.2

 3.84

 7.04

 1.3

 1.7

 7.0

± 0.3 12.7

± 0.7 9.1

± 0.7 17.6

± 1.2

 3.0

 4.2

± 0.3 1.7 2,8

± 0.2

± 0.1

± 0.1

± 0.54

± 0.29

181.4

± 16.7

 219.6

± 20.8

 2.14

 3.30

 1.2

 1.5

 5.2

 7.2

± 0.3 6.9

± 0.2 9.5

± 0.6

 3.8

 4.3

± 0.3 1.4 2.1

± 0.1

± 0.1

± 0.1

± 0.1

± 0.19

± 0.16

201.7

± 15.5

 235.9

± 60.1

 1.94

 2.09

 1.0

 0.9

 4.5

 5.7

± 0.6 4.3

± 2.2 6.1

± 1.1

 2.1

 2.6

± 0.4 1.3 1.1

± 1.1

± 1.3

± 0.2

± 0.1

± 0.39

± 0.15

327.5

± 43.7

 370.4

± 69.2

 4.38

 7.80

 1.3

 2.1

 8.6

± 0.3 13.9

± 0.7 12.4

± 0.7 27.9

± 1.6

 3.8

 7.5

± 0.4 1.2 2.6

± 0.2

± 0.1

± 0.1

± 0.45

± 0.23

19.2

± 10

 42.7

 0.59

 0.88

 3.1

 2.1

 3.0

 3.7

± 0.1 1.8

± 0.1 3.5

± 0.3

 9.4

 8.3

± 0.6 3.8 1.9

± 0.6

± 0.1

± 0.2

± 0.2

± 0.08

± 0.04

± 10.8

9.7

2.0

8.3

± 4.2

 29.2

 0.95

 1.21

 11.5

 4.1

 3.6

 4.0

± 0.2 2.6

± 0.3 4.2

± 0.4

 32

 14.2

± 1.5 8.2 1.5

± 3.3

± 0.2

± 0.1

± 1.1

± 0.02

± 0.09

± 5.6

± 0.8

 4.0

 0.74

 1.11

 36.4

± 4, 1 27.4

± 2.6

 3.3

 4.0

± 0.2 2.2

± 0.2 3.9

± 0.4

 105

 95.5

± 8.8 3.1 1.9

± 11.3

± 0.1

± 0.11

± 0.08

± 1.7

± 2.6

 11.2

 0.76

 1.21

 7.8

 10.8

 3.3

 4.0

± 0.2 2.0

± 0.2 3.8

± 0.3 20.5

± 1.8

 34.1

± 2.7 1.3 2.2

± 0.1

± 0.9

± 0.7

± 0.10

± 0.07

± 3.0

TABLE

2


Daily

primary

production

(PP),

bacterial

production

(BP),

and

the

resulting

BP/PP

ratios

for

10

◦C

and

1

+6

◦C

and

derived

Q10

values

for

PP

and

BP.

at 50% light intensity PP remained almost constant (Figure S3). The temperature increase of △+6 ◦C led to a stronger relative increase in PP at stations with a low phytoplankton biomass and a low rate of PP (stations 1–4; up to 250% at station 3) than at stations with high PP rates (stations 5–8). The increase at the latter stations during the incubation period was only 20% on average (**Figure 3A**, **Table 2**), with the strongest increase (36%) at station 8. The calculated Q<sup>10</sup> values were in the range of 1.2–1.7 for the southern, high-PP stations and 1.3–8.2 for the northern, low-PP stations (**Figure 2A**; **Table 2**). However, for the northern stations these values should be interpreted with caution as the PP values were very low and the measured increases were in some cases close to the detection limit.

With a warming of 1+6 ◦C, BP also increased significantly (paired t-test, p = 0.008, n = 8), although with a high variability (mean increase 59 ± 31%) (**Figure 3B**). At both the northern and the southern stations BP increased by 2- to 4-fold (**Figure 3B**, **Table 2**). The resulting Q<sup>10</sup> for bacterial production ranged between 1.1 and 2.8 (mean 2.2 ± 0.7) at high-PP stations (5–8) and were comparable with BP at low-PP stations which ranged between 1.5 and 2.2 (mean 1.9 ± 0.3 (**Table 2**). The calculated ratios of BP/PP and BCD/PP were much lower at the southern stations (5–8), where the phytoplankton blooms had developed, than at the northern non-bloom stations (1–4) (**Table 2**). This

was independent of the chosen model for calculating BCD (**Table 2**). The experimental temperature increase did not result in significantly different ratios of BP/PP and BCD/PP at any of the northern stations (paired t-test, p = 0.798, n = 8) (**Table 2**). However, for the southern stations only, warming resulted in clearly increased BP/PP and BCD/PP ratios.

#### Bacterial Community Composition

The northern stations (1–4) of the Baltic Sea were dominated by Actinobacteria (13.4–26.7% of the total sequences), Alphaproteobacteria (18.7–24.5%), and Bacteroidetes (10.5– 13.9%) (**Figure 4**). At the southern stations (5–8), where the diatom-dominated phytoplankton spring bloom occurred, the same phyla/classes predominated but the proportion of Bacteroidetes was much larger (15.5–40%). The contribution of Cyanobacteria (mainly Synechococcus) was maximal at station 5 (17.9%) and decreased with increasing phytoplankton biomass in the southern Baltic Sea (stations 7 and 8). This pattern was consistent with the cell counts of picocyanobacteria determined by flow cytometry (**Figure 2B**). Compared to the northern stations (1–4), the proportions of Bacteroidetes and Gammaproteobacteria nearly doubled at the southern stations (5–8), whereas those of Betaproteobacteria, Planctomycetes, and Verrucomicrobia decreased (**Figure 4**). In addition, the bacterial α-diversity decreased with increasing phytoplankton biomass (number of bacterial genera, Shannon index; Figure S2).

Among the 75% most abundant bacterial OTUs, LEfSe analysis identified 14 that differed significantly between stations with a high (stations 5–8) and low (stations 1–4) phytoplankton biomass (**Figure 5**). In the phytoplanktondominated samples of the southern stations, several OTUs belonging to Flavobacteria (unclassified Cryomorphaceae, Ulvibacter, unclassified Flavobacteriaceae) were particularly abundant, with OTU NS3a reaching a relative abundance of 16.8% at station 8 (**Figure 5**). LEfSe analysis also identified the enrichment of Candidatus Aquiluna (Actinobacteria) as well as representatives from the unclassified PeM15 group (Actinobacteria) and the BAL58 group (Betaproteobacteria) at the southern stations. The dominant OTUs from the northern stations were more diverse and included representatives from the Alphaproteobacteria, Betaproteobacteria, Acidobacteria, and Chloroflexi but also from the flavobacterial NS9 group. Additionally, several OTUs occurred in higher abundance at all eight of the investigated stations (**Figure 5**).

# DISCUSSION

#### Early Spring Plankton Development in the Baltic Sea

The Baltic Sea is characterized by annual phytoplankton spring blooms. These occur during late February to early April in the southern parts but may last into May in the northern parts (Groetsch et al., 2016). Sufficient light is essential for the onset of phytoplankton blooms. Diminished vertical mixing and the associated earlier onset of thermal stratification depend on the balance of surface warming and wind energy. In the Baltic Sea, there is a strong inter-annual variation in the springtime water temperature, which ranges between −1 ◦ and +5 ◦C (HELCOM, 2013). In the northern Baltic, low water temperatures prevent thermal stratification such that the phytoplankton spring bloom generally does not start before April (Spilling and Markager, 2008). However, shortterm warming periods can result in a temporal stratification, warmer surface temperatures, and the initiation of smaller, earlier phytoplankton blooms (e.g., Wasmund et al., 1998). Recent trends of ongoing sea surface warming have already been shown to impact the phytoplankton spring bloom in the Baltic Sea, in the form of earlier bloom onsets and changes in community composition (Klais et al., 2011, 2013; Wasmund et al., 2011). For example, during recent winter periods, the earlier stratification of the water column due to the increasing sea surface temperature frequently led to changes in phytoplankton composition and abundance, such that the proportions of coldwater species, including Diatomophyceae, declined and those of warm-water species, such as Dinophyceae and Cyanophyceae, increased.

Our study missed the beginning of the early spring bloom at the northern Baltic Sea stations, where we encountered pre-bloom conditions, but it did encompass a typical diatomdominated spring bloom (mainly Skeletonema costatum), which developed in late February and reached a peak in early March (Wasmund et al., 2013; Groetsch et al., 2016), at the sea's southern stations. Hence, this study detected the normal phytoplankton spring succession, whereby stations 5–7 probably covered the initial phase and station 8 the maximum of the spring bloom. Here the diatoms were probably already experiencing nutrient limitation, especially by nitrate availability (**Table 1**). This high phytoplankton biomass of 347 µg C L−<sup>1</sup> at station 8 was in accordance with previous phytoplankton peaks observed during springtime in the Baltic Sea (e.g., Wasmund et al., 2008). The dominance of one or few taxa during the phytoplankton spring blooms in the Baltic Sea, as observed here for Skeletonema costatum (e.g., ∼95% of total phytoplankton


biomass at station 8) is also a typical feature (Wasmund et al., 2013). Phytoplankton provides labile organic carbon, either directly through the exudation of DOC (which resembles dissolved PP) or indirectly through grazers or viral lysis, thereby activating the bacterioplankton community and triggering the growth of adapted bacterial taxa (Sarmento and Gasol, 2012; Teeling et al., 2012; Wear et al., 2015). However, at very low temperatures bacterial growth can be suppressed or delayed despite organic matter production by the phytoplankton spring blooms (Pomeroy and Deibel, 1986). The latter presumption has been challenged by studies that demonstrated considerable bacterial activity at low temperatures (Yager et al., 2001; Kirchman et al., 2009). Interactive effects between temperature and substrate supply complicate predictions of the response to temperature by bacterioplankton communities (Pomeroy and Wiebe, 2001; Hall and Cotner, 2007; Kritzberg et al., 2010).

In our field study, where the temperature at all stations was in the 0–3◦C range, activation of the bacterial communities by blooming phytoplankton was apparent at stations with high PP (**Figures 2B**, **3B**). Here, increases in bacterial abundance and production, as well as in the proportion of HNA bacteria, a common indicator of more active bacterial cells (Gasol et al., 1999), were recorded. A similar temperature experiment revealed that in situ BP and BR were positively related to temperature but BR responded more strongly to temperature than BP, indicating that increased temperature may result in a higher bacterial carbon demand and decreased growth efficiency (Kritzberg et al., 2010). Thus, overall there is considerable evidence that increasing temperature results in a higher carbon turnover in surface waters during phytoplankton spring blooms.

At the level of the main phylogenetic phyla/classes, there were only minor differences in bacterial community composition (BCC) across the stations sampled (**Figure 4**) whereas larger differences were visible at the genus level based OTU composition (**Figure 5**). Although salinity is a dominant factor for BCC in the Baltic Sea, major shifts occur when salinities drop to <3 or rise to >10 (Herlemann et al., 2011). Modest changes in main bacterial phyla in our study area can be explained by the fact that only stations in a salinity range of 5.2–10 were covered. Temperature is another strong factor for the BCC, but since the conditions of the sampled stations were also relatively similar in terms of temperature (0–3◦C), we assume that the presence of phytoplankton was probably the main driver for the differences in the dominating OTUs at the different stations. Both bacterial richness (here, the number of different OTUs) and the Shannon index decreased with increasing phytoplankton biomass (Figure S2), indicating that adapted bacteria became more dominant. OTUs belonging to Gammaproteobacteria and Bacteroidetes, which became more dominant at the stations with high phytoplankton biomass, are characteristic for diatomdominated phytoplankton blooms (Teeling et al., 2012; Buchan et al., 2014; Bunse et al., 2016). For example, at station 8 with the phytoplankton bloom peak, one OTU of the NS3a marine group within the Flavobacteriales dominated. The same taxon occurred in high abundance in a previous mesocosm experiment with a comparable diatom-dominated phytoplankton bloom (clone-157; 16% of all 16S rRNA clones; von Scheibner et al., 2014) as well as in another field study carried out in the same area of the Baltic Sea during a spring bloom (OTU\_000022; 16% of all 16S rRNA gene sequences; Bunse et al., 2016). This suggests a key role for this flavobacterial taxa in carbon processing during early diatom spring blooms in the Baltic Sea. Most of the other abundant OTUs in our study differed from those reported by Bunse et al. (2016), perhaps due to the fact that those authors encountered a higher phytoplankton biomass, with a different phytoplankton composition in the northern Baltic than was the case in our study. Generally, the OTUs that were abundant during the phytoplankton bloom at station 8 were more closely related to the OTUs from the southern Baltic Sea in the study of Bunse et al. (2016).

## Phyto-Bacterioplankton Coupling and the Impact of Warming

Although bacterial abundance and production were elevated at the southern stations (5–8), characterized by high PP rates, proportionally much stronger increases in phytoplankton and lower BP/PP ratios (1.0%–1.3%) than generally known for phytoplankton blooms were recorded. Despite our measurements of PP are probably only rough estimates of daily production, as changes in light and radiation could not be considered, and PP may change from day to day, we do not think that more extensive PP measurements would have provided very different BP/PP ratios. For example, the incubations at stations 1, 5 and 8 started later in the morning, and higher radiance over midday may have resulted in higher PP values, BP increased equivalent to PP, resulting in comparable BP/PP values. The Q10 values have to be interpreted with caution as the calculation was based on only two temperatures. However, the obtained Q10 values are well within the range of reported values from experimental warming experiments.

According to global observations, water temperature correlates positively with BP/PP ratios, which increase from 2 to 10% in cold and temperate climate zones up to roughly 40% at lower latitudes (Carlson and Ducklow, 1996; Ducklow et al., 1999; Hoppe et al., 2002). This suggests that the low temperature in our study area suppressed bacterial growth and therefore prevented stronger phytoplankton-bacteria coupling. The low BP/PP ratios are comparable to those reported in other studies of low temperature waters, in which ratios < 10% were interpreted as evidence of "uncoupled" BP and PP (Cole et al., 1988; Nielsen and Richardson, 1989). The low water temperatures during early spring blooms in the Baltic Sea probably delay the bacterial degradation of phytoplankton-derived organic matter and may result in the gradual temporal accumulation of DOC. A temporal delay of bacteria-phytoplankton coupling is often recorded at these low water temperatures, since the bacterioplankton peak follows the phytoplankton bloom peak often with a delay of 1–2 weeks and could be a critical factor for the low bacterial abundance and production in this study (e.g., Hoppe et al., 2008). Furthermore, high grazing pressure is often observed for controlling the bacterial growth during phytoplankton spring blooms at low temperatures and could be another factor for the reduction in phytoplankton-bacteria coupling (Lignell et al., 1992; von Scheibner et al., 2014).

Bacterial growth is more dependent on dissolved primary production (DPP) than on particulate primary production (PPP) (e.g., Morán et al., 2002). Data concerning the percent extracellular release (PER) of PP in the Baltic Sea during phytoplankton spring blooms are scarce, but studies from other systems indicate that PER is usually <20% of total PP (Nagata, 2000; Teira et al., 2001; Marañón, 2005; Morán et al., 2013). At a PER of 20%, the estimated BCD at the northern stations would exceed DPP whereas at the southern stations BCD could be entirely fueled by DPP. However, other mechanisms also contribute to the transfer of phytoplankton carbon to bacteria, such as sloppy feeding by zooplankton or viral lysis. As we did not measure DPP, we used PPP as a proxy for phytoplankton organic carbon production, which can fuel the microbial loop. Although BP/PP ratios are insufficient to completely describe the pelagic carbon flow, they are indicators of the current relation between autotrophic production and heterotrophic consumption of organic matter. Including measurements of DPP, bacterial respiration, and grazing rates in assessments of the carbon flow during the phytoplankton spring bloom would lead to more precise estimates but probably would not change the overall finding of low phytoplankton-bacteria coupling.

Warming (1+6 ◦C) stimulated both PP and BP, with distinctly higher average Q<sup>10</sup> values for BP (2.1 ± 0.7) than for PP (1.4 ± 0.2) during phytoplankton bloom conditions at the southern stations (**Table 2**). This positive effect of higher temperature on bacterial production rates is in agreement with many previous studies in the Baltic Sea and other marine locations in this climate zone (Vázquez-Domínguez et al., 2007; Hoppe et al., 2008; Panigrahi et al., 2013). However, in some studies the Q<sup>10</sup> for BP even exceeded the normal range of 2–3 known for heterotrophic organisms during phytoplankton spring blooms (Vaquer-Sunyer et al., 2015). PP increased significantly with higher temperatures in all samples, resulting in Q<sup>10</sup> values between 1.2 and 8.2 (**Table 2**). These were within the range reported in other field studies (e.g., Panigrahi et al., 2013) and in marine mesocosm experiments in the Baltic Sea (Hoppe et al., 2008; von Scheibner et al., 2014). In our study, the warming-related increase in PP suggested light saturation of phytoplankton, since light-limited rates of PP are insensitive to temperature changes (Tilzer et al., 1986), as also shown in mesocosm studies of marine plankton (e.g., Lewandowska and Sommer, 2010; Lewandowska et al., 2012; Sommer et al., 2012). Q<sup>10</sup> values between 1.2 and 1.7 at stations with high PP rates indicated that short-term warming did not greatly disturb biochemical processes. In addition, these values are in accordance with those of previous studies demonstrating lower temperature-sensitivity of phytoplankton growth and production (Marañón et al., 2014). A more detailed picture would be obtained by measuring not only the response of particulate but also that of dissolved primary production (e.g., Morán et al., 2006) as this is expected to be more related to bacterial substrate utilization. However, other temperaturesensitive factors, such as grazing pressure (Duarte et al., 2005; Aberle et al., 2007), qualitative and quantitative aspects of phytoplankton exudation (Zlotnik and Dubinsky, 1989; Morán et al., 2006), and substrate affinity (Nedwell, 1999), can strongly interfere with phytoplankton-bacteria interactions and will have to be considered for a more comprehensive understanding of the effects of temperature. Their inclusion would also help to explain the absence of a temperature effect on the BCD/PP ratio in samples obtained from the non-bloom conditions of the northern Baltic Sea (stations 1–4). In these waters, temperature more strongly stimulated PP than BP, such that carbon flow to heterotrophic decomposers was reduced by ∼1% per 1◦C. The high proportion of Dinophyceae (e.g., Scrippsiella sp.), which often establish a high biomass at low temperature in the northern Baltic Sea (Kremp, 2000; Kremp et al., 2008), may have been responsible for the reduced carbon flow to bacterioplankton. However, at these very low PP and BP rates, carbon budget calculations should be interpreted with caution, since small changes, close to the detection limit, nonetheless strongly impact BP/PP ratios.

# Comparison of Field Data With Experimental Warming Experiments

Our results from the Baltic Sea are consistent with those of other field studies performed at a similar cold temperature range (Hoppe et al., 2002; Morán et al., 2002; Duarte et al., 2005). They suggest that only a small fraction of the fixed carbon of phytoplankton enters the heterotrophic bacterial community during early bloom conditions at cold temperatures. Previous mesocosm studies of Baltic Sea spring communities, examined at a temperature range similar to that tested in the present work, showed that BP during phytoplankton spring bloom conditions was strongly stimulated by warming (Wohlers et al., 2009; von Scheibner et al., 2014), leading to a closer temporal coupling of autotrophic and heterotrophic processes as the temperature rose (Wohlers-Zöllner et al., 2012). The mean Q<sup>10</sup> of 2.2 for the bacterial response in our study was nearly identical to that reported in a previous experimental warming experiment (Q<sup>10</sup> = 2.4) (Hoppe et al., 2008).

However, bacterial stimulation by warming was much less in this field study than in previous mesocosm studies mentioned above, as BP/PP ratios were low (0.9–2.1%) even during the diatom blooms at the southern stations (**Table 2**), implying that BP was largely uncoupled from PP also at the elevated experimental temperatures. By contrast, in the mesocosm experiments the BCD/PP ratio during the bloom peak was ∼19% at the in situ temperature (∼2 ◦C) and ∼24% in response to a warming of +6 ◦C (von Scheibner et al., 2014). Major differences with our field study were the higher phytoplankton biomass (1,000–1,500 µg C L−<sup>1</sup> ) and PP in the mesocosm experiments (Hoppe et al., 2008; Wohlers et al., 2009; Lewandowska and Sommer, 2010), significant grazing by mesozooplankton and a longer duration. This might have provided higher substrate supply from phytoplankton origin for heterotrophic bacteria. Several studies already highlighted that bacterial growth is rather depended on the supply of dissolved PP than to particulate PP. Moreover, in experimental warming study with slight temperature increase of 2◦C resulted in massive increase of dissolved PP, whereas the particulate PP nearly stable (Morán et al., 2006). In the more unstable physical conditions of the Baltic Sea (e.g., higher vertical mixing or drifting), phytoplankton biomass is much lower (**Table 1**). This could explain the differences in the carbon flow patterns between the mesocosm experiments vs. in situ conditions in our study. Another reason for the distinct differences in phyto-bacterioplankton coupling might be that the duration of warming in this study was not sufficient to allow acclimatization of the bacterial communities to a similar extent as in the mesocosm studies, which were generally conducted for several weeks.

# CONCLUSIONS

Overall, the appearance of diatom-dominated phytoplankton blooms in the southern Baltic Sea led to a strong activation of the bacterial communities, in contrast to the northern Baltic Sea, where the phytoplankton biomass was low and dominated by dinoflagellates. The experimental temperature increase during the phytoplankton bloom significantly enhanced both PP and, especially, BP. This partially confirmed our hypothesis and previously published considerations that warming results in a stronger stimulation of bacterioplankton than of phytoplankton communities under cold-water conditions. However, according to the low BP/PP ratios, bacteria were still relatively uncoupled from phytoplankton growth, a situation that remained essentially unchanged after a temperature increase of 6◦C. This is in contrast to results from previous mesocosm experiments and suggests that during the early spring bloom in the Baltic Sea increased sea surface temperatures will not strengthen bacterio-phytoplankton coupling to a similar extent as in mesocosm experiments. The clear differences in carbon flow pattern between field conditions and experimental studies could be due to different factors such as physical conditions, food web structure and acclimation, which remain to be examined. The temporal response patterns of microbial communities to warming as well as possible bacterial adaptations are important avenues for future field and experimental research. By careful comparisons of the results of experimental mesocosm with those of in situ studies, including

the different time scales, a more realistic picture of the effects of global warming on the marine carbon budget will be obtained.

#### AUTHOR CONTRIBUTIONS

MvS and KJ planned, initiated and conducted the study and wrote the first draft of the manuscript. All authors discussed the results and edited the manuscript. DH analyzed bacterial community composition and AL assessed phytoplankton composition and biomass.

#### FUNDING

This work was funded by the German Science Foundation (DFG), grants JU367/7-3 (priority program AQUASHIFT) and

## REFERENCES


JU 367/15-1 to KJ. DH was supported by the grant MOBTT24 (Estonian Research Council, and European Union).

#### ACKNOWLEDGMENTS

We thank the crew and captain of the RV Alkor (AL332) for support during the cruise, and Marco Nack for technical assistance during the incubation experiments, as well as Annett Grüttmüller for flow cytometry analysis.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars. 2018.00231/full#supplementary-material

antarctica bloom in the Ross Sea, Antarctica. Aquat. Microb. Ecol. 19, 215–227. doi: 10.3354/ame019215


and bacteria during the spring bloom. A mesocosm study. Aquat. Microb. Ecol. 51, 105–115. doi: 10.3354/ame01198


**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 © 2018 von Scheibner, Herlemann, Lewandowska and Jürgens. 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(s) 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.

# Diatom Transcriptional and Physiological Responses to Changes in Iron Bioavailability across Ocean Provinces

Natalie R. Cohen<sup>1</sup> , Kelsey A. Ellis <sup>1</sup> , Robert H. Lampe<sup>1</sup> , Heather McNair <sup>2</sup> , Benjamin S. Twining<sup>3</sup> , Maria T. Maldonado<sup>4</sup> , Mark A. Brzezinski <sup>2</sup> , Fedor I. Kuzminov <sup>5</sup> , Kimberlee Thamatrakoln<sup>5</sup> , Claire P. Till 6, 7, Kenneth W. Bruland<sup>6</sup> , William G. Sunda<sup>1</sup> , Sibel Bargu<sup>8</sup> and Adrian Marchetti <sup>1</sup> \*

*<sup>1</sup> Department of Marine Sciences, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, <sup>2</sup> The Marine Science Institute and the Department of Ecology Evolution and Marine Biology, University of California, Santa Barbara, Santa Barbara, CA, United States, <sup>3</sup> Bigelow Laboratory for Ocean Sciences, East Boothbay, ME, United States, <sup>4</sup> Department of Earth, Ocean, and Atmospheric Sciences, University of British Columbia, Vancouver, BC, Canada, <sup>5</sup> Department of Marine and Coastal Sciences, Rutgers, the State University of New Jersey, New Brunswick, NJ, United States, <sup>6</sup> Department of Ocean Sciences, University of California, Santa Cruz, Santa Cruz, CA, United States, <sup>7</sup> Chemistry Department, Humboldt State University, Arcata, CA, United States, <sup>8</sup> Department of Oceanography and Coastal Sciences, College of the Coast and Environment, Louisiana State University, Baton Rouge, LA, United States*

#### Edited by:

*Kristian Spilling, Finnish Environment Institute (SYKE), Finland*

#### Reviewed by:

*Ulisses Miranda Azeiteiro, University of Aveiro, Portugal Penelope Ann Ajani, University of Technology Sydney, Australia*

> \*Correspondence: *Adrian Marchetti amarchetti@unc.edu*

#### Specialty section:

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

Received: *13 August 2017* Accepted: *26 October 2017* Published: *14 November 2017*

#### Citation:

*Cohen NR, Ellis KA, Lampe RH, McNair H, Twining BS, Maldonado MT, Brzezinski MA, Kuzminov FI, Thamatrakoln K, Till CP, Bruland KW, Sunda WG, Bargu S and Marchetti A (2017) Diatom Transcriptional and Physiological Responses to Changes in Iron Bioavailability across Ocean Provinces. Front. Mar. Sci. 4:360. doi: 10.3389/fmars.2017.00360* Changes in iron (Fe) bioavailability influence diatom physiology and community composition, and thus have a profound impact on primary productivity and ecosystem dynamics. Iron limitation of diatom growth rates has been demonstrated in both oceanic and coastal waters of the Northeast Pacific Ocean and is predicted to become more pervasive in future oceans. However, it is unclear how the strategies utilized by phytoplankton to cope with low Fe bioavailability and resupply differ across these ocean provinces. We investigated the response of diatom communities to variable Fe conditions through incubation experiments performed in the Fe mosaic of the California Upwelling Zone and along a natural Fe gradient in the Northeast Pacific Ocean. Through coupling gene expression of two dominant diatom taxa (*Pseudo-nitzschia* and *Thalassiosira*) with biological rate process measurements, we provide an in-depth examination of the physiological and molecular responses associated with varying Fe status. Following Fe enrichment, oceanic diatoms showed distinct differential expression of gene products involved in nitrogen assimilation, photosynthetic carbon fixation, and vitamin production compared to diatoms from low-Fe coastal sites, possibly driven by the chronic nature of Fe stress at the oceanic site. Genes of interest involved in Fe and N metabolism additionally exhibited divergent expression patterns between the two diatom taxa investigated, demonstrating that diverse diatoms may invoke alternative strategies when dealing with identical changes in their environment. We report here several mechanisms used distinctly by coastal or oceanic diatom communities as well as numerous taxaspecific strategies for coping with Fe stress and rearranging nutrient metabolism following Fe enrichment.

Keywords: diatoms, Thalassiosira, Pseudo-nitzschia, iron, metatranscriptomics, California Upwelling Zone, Northeast Pacific Ocean

# INTRODUCTION

Phytoplankton growth is limited by iron (Fe) availability in ∼30–40% of the ocean (Moore et al., 2001, 2004). The subarctic Northeast (NE) Pacific Ocean is one of the most well-characterized of these high-nutrient, low chlorophyll (HNLC) regions. Productivity in the NE Pacific Ocean remains low as a result of low Fe concentrations regardless of sufficient nitrate (NO<sup>−</sup> 3 ) levels and is typically dominated by small cells such as the cyanobacterium Synechococcus and eukaryotic picophytoplankton (Varela and Harrison, 1999). In this region, Fe is supplied to surface waters mainly through atmospheric deposition of dust from arid continental regions and volcanic emissions, with Fe inputs from continental margin sediments fueling winter phytoplankton blooms when atmospheric deposition is low (Lam et al., 2006; Lam and Bishop, 2008). A gradient in surface nutrient concentrations is observed from this oceanic region eastwards toward the continent; bioavailable Fe increases and supports higher phytoplankton biomass while NO<sup>−</sup> 3 concentrations in the upper mixed layer decrease to limiting levels on the continental shelf (Taylor and Haigh, 1996; Harris et al., 2009; Ribalet et al., 2010).

Iron-limited growth of phytoplankton may also occur in coastal waters, notably in regions of the California Upwelling Zone (CUZ; Hutchins et al., 1998; Bruland et al., 2001). These regions of the CUZ are characterized by high concentrations of upwelled macronutrients, but relatively low dissolved Fe (dFe) such that phytoplankton blooms ultimately become Fe-stressed. Low Fe levels result from the lack of Fe inputs from rivers and narrow continental shelves that prevent mixing of upwelled water with Fe derived from Fe-rich shelf sediments (Johnson et al., 1999; Bruland et al., 2001) and consequently, the primary Fe source in the CUZ is winter river sediment deposition (Hutchins et al., 2002; Chase et al., 2005).

Phytoplankton that subsist in Fe-limited environments are equipped with strategies to sustain growth during periods of physiological Fe stress and to rapidly respond to sudden increases in bioavailable Fe. Strategies employed by phytoplankton include replacement of Fe-containing proteins with Fe-independent ones to decrease cellular Fe requirements (La Roche et al., 1996; Peers and Price, 2006; Allen et al., 2008; Lommer et al., 2012), increasing Fe uptake rates through induction of high affinity Fe uptake systems (Maldonado and Price, 2001; Morrissey et al., 2015) and using Fe storage through specialized proteins or vacuoles (Marchetti et al., 2009; Nuester et al., 2012). In some diatom laboratory isolates and natural communities, these low-Fe strategies are rapidly reversed when Fe concentrations increase (Kustka et al., 2007; Lommer et al., 2012), whereas in others these strategies are permanent adaptations (Lommer et al., 2010; Marchetti et al., 2012). Phytoplankton species from low-Fe oceanic environments generally have lower growth requirements for cellular Fe than species from higher Fe coastal waters, largely linked to differences in Fe-containing photosynthetic proteins and complexes (Sunda and Huntsman, 1995; Strzepek and Harrison, 2004; Peers and Price, 2006; Behrenfeld and Milligan, 2013). While we have an understanding of how a few phytoplankton species alter their nutrient metabolism in response to chronic Fe limitation from laboratory experiments, how the nutrient strategies invoked by intermittently Fe-limited coastal taxa might differ from those used by species residing in chronically Fe-limited regions of the open ocean has not been directly compared.

A large amount of genetic diversity exists among diatom taxa, possibly due to differences in environmental pressures at the time of evolutionary emergence (Sims et al., 2006; Armbrust, 2009; Rabosky and Sorhannus, 2009). A genomic comparison between the evolutionarily older centric Thalassiosira pseudonana and the more recently evolved pennate Phaeodactylum tricornutum demonstrates the two diatoms share only 57% of their genes with each other, suggesting a tremendous amount of genomic diversity exists between members of these two diatom lineages (Bowler et al., 2008). Furthermore it is often observed that pennate diatoms, especially those belonging to the genus Pseudo-nitzschia, tend to dominate large Fe-induced blooms in HNLC waters (de Baar et al., 2005; Marchetti et al., 2012). These observations may suggest that the pennate diatoms have evolved distinct strategies for optimizing their potential for rapid growth when transitioning from low to relatively high Fe conditions, resulting in a competitive advantage over older lineages of diatoms as well as other types of phytoplankton.

To better understand whether major diatom genera from coastal and oceanic regions differ in their gene expression responses to changes in Fe availability, a comparative analysis across distinct nutrient regimes was performed through a combination of metatranscriptomic and physiological approaches. Microcosm incubation experiments were conducted at geographically diverse sites with different Fe regimes, macronutrient concentrations, and phytoplankton community compositions—at an Fe-limited oceanic site and a coastal site in the subarctic NE Pacific Ocean, and at three biogeochemically distinct sites within the Fe mosaic of the coastal CUZ. For our study, we focused on the changes in gene expression patterns between two dominant taxa across all sites—the pennate diatom Pseudo-nitzschia and the centric diatom Thalassiosira. These two taxa were classified by the Tara Oceans circumnavigation expedition to be two of the eight most abundant diatom genera in the global ocean (Malviya et al., 2016). Given the large amount of genetic and physiological variation observed between major diatom groups (Bowler et al., 2008; Marchetti et al., 2009; Sutak et al., 2012; Alexander et al., 2015), differences in molecular responses to changing Fe availabilities across the NE Pacific Ocean and CUZ were anticipated.

#### MATERIALS AND METHODS

#### Experimental Design

Incubation experiments were conducted on two separate cruises: within regions of the CUZ during July 3–26th 2014 onboard the R/V Melville and along the Line-P transect of the subarctic NE Pacific Ocean during June 7–23rd 2015 onboard the Canadian Coast Guard Ship (CCGS) John P. Tully (**Figure 1**). The incubated phytoplankton community response was assessed using a combination of physiological measurements and metatranscriptomics to examine the effects of Fe status

on diatom physiology and gene expression. Each experiment included three treatments: (1) a 5 nmol L−<sup>1</sup> FeCl<sup>3</sup> addition (Fe), (2) a 200 nmol L−<sup>1</sup> desferroxamine B (DFB) addition, and (3) an unamended control (Ctl), each sampled at two time points.

During the CUZ cruise, three incubation experiments were performed at separate locations corresponding to distinct Fe and macronutrient regimes (Supplementary Table 1), including sites of high dFe, macronutrients, and phytoplankton biomass (C1: 38◦ 39.30 N, 123◦ 39.87 W), relatively low dFe, high macronutrients and high phytoplankton biomass (C2: 38◦ 15.31 N, 123◦ 57.98 W), and low dFe with high macronutrients and low phytoplankton biomass (C3: 42◦ 40.00 N, 125◦ 0.00 W) (**Figure 1**). Near-surface seawater was collected from a depth of ∼15 m using a trace-metal clean sampling system consisting of a tow-fish sampler attached to KevlarTM line, PFA Teflon tubing, and a Teflon dual-diaphragm pump that pumped seawater directly into a positive pressure trace-metal clean bubble. The seawater was placed in a large 200 L acid-cleaned HDPE drum for homogenization before being distributed into 10 L flexible acid-cleaned polyethylene cubitainers (Hedwin Corporation). Cleaning protocols for the cubitainers included successive soaks in 1.2 mol L−<sup>1</sup> hydrochloric acid (reagent grade) for 3 days, 1.2 mol L−<sup>1</sup> hydrochloric acid (trace metal grade) for 1 week and 0.1 mol L−<sup>1</sup> acetic acid (trace-metal grade) until use. Prior to filling the cubitainers with seawater, the dilute acetic acid was removed and the cubitainers were rinsed thoroughly three times with ambient seawater from the collection site. The primary objective of these experiments was to elucidate the responses of target diatom genera and the phytoplankton community to variable Fe conditions. Therefore, sites were targeted that would ensure adequate macronutrient concentrations to support phytoplankton growth. However, at C2, 15µmol L−<sup>1</sup> of Si(OH)<sup>4</sup> was added to all cubitainers to support growth of diatoms due to the initially low Si(OH)<sup>4</sup> concentration (<4.7µmol L−<sup>1</sup> ).

During the Line-P cruise, incubation experiments were conducted at the low NO<sup>−</sup> 3 coastal station P4 (48◦ 39 N, 126◦ 40 W; referred to as C4 in this analysis) and at the chronically Fe-limited, HNLC oceanic station P26, also known as Ocean Station Papa (OSP, 50◦ 00 N, 145◦ 00 W; Harrison, 2002; referred to as O5). Seawater was collected at depths corresponding to ∼30% of incident irradiance (8–12 m) at both stations using a trace-metal clean sampling system consisting of a Teflon air bellows pump and PTFE lined KevlarTM tubing attached to a KevlarTM line. The seawater was pumped directly into 10 L acidcleaned polyethylene cubitainers placed within an on-deck tracemetal clean positive pressure flowhood. At site C4, 10µmol L−<sup>1</sup> of NO<sup>−</sup> <sup>3</sup> was added to all cubitainers to support growth of diatoms due to the initially low NO<sup>−</sup> 3 concentration (<1.5µmol L−<sup>1</sup> ).

At the start of the experiments, ambient seawater was filtered for all initial measurements (T0). For each incubation experiment, cubitainers were filled to serve as a control (Ctl) or amended with FeCl<sup>3</sup> or DFB just prior to dawn. All cubitainers were placed in on-deck Plexiglass incubators with flow-through seawater to maintain near-ambient surface water temperatures. Incubators were covered with neutral density screening to achieve ∼30% of the incident irradiance (Supplementary Figure 1). Following 24–96 h of incubation (Supplementary Table 1; depending on the measured macronutrient drawdown) the cubitainers for a specific time point were removed from the incubators and filtered immediately. The goal for each time point was to achieve measureable drawdowns in macronutrients that would infer stimulation of phytoplankton growth without complete macronutrient depletion. However, for some experiments and time points, depletion of NO<sup>−</sup> 3 or other macronutrients occurred. All filtrations were conducted at dawn. Subsamples for dissolved and particulate nutrients, size-fractionated uptake rates of dissolved inorganic carbon (DIC) and NO<sup>−</sup> 3 , sizefractionated chlorophyll a, Fv/Fm, and RNA were collected at T<sup>0</sup> and from each cubitainer according to the protocols described below.

#### Nutrient Concentrations, Uptake Rates, and Biogenic Silica Concentrations

For CUZ experiments, dissolved nitrate and nitrite (NO<sup>−</sup> 3 + NO<sup>−</sup> 2 ), phosphate (PO3<sup>−</sup> 4 ), and silicic acid (H4SiO4) concentrations were measured onboard using a Lachat Quick Chem 8000 Flow Injection Analysis system (Parsons et al., 1984) with detection limits of 0.05µM for NO<sup>−</sup> <sup>3</sup> <sup>+</sup> NO<sup>−</sup> 2 , 0.03µM for PO3<sup>−</sup> 4 , and 0.2µM for H4SiO<sup>4</sup> (Bruland et al., 2008). Particles were removed by filtration through a Whatman GF/F filter (25 mm). Reference standards for nutrients in seawater were run for quality control. During Line-P sampling, ∼15 mL of seawater was filtered through a Whatman GF/F filter into acid-rinsed polypropylene tubes and frozen at −20◦C in aluminum blocks until onshore analysis. Shortly following the cruise, the dissolved NO<sup>−</sup> <sup>3</sup> <sup>+</sup> NO<sup>−</sup> 2 , PO3<sup>−</sup> 4 , and H4SiO<sup>4</sup> concentrations were determined using an Astoria nutrient analyzer (Barwell-Clarke and Whitney, 1996). Nutrient detection limits were 0.2µM for NO<sup>−</sup> <sup>3</sup> <sup>+</sup> NO<sup>−</sup> 2 , 0.02µM for PO3<sup>−</sup> 4 , and 0.5µM for H4SiO<sup>4</sup> (Frank Whitney and Mark Belton [IOS], pers. comm.).

For biogenic silica (bSi) measurements, 335 mL (CUZ) or 250 mL (Line P) of seawater was filtered onto polycarbonate filters (1.2µm pore size for CUZ and 0.6µm pore size for Line-P, 25 mm), digested with NaOH in Teflon tubes, and measured with the colorimetric ammonium molybdate method (Krause et al., 2013).

Size-fractionated particulate nitrogen (PN), particulate carbon (PC), and NO<sup>−</sup> 3 uptake rates were obtained by adding <sup>15</sup>N-NaNO<sup>3</sup> to 618 mL subsample of experimental seawater placed within clear polycarbonate bottles. The concentration of NO<sup>−</sup> 3 added was no more than 10% of ambient NO<sup>−</sup> 3 level within CUZ incubations, and was 1 µmol L−<sup>1</sup> within Line-P incubations (corresponding to NO<sup>−</sup> 3 levels of 68% at T<sup>0</sup> and 10% within NO<sup>−</sup> 3 -amended incubations at C4, and ∼10% at O5). DIC uptake within Line-P incubations was measured by additionally spiking subsamples with 120µmol L−<sup>1</sup> NaH13CO3. Bottles were incubated in the same flow-through Plexiglass incubators where cubitainers were kept. Following 8 h of incubation, seawater samples were filtered in series through a polycarbonate filter (5µm pore size, 47 mm) via gravity filtration, and then through a pre-combusted (450◦C for 5 h) GF/F filter by gentle vacuum (<100 mg Hg). Particulates collected on the 5µm polycarbonate filter were then rinsed onto a separate pre-combusted GF/F filter using an artificial saline solution. Filters were stored at −20◦C until onshore analysis. In the laboratory, filters were heated to 60◦C for 24 h and pelletized in tin capsules (Elemental Microanalysis) in preparation for analysis of the atom % <sup>15</sup>N, atom % <sup>13</sup>C (for Line-P), particulate nitrogen (PN), and particulate carbon (PC) using an elemental analyzer paired with an isotope ratio mass spectrometer (EA-IRMS). Biomass-normalized NO<sup>−</sup> 3 uptake rates (PN-VNO3) and DIC uptake rates (PC-VDIC) for the Line-P experiments were obtained by dividing the measured NO<sup>−</sup> 3 and DIC biological uptake rates by PN and PC concentrations, respectively.

To quantify VDIC in CUZ incubations, incorporation of <sup>14</sup>C was determined using a protocol adapted from Taylor et al. (2013). Briefly, 60 mL of seawater from each cubitainer was distributed into acid-cleaned light and dark polycarbonate bottles. In each bottle, 1.2µCi of NaH14CO<sup>3</sup> was added. Bottles were incubated in the same flow-through Plexiglass incubators where cubitainers were kept for 6.5–8 h. Following incubation, samples were filtered through stacked 47 mm polycarbonate filters (5 and 1µm) separated with a mesh spacer during filtration. Filters were vacuum dried, placed in 7 mL scintillation vials containing 0.5 mL of 6 M HCl and permitted to degas for 24 h. Disintegrations per minute (DPM) retained on the filters were measured using a Beckman Coulter LS 6500 scintillation counter. Reported values are light bottle DPMs minus dark bottle DPMs. To obtain VDIC, DIC uptake rates were normalized to PC concentrations obtained as part of the NO<sup>−</sup> 3 uptake measurements within each incubation and size fraction.

#### Dissolved Iron Concentrations

Seawater samples for Fe analysis within the CUZ were acidified at sea with the equivalent of 4 mL 6 N quartz-distilled HCl per L of seawater (pH ∼ 1.7) and stored in acid-cleaned LDPE bottles for at least 2 months prior to analysis. Samples were analyzed using an adaption of Biller and Bruland (2012) as described in Parker et al. (2016). Briefly, this method involves preconcentrating the Fe from buffered (pH 6.0) seawater on Nobias-chelate PA1 resin and eluting with 1 N quartz-distilled HNO3. The eluent was analyzed with a Thermo-Element high resolution XR ICP-MS in counting mode. Line-P dissolved Fe samples were stored in acidcleaned LDPE bottles, acidified post-cruise with Optima-grade HCl (1 mL 12 N HCl per L of seawater), and allowed to sit for >3 months. Dissolved Fe was measured via ICP-MS by P. Morton at Florida State University following resin preconcentration using the protocol of Milne et al. (2010).

## Chlorophyll a

Four hundred milliliters of seawater was gravity-filtered through a polycarbonate filter (5µm pore size, 47 mm diameter) followed by vacuum filtration through a GF/F filter (0.7µm nominal pore size, 25 mm diameter) using a series filter cascade for size fractionation. Filters were frozen at −80◦C until analysis. Chlorophyll a extraction was performed using 90% acetone at −20◦C for 24 h and the extracted Chl a was quantified by fluorometry with a Turner Designs 10-AU fluorometer using the acidification method (Parsons et al., 1984).

# Domoic Acid

Approximately 250 mL of seawater from each CUZ site was filtered through GF/F filters (25 mm) via vacuum pressure (<100 mm Hg) and the filters were frozen at −80◦C. The filters were extracted with 20% methanol (MeOH) in water. The mixture was sonicated in an ice bath for 2 min at 30– 40 W with a Sonicator 3000, followed by centrifugation (10 min, 1,399 × g). The supernatant was collected and passed through a 0.22µm syringe filter. Samples were stored at −20◦C until analysis. Concentrations with a detection limit of 0.01 µg L−<sup>1</sup> were obtained using an enzyme-linked immunosorbent assay (Abraxis, Warminster, PA, USA) following the manufacturer's protocol, including running each sample in duplicate at several dilutions. Final concentrations (pg DA mL extract−<sup>1</sup> ) were calculated using the manufacturer supplied analysis spreadsheet.

# Photophysiology

The maximum photochemical yield of PSII (Fv/Fm) was measured by fast repetition rate fluorometry (FRRF) using a custom-built fluorescence-induction and relaxation system (Kolber et al., 1998; Gorbunov and Falkowski, 2004). Before each measurement, a 5 mL subsample of seawater from each cubitainer was acclimated to low light for 20 min. A saturating pulse (20,000µmol photons m−<sup>2</sup> s −1 ) of blue light (450 nm) was applied to dark-acclimated cells for a duration of 100– 200µs. Measurements were obtained using the single-turnover flash (STF) setting with the average of 50 iterations for the CUZ experiments, and a single iteration for the Line-P experiments. Data were blank corrected using 0.2µm filtered seawater.

# RNA Extraction and Bioinformatic Analysis

Phytoplankton in seawater samples were filtered onto 0.8µm Pall Supor filters (142 mm) via peristaltic pumping, immediately flash frozen in liquid nitrogen and stored at −80◦C until extraction onshore. The filters were briefly thawed on ice before being extracted individually using the ToTALLY RNA Kit (Ambion). The extraction procedure followed manufacturer protocols with the modified first step of glass bead addition and vortexing to facilitate disruption of cells. Removal of DNA was performed with one round of DNAse I (Ambion) incubation. For the Line P experiments, due to low yields in treatments, RNA from the triplicate cubitainers was pooled prior to sequencing. Within CUZ experiments all triplicate incubation samples were sequenced separately. At the oceanic site O5, RNA yields were too low to successfully sequence metatranscriptomes at the T<sup>1</sup> timepoint, and consequently, transcriptomic analyses were performed using the T0, T<sup>2</sup> Fe, and T<sup>2</sup> Ctl treatments. Metatranscriptomic library preparation was performed with the Illumina TruSeq Stranded mRNA Library Preparation Kit and HiSeq v4 reagents. Samples were barcoded and run across three lanes of Illumina HiSeq 2000 (125 bp, paired-end) yielding on average 23 million paired-end reads per sample (Supplementary Table 2). The RNA-seq data reported here has been deposited in the National Center for Biotechnology (NCBI) sequence read archive (SRA) under the BioProject accession no. PRJNA320398 and PRJNA388329.

Raw reads were trimmed for quality bases and removal of adapters using Trimmomatic v0.32 (paired-end mode, adaptive quality trim with 40 bp target length, and strictness of 0.6, minimum length of 36 bp; Bolger et al., 2014). Trimmed paired reads were merged into single reads with BBMerge v8.0. For each site, the resulting merged pairs and non-overlapping pairedend reads were assembled using ABySS v1.5.2 with a multi-kmer approach (Birol et al., 2009). The different k-mer assemblies were merged to remove redundant contigs using Trans-ABySS v1.5.3 (Robertson et al., 2010). Read counts were obtained by mapping raw reads to assembled contigs with Bowtie2 v2.2.6 (end-to-end alignment; Langmead and Salzberg, 2012). Alignments were filtered by mapping quality score (MAPQ) of 10 or higher as determined by SAMtools v1.2 (Li et al., 2009). Taxonomic and functional annotations were assigned based on sequence homology to reference databases via BLASTx v2.3.0 with an e-value cutoff of 10−<sup>3</sup> (Altschul et al., 1990). Functional annotations were assigned according to the top hit using the Kyoto Encyclopedia of Genes and Genomes (KEGG; Release 75), while taxonomic assignments were obtained according to the top hit using MarineRefII (Laboratory of Mary Ann Moran, University of Georgia), a custom-made database comprised of protein sequences of marine prokaryotes and eukaryotes including all sequenced transcriptomes from Marine Microbial Eukaryote Transcriptome Sequencing Project (MMETSP) (Keeling et al., 2014). Taxonomic information was obtained from NCBI's Taxonomy Database (each isolate in MarineRefII is assigned a NCBI taxonomic ID). The information from NCBI was manually curated to ensure proper assignment and use of common phytoplankton taxonomic ranks. For our analysis, we have grouped diatom-associated sequences at the genus level. Therefore, the patterns in gene expression observed could be driven by one dominant species or many equally distributed species belonging to a genus within each site.

All diatom-assigned counts were summed to both the genus taxonomic rank and KEGG Orthology (KO) functional annotation level. For genes of interest without a KO assignments but with an annotated gene definition (i.e., ISIPs and rhodopsin), raw counts corresponding to KEGG gene definitions were summed. EdgeR v3.12.0 was used to calculate Pseudo-nitzschiaor Thalassiosira-specific normalized fold change and countsper-million (CPM) from pairwise comparisons using the exactTest (Robinson and Smyth, 2008; Robinson and Oshlack, 2010; Robinson et al., 2010; Klingenberg and Meinicke, 2017). Significance (p < 0.05) was calculated with edgeR's estimate of tagwise dispersions across all samples within CUZ sites. Heatmaps were produced with the R package pheatmap v1.0.8, and dendrograms created using Euclidean distance and hierarchical clustering. Assembled contigs, read counts, and functional annotations of contigs are available at marchettilab.web.unc.edu/data.

In order to directly compare transcript abundance across locations for principal component analyses (PCA), the assemblies for all sites were merged with Trans-ABySS. The removal of redundant contigs was verified with GenomeTools v.1.5.1. Counts were obtained by aligning raw reads against this merged metatranscriptome using Salmon v.0.7.3-beta. Normalized counts were then obtained with edgeR v3.12.0. PCA biplots were created using log-transformed normalized counts for genes of interest with ggbiplot v.0.5.

#### Phylogenetic Analysis of Environmental Sequences

Environmental Pseudo-nitzschia and Thalassiosira contigs functionally annotated as RubisCO (RBCL), rhodopsin (RHO), or superoxide dismutase (SOD) and containing a large number of mapped reads were compared to diatom reference sequences for phylogenetic characterization. Diatom sequences used in reference alignments were obtained through a sequence homology search using BLASTx v2.2.28 with Pseudo-nitzschia RBCL, RHO, and SOD against the database MMETSP using an E-value cutoff of 10−<sup>5</sup> (Altschul et al., 1990). Sequences were aligned using MUSCLE within Geneious v5.6.4 software (Edgar, 2004).

# RESULTS

#### Nutrient Regimes of Experimental Sites

CUZ site C1 (**Figure 1**) was characterized by high macronutrient and dFe concentrations in the mixed layer supporting a high biomass, nutrient-replete phytoplankton community. The community was dominated by phytoplankton cells in the >5µm chlorophyll a (chl a) size fraction, constituting 88% of the total chl a concentration (**Figure 2B**; Supplementary Table 1). Macronutrient concentrations were rapidly consumed during the first 24 h of incubation (T1), with near complete depletion of the NO<sup>−</sup> 3 (≤1 µmol L−<sup>1</sup> remaining by 48 h [T2]; **Figure 2A**). The initially Fe-replete phytoplankton community (dFe: 3.57 nmol L −1 ) was mostly unaffected by the additions of Fe or DFB as demonstrated through relatively constant Fv/Fm, phytoplankton biomass, particulate nitrogen (PN)-specific nitrate uptake rates (VNO3, or nitrate assimilation rates), and particulate carbon (PC)-specific dissolved inorganic carbon uptake rates (VDIC, or carbon assimilation rates) across treatments at each time point (**Figures 2B–E**). Furthermore, the NO3:Fe ratio of the initial (T0) seawater (3.8µmol:nmol. Supplementary Table 1) was substantially below the predicted threshold ratio for eventual Fe stress of 12µmol:nmol for phytoplankton in this region as proposed by King and Barbeau (2007), albeit this ratio is subject to variation as a function of phytoplankton iron demands (Bruland et al., 2001), suggesting this phytoplankton community was not likely to be driven into Fe limitation prior to complete NO<sup>−</sup> 3 utilization. However, indications of molecularlevel responses to Fe and DFB additions were observed; 74 genes were differentially expressed (p < 0.05) in Pseudo-nitzschia between the Fe and DFB treatments (Supplementary Figure 2A). Fe-stress bioindicator genes (FLDA, PETE, and ISIP2A; Whitney et al., 2011; Morrissey et al., 2015; Graff van Creveld et al., 2016) increased in expression following the addition of DFB relative to the added Fe treatment, suggesting the onset of Fe stress following the addition of DFB by the end of the first time point.

CUZ site C2 was located in close geographical proximity to C1 (**Figure 1**), yet exhibited different mixed layer properties in relation to phytoplankton biomass, silicic acid (Si[OH]4) and dFe concentrations (0.44 nmol L−<sup>1</sup> ). Nitrate and ortho-phosphate (PO3<sup>−</sup> 4 ) concentrations were similarly high (10.3 and 0.96µmol L −1 , respectively) as found at site C1, although Si(OH)<sup>4</sup> levels were appreciably lower (4.7µmol L−<sup>1</sup> ) and possibly growthlimiting to certain diatoms (Nelson et al., 1996). Therefore, incubations were amended with 15µmol L−<sup>1</sup> Si(OH)<sup>4</sup> to support potential diatom growth with added Fe (Brzezinski, 1985). Although the chl a concentration in the >5µm size fraction was initially <1µg L−<sup>1</sup> and biogenic silica (bSi) concentrations were <3µmol L−<sup>1</sup> , by 48 h (T1) the >5µm chl a fraction reached 5– 8µg L−<sup>1</sup> , and bSi increased to 10–15µmol L−<sup>1</sup> in all treatments, accompanied by appreciable decreases in NO<sup>−</sup> 3 , PO3<sup>−</sup> 4 , and Si(OH)<sup>4</sup> concentrations (**Figures 2A,B**). Since this community quickly depleted NO<sup>−</sup> 3 concentrations during the experimental period, this site presented an opportunity to couple the physiological indicators of NO<sup>−</sup> 3 stress with N-related transport and assimilation genes observed to be elevated in NO3-starved laboratory diatom cultures (Hildebrand, 2005; Song and Ward, 2007; Bender et al., 2014; Rogato et al., 2015). Apart from Fv/F<sup>m</sup> reaching relatively low values in the DFB treatments, indications of Fe stress in bulk physiological measurements across treatments were absent (**Figure 2C**). However, the initial seawater NO3:Fe ratio of 23.4µmol:nmol suggests this community may have been driven into Fe limitation provided sufficient Si(OH)<sup>4</sup> was present. Additionally, a total of 414 Pseudo-nitzschia–associated genes were differentially expressed (p < 0.05) by T<sup>1</sup> between the Fe and DFB treatments (Supplementary Figure 2). This greater number of differentially expressed genes in Pseudo-nitzschia when compared to C1 suggests the C2 diatom community in the DFB treatment experienced a higher degree of Fe stress during the incubation period. The initially low dissolved Si(OH)4:NO<sup>3</sup> ratio at this site furthermore implies a possible increase in the Si:N ratios of Fe-stressed diatoms (Hutchins and Bruland, 1998; Marchetti and Cassar, 2009; Brzezinski et al., 2015). Interestingly, concentrations of domoic acid (DA), a neurotoxin produced by Pseudo-nitzschia, was 90 pg mL−<sup>1</sup> in initial seawater (T0) and exceeded 3,000 pg mL−<sup>1</sup> in the control treatment by T<sup>1</sup> (Supplementary Figure 3). This increase in DA concentration may be linked to both the increase in Pseudo-nitzschia abundance and depletion of Si(OH)<sup>4</sup> resulting in Si-limited cells which has been shown to greatly enhance DA production (Pan et al., 1996).

Site C3 (**Figure 1**) contained the lowest dFe concentrations (0.31 nmol L−<sup>1</sup> ) among the CUZ sites along with high macronutrient concentrations [17µmol L−<sup>1</sup> NO<sup>−</sup> 3 , 19µmol L −1 Si(OH)4, and 1.5µmol L−<sup>1</sup> PO3<sup>−</sup> 4 ; **Figure 2A**]. The corresponding NO3:Fe ratio of the initial seawater was

∼54.9µmol:nmol (Supplementary Table 1). Following incubation, the chl a, bSi, PN-specific VNO3, and PC-specific VDIC were all higher in the Fe-amended treatment relative to the unamended control by T<sup>1</sup> (**Figures 2B,D,E**). By 72 h, NO<sup>−</sup> <sup>3</sup> was completely drawn down within the Fe treatment (T2). Despite the pronounced influence of Fe enrichment on bulk parameters, Fv/F<sup>m</sup> values were only slightly higher in the Fe treatment than the control, but they were substantially higher than in the DFB treatment (**Figure 2C**). This is likely a reflection of the different phytoplankton composition at this location compared to site C2, which did not show indications of an Fe-addition response on the measured bulk parameters, but did demonstrate elevated Fv/F<sup>m</sup> values in the added Fe treatment. Site C3 represented the only phytoplankton community in the CUZ that displayed a definite physiological response to Fe addition relative to the control treatment (Supplementary Table 1). The Fe-induced molecular response in diatoms was demonstrated by the differential expression of 458 genes in Pseudo-nitzschia and 1,223 genes in Thalassiosira between the Fe and DFB treatments (Supplementary Figure 2C), and 365 genes in Pseudo-nitzschia and 837 genes in Thalassiosira between the Fe and Ctl treatments (p < 0.05).

Coastal site C4 was located at station P4 of the Line-P transect in the subarctic NE Pacific Ocean (**Figure 1**). Initial mixed-layer seawater properties were characterized by low concentrations of macronutrients and dFe, which supported a low phytoplankton biomass. Nitrate concentrations were initially 1.5µmol L−<sup>1</sup> (**Figure 2A**). To facilitate a potential phytoplankton growth response to added Fe, 10µmol L−<sup>1</sup> of NO<sup>−</sup> <sup>3</sup> was added to each treatment. Si(OH)<sup>4</sup> concentrations were also initially low (2.2µmol L−<sup>1</sup> ) and incubation concentrations dropped to <2µmol L−<sup>1</sup> in most treatments by the second time point (T2; **Figure 2A**). These low concentrations restricted biomass accumulation as bSi (**Figure 2B**) and it is likely that the resulting diatom community experienced Si(OH)<sup>4</sup> limitation by the end of the incubation period. Despite its relatively close proximity to land and relatively high dFe concentration (0.64 nmol L−<sup>1</sup> ), there was a pronounced response to Fe addition at C4 as demonstrated through higher Fv/Fm, PN-specific VNO3, and PC-specific VDIC in the Fe treatment compared to values in the unamended control by T<sup>1</sup> (**Figures 2D,E**; Supplementary Table 1). The NO3:Fe ratio following artificial NO<sup>−</sup> 3 addition was 18.8µmol:nmol, sufficiently high to cause Fe stress with phytoplankton growth following an increase in phytoplankton biomass.

Oceanic site O5 was located at Ocean Station Papa (OSP), station P26 of the Line-P transect (**Figure 1**). This site demonstrated characteristically high macronutrients and low dFe (0.05 nmol L−<sup>1</sup> ), resulting in the highest NO3:Fe ratio observed across all experimental sites (234µmol:nmol; Supplementary Table 1). Phytoplankton biomass was initially low, consistent with historical observations from this well-characterized Felimited region (**Figure 2A**; Supplementary Table 1; Boyd and Harrison, 1999). In contrast to most of the coastal sites, the majority of the phytoplankton biomass was dominated by picophytoplankton and other small cells (<5µm) initially and throughout the incubation period (Supplementary Table 1; **Figure 2B**). Biogenic Si concentrations only increased after 96 h with similar responses in controls and Fe treatments (**Figure 2B**). Both large and small chl a size fractions, Fv/Fm, PN-specific VNO3, and PC-specific VDIC were higher in the Fe treatment than in the unamended control (Ctl), confirming that the phytoplankton community in the initial seawater and in all incubation treatments without added Fe were experiencing Fe limitation (**Figures 2B–E**).

#### Community Composition across Sites

Metatranscriptomic assembly of sequence data and subsequent taxonomic annotation yielded the relative transcript proportions of phytoplankton functional groups (**Figure 3**). The CUZ site C1 was predominantly comprised of diatom transcripts at T0; however, there was a 26% decrease in diatom transcripts in both the Fe and DFB treatments by T1, accompanied by genus-level shifts within the diatoms. In contrast, CUZ site C2 initially yielded a phytoplankton community transcript pool dominated equally by diatoms (30%) and prasinophytes (28%), with diatoms remaining a dominant taxa following incubation (26–28%) and prasinophyte transcripts substantially decreasing from 28 to 3–8% in both Fe and DFB incubations. CUZ site C3 contained a phytoplankton community transcript pool almost equally represented by diatoms, prasinophytes, haptophytes, and dinoflagellates with little change in community composition among treatments following incubation. The coastal subarctic Pacific site C4 yielded an initial phytoplankton community transcript pool dominated by dinoflagellate-assigned sequences (24%), although these sequences decreased by ∼10% in the Fe treatment, concurrent with a 9% increase in diatom transcripts. At the oceanic site O5, there were initially equal proportions of prasinophyte (22%) and haptophyte (23%) transcripts, with little representation by diatoms (4%). However, diatom-assigned transcripts constituted 9% of the community transcript pool by T<sup>2</sup> in the Fe addition treatment. Pseudo-nitzschia and Thalassiosira were among the top five diatom genera at all sites examined based on relative transcript abundance (**Figure 3**). These two genera together constituted between 9 and 53% of the transcript proportions in the initial diatom communities, and 25–58% of the Fe-enriched diatom communities.

### Gene Expression Responses to Fe Status across Sites

Gene expression responses among sites were compared using Euclidian distance similarity analyses between Fe and DFB treatments (Fe/DFB, Fe/Ctl for O5) within the diatom genera Pseudo-nitzschia and Thalassiosira (**Figure 4**). Expression responses within coastal sites clustered together, while the oceanic site O5 displayed distinctly different patterns in both taxa. At site O5, 83 out of 1,334 KEGG Orthology genes (KOs) in Pseudo-nitzschia demonstrated >16-fold higher expression in the added Fe treatment than in the Fe-limited control treatment (**Figure 4**, Supplemental Table 4). By comparison, 155 out of 1,241 KOs in Thalassiosira showed >16-fold higher expression in the added Fe treatment compared to the low Fe control treatment. The most highly differentially expressed genes in oceanic Pseudo-nitzschia following Fe enrichment were ferritin (FTN, 290-fold), a metal transporter (CNNM, 32-fold), a putative bicarbonate (HCO3) transporter (ICTB, 133-fold), and an NADPH-dependent glutamate synthase (GLT; 146-fold). In oceanic Thalassiosira, highly differentially expressed genes included ferredoxin-dependent sulfite reductase (Fd-SIR, 74-fold) and ferredoxin-dependent glutamate synthase (Fd-GLT; 416-fold). Fe addition induced both genera to increase the expression of several genes involved in photosynthesis by >16-fold exclusively at this location. Both taxa overexpressed gene products involved in vitamin biosynthesis, including the Fedependent vitamin B<sup>7</sup> synthesis protein biotin synthase (BIOB), which increased expression in the Fe enriched treatment by 84 and 49-fold in Pseudo-nitzschia and Thalassiosira, respectively.

provided as the Fe-limited comparison.

Furthermore, Pseudo-nitzschia increased expression of the vitamin B<sup>1</sup> (thiamine) biosynthetic gene THIC (by 179-fold) and vitamin B<sup>6</sup> (pyridoxine) biosynthetic genes pyridoxine kinase (PDXK; by 74-fold) and pyridoxine 4-dehydrogenase (PLDH; by 152-fold) following Fe enrichment at the oceanic site.

A number of genes demonstrated higher expression in the Fe-limited control treatment at O5. Forty-eight out of 1,334 genes in Pseudo-nitzschia and 77 out of 1,241 genes in Thalassiosira showed >16-fold higher expression in the Ctl treatment than in the added Fe treatment, patterns that were not found in diatoms from the examined coastal sites (**Figure 4**). In Thalassiosira, these genes encode proteins such as the copper (Cu)/zinc (Zn) superoxide dismutase (Cu-Zn SOD), an enzyme that removes toxic superoxide radicals by dismuting them into molecular oxygen and hydrogen peroxide, and a divalent metal transporter belonging to the ZIP family (ZIP7) (Marchetti and Maldonado, 2016). In both taxa, ribulose-1,5-bisphosphate carboxylase oxygenase (RubisCO; large subunit; RBCL), which catalyzes C-fixation in the Calvin cycle, had ≥24-fold higher expression in the Ctl treatment at O5.

# Influence of Fe Availability on Fe Metabolism

The expression of genes involved in cellular growth and function, including N and C assimilation, vitamin synthesis, Fe-related metabolism, and trace metal acquisition, were compared in the dominant diatom genera Pseudo-nitzschia and Thalassiosira between the Fe and DFB/Ctl treatments (**Figure 5**). Genes encoding proteins involved in metal transport were detected at all locations, with expression patterns varying depending on site and taxa. Pseudo-nitzscha increased expression of the Fe transporter ABC.FEV.S by >2-fold under Fe enrichment at all locations where incubated communities showed a physiological Fe effect (C3, C4, O5; Supplementary Table 1). Transcripts for another Fe uptake protein, the high affinity iron permease FTR, were generally more abundant in the DFB/Ctl treatments in Thalassiosira, although the gene was more highly expressed following Fe enrichment in Pseudo-nitzschia at sites C2, C3, and O5 (**Figure 5**). The putative metal transporter CNNM was 32 fold more highly expressed following Fe enrichment in Pseudonitzschia at the oceanic site, but was not detected in Thalassiosira. Conversely, the non-specific metal transporter ZIP7 was 21 fold more highly expressed under Fe-limiting conditions in

oceanic Pseudo-nitzschia and similarly not detected in oceanic Thalassiosira. Transcripts for Fe starvation induced proteins (ISIPs), including the recently-identified Fe acquisition protein ISIP2A that binds Fe at the cell surface and is thought to be involved in intracellular Fe transport (Morrissey et al., 2015), were highly abundant in Fe-stressed treatments (e.g., DFB and/or Ctl depending on the site) across all sites and in both taxa (**Figure 5**). Although their specific functions in diatoms are unclear, other ISIPs were markedly abundant and differentially expressed in the DFB/Ctl treatments, with ISIP1 one of the most differentially expressed genes between Fe-replete and Felimited treatments at each experimental site and in both taxa (Supplementary Figure 2).

Other Fe-related metabolic processes similarly varied depending on both site and taxa. Differences in expression patterns between taxa were generally greater for these Fe-related genes than in the N- and C-related genes investigated (**Figure 5**). At most sites, transcripts for the Fe storage protein ferritin (FTN) were higher in the Fe addition treatments than in the DFB/Ctl treatments. However, at two sites (C2 and C4), FTN transcripts were more abundant in the DFB treatment compared to the Fe addition treatment for one of the two genera (e.g., at site C2, 3.5-fold higher in Pseudo-nitzschia, p = 1 × 10−<sup>3</sup> and at site C4, 90-fold higher in Thalassiosira). SODs were additionally differentially expressed, but they showed different expression patterns depending on the enzymes' metal cofactor(s) and the diatom genus. Cu-Zn SOD, which contains both Cu and Zn at its active site, showed a >100-fold higher expression in Thalassiosira in the Fe-limited control than in the added Fe treatment at the Fe-limited site O5. In contrast, in the same Fe-limited control treatment at this location, Pseudo-nitzschia demonstrated 2-fold higher expression of Fe-Mn SOD, which contains either Fe or manganese (Mn) as its metal cofactor. Based on the presence of Mn-coordinating amino acids at sites G-77 and Q-146 of the most highly expressed Fe-Mn SOD contigs, this Pseudo-nitzschia SOD was determined to specifically utilize Mn as its metal cofactor (Crowley et al., 2000; Groussman et al., 2015) (Supplementary Figure 4C).

Transcriptional responses of genes encoding Fe-dependent proteins and their functional replacements in photosynthetic electron transport were examined in both diatom genera (**Figure 5**). Transcripts for the Fe-independent protein flavodoxin (FLDA), which functionally replaces the Fe-protein ferredoxin (PETF) in photosynthetic electron transport, were generally more abundant in the DFB/Ctl treatments than in the Fe treatments in both genera (**Figure 5**). Conversely, transcripts of PETF were >2-fold higher in the high-Fe treatment only in Thalassiosira and across all sites. In Pseudo-nitzschia, PETF transcripts were either constitutively expressed (C3 and C4), more highly expressed in the DFB treatment (C1), or not present (C2 and O5) (**Figure 5**). Transcripts of cytochrome c<sup>6</sup> (PETJ) and its functional non-Fe replacement, the copper-protein plastocyanin, also showed differences in gene expression. PETJ transcripts were more abundant in the high Fe treatment at

#### FIGURE 5 | Continued

the log2 fold change of gene expression within the Fe addition treatment relative to the DFB treatment at each site. For site O5, the T2 control (Ctl) treatment is used as the Fe-limited comparison. Gray boxes indicate transcripts were not detected in either treatment. White boxes signify no change in expression between treatments. A schematic representation of select N, C, Fe, metal transport, and vitamin-related processes within a diatom cell, color-coded by genes of interest included in (A) is provided (B). Adjacent proteins with black borders indicate similar cellular functions (e.g., *FLDA, PETF*). Gene abbreviations are NRT2, nitrate transporter; AMT, ammonium transporter; URTA, urea transporter; NR, nitrate reductase; NIRA, ferredoxin-nitrite reductase; NIRB, nitrite reductase; NIT-6, nitrite reductase; GLT, glutamate synthase; Fd-GLT, ferredoxin-glutamate synthase; α-CA, carbonic anhydrase (α family); SLC4A, solute carrier family (bicarbonate transporters); ICTB, putative bicarbonate transporter; PEPC, phosphoenolpyruvate carboxylase; RBCL, RubisCO large subunit; RBCS, RubisCO small subunit; PGK, phosphoglycerate kinase; TPI, triseophosphate isomerase; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; FBP, fructose-1,2-bisphosphatase I; TKL, transketolase; RPE, ribulose-phosphate 3-epimerase; PRK, phosphoribulokinase; FBA class I, fructose bisphosphate aldolase (class I); FBA class II, fructose bisphosphate aldolase (classII); FTR, high affinity iron permease; ABC.FEV.S: iron complex transport system substrate-binding protein; ZIP7, zinc transporter 7; CNNM, metal transporter; ISIP2A, iron starvation induced protein 2A; ISIP1, iron starvation induced protein 1; ISIP2B, iron starvation induced protein 2B; ISIP3, iron starvation induced protein 3; FTN, ferritin; FLDA, flavodoxin I; PETF: ferredoxin; PETH, ferredoxin-NADP+ reductase; PETE, plastocyanin; PETJ, cytochrome c6; PETC, cytochrome *b*6*f* complex; Cu-Zn SOD, superoxide dismutase containing Cu and Zn as cofactors; Fe-Mn SOD, superoxide dismutase containing Fe or Mn as cofactor; Fd-SIR, ferredoxin-sulfite reductase; RHO, rhodopsin (note the localization of RHO within the vacuole membrane is speculative); BIOB, biotin synthase; PDXK, pyridoxal kinase; PLDH, pyridoxal 4-dehydrogenase; THIC, phosphomethylpyrimidine synthase.

all sites and in both genera, except O5, where it was slightly more abundant in the Fe-limited control treatment (**Figure 5**). By contrast, transcripts for plastocyanin (PETE) displayed inconsistent expression trends in response to Fe status across sites, being relatively more abundant following Fe enrichment in both genera at C3 (1.4-fold in Pseudo-nitzschia; 1.9-fold in Thalassiosira, p = 5 × 10−<sup>4</sup> ) and at the initially Fe-limited oceanic site, O5 (1.4-fold in Thalassiosira; **Figure 5**). At all other locations PETE transcripts were either more abundant under DFB conditions or not detected.

Transcripts for the proton-pumping protein rhodopsin (RHO) furthermore demonstrated differences in expression patterns among genera. This protein can supplement Feintensive photosynthesis in the light-driven production of membrane proton gradients and ATP in some diatoms (Marchetti et al., 2015). Rhodopsin was not detected in Thalassiosira at any location while its expression increased in Pseudo-nitzschia by >2-fold in the DFB/Ctl treatments relative to the Fe treatment at the two lowest dFe sites [C3 (p = 0.01) and O5; **Figure 5**; Supplementary Table 1]. At the other sites RHO expression was constitutive. These rhodopsin contigs were structurally similar to diatom rhodopsins identified within the MMETSP database (≥55% similarity; Supplementary Figure 4B).

Relationships among Fe-related transcript abundance, experimental site and treatment were determined using Principal Components Analysis (PCA) individually for each diatom genus. Principle components P1 and P2 explained 54% of the variation in transcript abundance in Pseudo-nitzschia and 76% in Thalassiosira (**Figure 6C**). In Pseudo-nitzschia, transcripts for the photosynthetic genes ferredoxin-NADP+ reductase (PETH), PETJ, a cytochrome b6/f complex protein (PETC), FTN, and Cu-Zn SOD were in higher relative abundance within Fe addition treatments while RHO, ISIPs, FLDA, PETE, and FTR were generally more abundant in the Ctl and/or DFB treatments, as the principle component P1 separated these samples based on Fe treatment. In Thalassiosira, a similar response was observed, although RHO was not detected, and PETF, which was sporadically found and not abundant in Pseudo-nitzschia, strongly co-varied with the other genes highly expressed in the treatments where Fe was added (**Figure 6C**).

# Influence of Fe Availability on N Metabolism

Genes involved in N transport and metabolism were investigated to assess the influence of varying Fe status on N assimilation. Transcripts for genes encoding nitrate (NRT2) and ammonium (AMT) transporters were detected at all locations, with NRT2 increasing in expression by >2-fold in response to Fe addition relative to the DFB/Ctl treatment at the majority of sites in both taxa, while AMT expression varied depending on site (**Figure 5**). For instance, C4 was the only location with a >2-fold increase in AMT expression in the DFB treatment in both Pseudonitzschia and Thalassiosira. Transcripts corresponding to genes encoding components of NO<sup>−</sup> 3 assimilation, including nitrate (NR) and nitrite reductases (NIRA, NIRB, NIT-6) were generally more abundant in the treatments with added Fe, although NIRA and NIRB displayed opposite expression patterns in Pseudonitzschia and Thalassiosira at site C3 (**Figure 5**). Furthermore Pseudo-nitzschia increased gene expression of one group of nitrite reductases [NIRB and NIT-6, which use NADPH as the reductant (Brown et al., 2009)] by 11- and 3.6-fold, respectively, following added Fe while Thalassiosira conversely increased NIRB expression by 3.7-fold in the DFB treatment (**Figure 5**). In addition, Thalassiosira increased gene expression of another form of nitrite reductase (NIRA, which uses ferredoxin/flavodoxin as reductant; Brown et al., 2009) by 8-fold (p = 3 × 10−22) following Fe enrichment while Pseudo-nitzschia constitutively expressed NIRA at this location. Noticeably, transcripts for the genes encoding NIRB and NIT-6 were present in at least one of the two diatom taxa examined at all sites except the oceanic site, O5.

The relationships among transcript abundance for N uptake and assimilation-related genes, experimental sites, treatments and PN-specific VNO<sup>3</sup> measurements within the >5µm size fraction of the phytoplankton community were examined via PCA bi-plots. Principle components P1 and P2 explained 86% of the variation in N-related transcript abundance in Pseudo-nitzschia and 88% in Thalassiosira (**Figure 6A**). Sites generally contained high transcript abundances of NRT2 and NR in the added Fe treatment, with the two genes strongly co-varying with one another in both Pseudo-nitzschia and Thalassiosira. Furthermore, the Fe addition treatments at two

sites that experienced NO<sup>−</sup> 3 depletion following incubation, C2 and C3, clustered together and contained the highest AMT transcript abundance at T<sup>1</sup> and T2, respectively. Phytoplankton communities within these incubation treatments concomitantly displayed low PN-specific VNO<sup>3</sup> (0.03–0.13 day−<sup>1</sup> ; **Figure 6A**). The highest PN-specific VNO<sup>3</sup> were observed in the added Fe treatment at site C4 at T<sup>1</sup> and at C1 within the initial (T0) phytoplankton community (1.4 day−<sup>1</sup> ), which coincided with high abundances of NIRA transcripts in both genera at these locations.

# Influence of Fe Availability on C Metabolism

To further gain insight into how variable Fe status influences macronutrient resource utilization and regional biogeochemistry, genes involved in C transport and fixation were examined among sites and between diatom genera. Transcripts corresponding to a carbonic anhydrase belonging to the α-family (α-CA), involved in the carbon concentrating mechanism (CCM) within photosynthetic eukaryotes (Reinfelder, 2011), were either constitutively expressed, not detected, or more highly expressed in the DFB treatment at all locations apart from C1, where expression was 7-fold higher following Fe addition in Thalassiosira (**Figure 5**).

Members of the solute carrier (SLC) family of bicarbonate transporters (SLC4A-1, -2, and -4), which import bicarbonate ions from the environment also thought to be involved in the CCM (Nakajima et al., 2013), were detected intermittently among sites, though in low transcript abundance (**Figure 5**). These genes share sequence homology with the P. tricornutum genes PtSLC4-1, -2, and -4 (BLASTP; E < 2 × 10−69) and displayed inconsistent patterns of gene expression with each another, with no clear relationship to carbon assimilation rates. Another putative bicarbonate transporter (ICTB) was detected intermittently across sites and solely in Pseudo-nitzschia, where it was notably more highly expressed by 128-fold following Fe addition at O5. Conversely in Thalassiosira, the gene encoding phosphoenolpyruvate carboxylase (PEPC), which is part of a C4- CCM in some species of this genus (Reinfelder, 2011), was more highly expressed by 73-fold following Fe addition at O5.

Gene expression of RubisCO (RBCL) was higher by >24 fold in the Fe-limited control treatment in both genera at site O5 while at other sites the gene was either constitutively expressed, increased expression in the added Fe treatment, or not detected (**Figure 5**). In addition, other genes involved in the Calvin Cycle, including phosphoglycerate kinase (PGK), transketolase (TKL), ribulose-phosphate 3-epimerase (RPE), and phosphoribulokinase (PRK), generally increased in expression following Fe addition compared to the DFB/Ctl treatment at one or more of the three sites experiencing some degree of Fe limitation (C3, C4, and O5; Supplementary Table 1; **Figure 5**). At the CUZ sites C1 and C2, transcripts for these genes were either not differentially expressed or were more abundant in the DFB treatment within both diatom genera. Fructose-bisphosphate aldolases (FBA), involved in the Calvin Cycle, glycolysis, and gluconeogenesis, demonstrated strong Fe-dependent transcriptional patterns regardless of site and taxa (**Figure 5**). Transcripts corresponding to class II FBA, likely a metal-dependent aldolase, increased by 1.5 to 69-fold in the added Fe treatment as compared to DFB treatments with the largest fold change attributed to Pseudo-nitzschia from O5. Class II FBA has been previously demonstrated to be abundant under high-Fe conditions in diatoms and is hypothesized to contain Fe2<sup>+</sup> as a metal cofactor (Horecker et al., 1972; Allen et al., 2012; Lommer et al., 2012). Transcripts corresponding to class I FBA, the metal-independent version of class II FBA, conversely increased by 1.3 to 16-fold in DFB compared to Fe treatments.

The relationships in transcript abundance among C fixationrelated genes, experimental sites, incubated treatments and PCspecific VDIC measurements were assessed using PCA bi-plots. Principle components P1 and P2 together explained 80% of the variation in C-related transcript abundance in Pseudo-nitzschia and 78% in Thalassiosira (**Figure 6B**). Site C4 contained some of the highest PC-specific VDIC measurements within the >5µm size fraction (0.65–1.6 day−<sup>1</sup> ), and coincided with the highest transcript abundances of PGK, PRK, FBP, TKL, RPE, and GAPDH in Pseudo-nitzschia (**Figure 6B**). Conversely, Fe-limited treatments from C3 and O5 had the lowest transcript abundances of these genes in both Pseudo-nitzschia and Thalassiosira, with principle component P1 separating these samples from other sites and treatments (**Figure 6C**). Fe-limited sites C3 and O5 phytoplankton communities additionally displayed some of the lowest PC-specific VDIC observed (0.11–0.17 day−<sup>1</sup> ).

# DISCUSSION

Prior to this study, our understanding of the strategies utilized by phytoplankton to cope with low Fe bioavailability and resupply across different coastal and oceanic regions was limited. Furthermore, whether diverse diatom genera from identical environments would respond similarly when exposed to changes in Fe availability was unresolved. The gene expression patterns presented here demonstrate that the cosmopolitan diatom genera Pseudo-nitzschia and Thalassiosira rely on diverse sets of strategies to handle Fe stress, and that oceanic diatoms from both groups are highly responsive to changes in Fe availability with a greater degree of differentially expressed genes involved in nitrate assimilation, carbon fixation, and vitamin production compared to their coastal counterparts.

#### Iron-Related Gene Expression Responses across Sites

Differences in gene expression patterns in response to Fe status were observed between the coastal (C1-C4) and oceanic sites (O5) examined in this study. This included the >16-fold higher expression of genes in the added Fe treatment relative to the Fe-limited control encoding proteins involved in B<sup>7</sup> synthesis (BIOB) in both taxa, and B<sup>1</sup> (THIC) and B<sup>6</sup> (PDXK, PDLH) synthesis in Pseudo-nitzschia. These increases are consistent with previous field observations demonstrating that Fe enrichment of previously Fe-limited oceanic diatom communities stimulates B-vitamin transcript production (Cohen et al., 2017). Genes encoding an Fe storage protein (ferritin [FTN]) and components of amino acid metabolism (glutamate synthase [GLT] in Pseudo-nitzschia; ferredoxin-dependent glutamate synthase [Fd-GLT] in Thalassiosira) were similarly more highly expressed by >16-fold following Fe addition exclusively at site O5. Conversely, in the Fe-limited control, we observed the >16-fold higher expression of genes encoding the proteins Cu-Zn superoxide dismutase (Cu-Zn SOD) and RubisCO (RBCL), in either one or both taxa investigated. These distinct transcriptomic patterns of genes involved in diverse metabolic processes reflect differences in environmental factors selecting for diatom growth between the chronically Fe-limited open ocean and sporadically Fe-limited coastal regions.

In contrast, many photosynthetic genes were highly expressed following Fe addition regardless of location. A subset of these genes displayed distinct expression responses depending on whether the incubated communities experienced Fe limitation of growth rate (e.g., C3 and O5) or only Fe stress (C4; Supplementary Table 1). One such gene encodes the putative Fe transporter ABC.FEV.S, in which expression increased following Fe addition in Pseudo-nitzschia only at sites C3, C4, and O5. Additional genes include flavodoxin (FLDA) and plastocyanin (PETE), in which transcripts were generally more abundant in the DFB or Fe-limited Ctl treatments, consistent with flavodoxin's role as an Fe-independent photosynthetic electron carrier and plastocyanin's role as a Cu-dependent replacement for Fe-dependent cytochrome c6. At the Fe-stressed CUZ site C3 however, FLDA was either constitutively expressed or slightly more abundant after Fe addition, depending on the diatom genus. Plastocyanin (PETE) transcripts were similarly more abundant after Fe addition in both diatom genera at C3 and in Thalassiosira at O5. This pattern suggests coastal diatoms from higher-Fe systems tend to temporarily replace Fedependent photosynthetic proteins with Fe-independent ones, while certain diatoms in chronically Fe-limited environments may rely exclusively on the Fe-free alternatives (Marchetti et al., 2012).

# Fe-Related Gene Expression Responses Between Diatom Taxa

Pseudo-nitzschia and Thalassiosira demonstrated several distinct responses to changes in Fe status despite co-existing under identical environmental conditions. Ferredoxin (PETF), ferredoxin-dependent glutamate synthase (Fd-GLT), and ferredoxin-dependent sulfite reductase (Fd-SIR) transcripts were more abundant in Thalassiosira at oceanic site O5 following Fe addition with these responses absent in Pseudo-nitzschia. In contrast, ferredoxin-related transcripts in oceanic Pseudonitzschia were constitutively expressed or not detected. These patterns may suggest oceanic Thalassiosira strongly utilizes ferredoxin and ferredoxin-dependent proteins following Fe addition while Pseudo-nitzchia relies on Fe-independent machinery. Site O5 was additionally the only location in which Thalassiosira increased gene expression of Cu-Zn SOD under Fe-limitation. This pattern was not evident in oceanic Pseudo-nitzschia, where gene expression of this protein was constitutive, or by either genus at coastal sites, suggesting that the oceanic Thalassiosira species have distinctly evolved to rely on this Cu- and Zn-containing enzyme as the preferred superoxide dismutase in their Fe-limited environment. Pseudonitzschia conversely increased expression of Mn SOD following Fe addition, likely as a result of iron-induced increases in photosynthetic rates and photosynthetic production of superoxide radicals (Asada, 2006). These patterns highlight differences in preferred metal cofactors as a function of Fe status and transcriptional tendencies between the two taxa.

Transcripts corresponding to rhodopsin (RHO) increased in abundance within Pseudo-nitzschia in the DFB/Ctl treatments at the two sites experiencing pronounced Fe limitation (C3 and O5), but were not identified in Thalassiosira at any location. This is consistent with rhodopsin being undetected in sequenced Thalassiosira spp. transcriptomes (Marchetti et al., 2015) and supports the notion that Pseudo-nitzschia may have a competitive advantage over non-rhodopsin containing taxa, allowing for an Fe-independent alternative to photosynthesis for ATP generation during times of Fe stress. Ferritin (FTN) gene expression patterns furthermore diverged between the two taxa at the coastal sites C4 (Line-P) and C2 (CUZ). This supports laboratory findings suggesting FTN may exhibit different expression patterns among diverse phytoplankton (Marchetti et al., 2009; Botebol et al., 2015), even between taxa residing in the same location. Lastly, ABC.FEV.S, encoding a membrane Fe transport system protein, displayed divergent expression patterns between the examined genera with only Pseudo-nitzschia increasing ABC.FEV.S expression after Fe addition in all incubations exhibiting signs of iron limitation (C3, C4, and O5).

Taken together, these patterns in gene expression demonstrate that members of the pennate diatom genus Pseudo-nitzschia and the centric diatom genus Thalassiosira restructure their functional metabolisms in response to changes in Fe availability in distinct manners, possibly allowing both species to co-exist in the same environment. Both taxa are equipped with strategies to sustain growth under chronic Fe limitation in the open ocean, as supported by their equal transcript abundance during initial sampling. Following pulse Fe additions however, oceanic Pseudonitzschia relies in part on the strategies discussed above to gain a competitive advantage over Thalassiosira and quickly dominates the phytoplankton community. It remains unclear however which combination of environmental factors in the NE Pacific Ocean would select for the preferential growth of Thalassiosira over Pseudo-nitzschia. We conclude that substantial differences in molecular responses to changes in Fe status are observed across taxonomic groups, and patterns in gene expression should not be assumed universal across diverse taxa or environments.

# Nitrogen-Related Gene Expression as a Function of Fe Status

The majority of N transport and assimilation genes investigated increased in expression following Fe addition in both Pseudonitzschia and Thalassiosira. Several site- and taxa-specific patterns were identified, with some trends also possibly explained by each site's initial NO<sup>−</sup> 3 concentration. For example, most gene copies encoding the NO<sup>−</sup> 3 transporter, NRT2, have been demonstrated in laboratory cultures to increase in expression in NO3-stressed diatoms (Bender et al., 2014; Rogato et al., 2015), and transcripts corresponding to this gene were some of the most abundant in both Pseudo-nitzschia and Thalassiosira at C2—the CUZ site where NO<sup>−</sup> 3 concentrations were depleted in all incubations by the first sampling time point. This gene also showed expression trends that correlated with Fe status; NRT2 transcripts were more abundant after Fe addition at all locations, regardless of initial NO<sup>−</sup> 3 concentrations. Based on these observations, NRT2 in diatoms also appears to be linked to Fe status and follows the expression of other N-related genes involved in Fe-dependent NO<sup>−</sup> 3 assimilation, including those encoding nitrate reductase (NR) and nitrite reductase (NIRA; Marchetti et al., 2012).

Diatoms were perhaps relying on NH<sup>4</sup> in place of NO<sup>−</sup> 3 as a source of N based on gene expression patterns at several CUZ sites. Fe-enriched treatments at C2 contained the lowest NO<sup>−</sup> 3 after 48 h of incubation (0.06 µmol L−<sup>1</sup> ), and the genes encoding the ammonium transporters AMTs concomitantly increased in expression in the Fe relative to DFB treatment (**Figure 6**). Furthermore at C3, Fe-enriched communities entered NO<sup>−</sup> 3 stress by the end of the incubation period, and AMT expression simultaneously increased in both Pseudo-nitzschia and Thalassiosira. This negative relationship between NO<sup>−</sup> 3 concentrations and AMT transcript abundance in natural diatom assemblages is consistent with those in laboratory Pseudonitzschia multiseries and Fragilariopsis cylindrus cultures (Bender et al., 2014; Rogato et al., 2015), and is reported here as one of the first observations of this relationship in natural phytoplankton communities.

High AMT transcript abundance at some of these locations may also represent NH<sup>4</sup> rather than NO<sup>−</sup> 3 being preferred as an N source by Fe-stressed diatoms conserving their cellular Fe supply, as NO<sup>−</sup> 3 assimilation depends on various Fe-dependent processes (Milligan and Harrison, 2000). This is supported by the increased expression of AMT transcripts in both Pseudonitzschia and Thalassiosira from the Fe-stressed coastal Line-P incubations at C4. Pseudo-nitzschia from the Fe-limited site O5 also exhibited this pattern whereas Pseudo-nitzschia from C3 and Thalassiosira from both C3 and O5 did not, suggesting other environmental parameters aside from Fe status are influencing whether diatoms utilize NH4- or NO3-specific N uptake pathways.

Similar to our Fe-related gene expression results, several N-related genes demonstrated divergent expression responses between Pseudo-nitzschia and Thalassiosira. Expression of the NO<sup>−</sup> 2 reductase genes, NIRA and NIRB, displayed opposite patterns between the two genera at the CUZ site where Fe-stress occurred in incubations (C3), with Pseudo-nitzschia highly expressing the gene encoding non-ferredoxin-utilizing NO<sup>−</sup> 2 reductase (NIRB) following Fe addition, and Thalassiosira highly expressing the gene encoding the ferredoxin-utilizing nitrite reductase (NIRA). Furthermore at site O5, Pseudo-nitzschia increased expression of AMT and NADPH-dependent glutamate synthase (GLT) following Fe addition while Thalassiosira increased expression of NRT2 and ferredoxin-dependent glutamate synthase (Fd-GLT). These transcriptomic patterns may suggest Pseudo-nitzschia continues to rely on the non-Fe requiring metabolic pathways for assimilating N once Fe becomes available (AMT, NIRB, GLT), whereas Thalassiosira shifts over to Fe-dependent ones (NRT2, NIRA, Fd-GLT) upon Fe resupply.

These expression patterns furthermore support that substantial variations exist between the two diatom taxa in terms of N acquisition and assimilation strategies following changes in Fe supply. Both Pseudo-nitzschia and Thalassiosira are equipped with distinct strategies to compete under a variety of Fe and N conditions, and this may contribute to how multiple diatom species relying upon the same limiting resources in identical environments co-exist (i.e., paradox of the plankton; Hutchinson, 1961). These patterns are consistent with previous reports of resource partitioning among diatoms based on N and phosphate utilization (Alexander et al., 2015). Varying environmental pressure likely maintain populations of diverse diatom genera in the open ocean, with certain species outcompeting others depending on specific sets of external factors, including both macro- and micronutrients (Godhe and Rynearson, 2017).

# Carbon-Related Gene Expression Responses as a Function of Fe Status

Genes encoding proteins involved in C uptake and assimilation were surveyed in order to determine the influence of Fe addition or stress on C metabolism. We observed site-specific expression patterns of the diatom RubisCO large subunit protein (RBCL), where gene expression was substantially elevated at site O5 in the Fe-limited control treatment relative to the Fe addition response in both diatom genera. A sequence analysis of RubisCO contigs obtained across experimental sites demonstrates that O5 protein sequences are structurally less similar to known Pseudo-nitzschia and Thalassiosira RubisCO protein sequences within the MMETSP database than those at the four coastal sites (Supplementary Figure 4A; Supplementary Table 3). This distinction in both protein structure and transcriptional expression may indicate a distinct adaptation and utilization of RubisCO in the oceanic diatoms than in those from high-Fe coastal waters. Phylogenetically diverse diatom species have been demonstrated to vary in their RubisCO enzyme kinetics in laboratory cultures, with their RubisCO content inversely linked to the strength of their carbon concentrating mechanism (CCM; Young et al., 2016). The CCM increases CO<sup>2</sup> concentrations in chloroplast stroma in the vicinity of RubisCO and is fueled by the energy (ATP) generated from the Fe-intensive process of photosynthesis (Reinfelder, 2011; Young et al., 2016). We hypothesize that chronically Fe-limited oceanic diatoms are ATP-limited by the scarcity of Fe needed to support photosynthesis, and instead increase their RubisCO protein content to maintain high rates of carbon fixation rather than allocate scarce energy resources to the CCM. Furthermore, the genes encoding a putative bicarbonate transporter (ICBT) and a C4-CCM component (PEPC; Reinfelder et al., 2000; Sage, 2004; Reinfelder, 2011) were highly expressed following Fe addition in Pseudo-nitzschia and Thalassiosira, respectively, exclusively at O5. This supports that diatoms may be capable of shuffling energy pools into either the CCM or RubisCO production depending on Fe bioavailability. Interestingly, in laboratory-based proteomic analyses with cultures of the coastal diatom T. pseudonana, RubisCO was similarly more highly expressed under Fe limitation, while PEPC protein levels were higher under Fe-replete conditions (Nunn et al., 2013). Consistent with our hypothesis, Hopkinson et al. (2010) attributed increases in biomass following CO2-enrichment of an Fe-limited phytoplankton community in the HNLC Northeast Pacific Ocean to downregulation of the CCM in order to conserve iron and photosynthetically-produced energy. Laboratory-based RubisCO kinetic work with cultured diatom isolates is needed to confirm whether diatoms from HNLC regions minimize their photosynthetic demand for Fe by synthesizing more RubisCO enzymes rather than allocating scarce energy resources into the CCM.

Other C fixation-related gene expression patterns were largely consistent with C assimilation rates, and generally varied as a function of both Fe status and ocean province. The genes PGK, TKL, RPE, and PRK did not exhibit site-specific expression patterns similar to RBCL, and instead increased in expression following Fe enrichment at sites where Fe addition increased C assimilation rates (C3, C4, and O5). Increased expression of these genes is expected with Fe stimulation of C-fixation and growth. These expression patterns are in agreement with laboratory cultures of the diatom P. tricornutum, which increased expression of genes involved in C fixation during the light portion of their diel cycle, when DIC is being taken up to support photosynthesis (Chauton et al., 2013).

# CONCLUSION

Gene expression characterization coupled with biological rate processes across geographically diverse communities suggests regional and taxa-specific strategies are utilized by diatoms when rapidly responding to variations in environment. Our analysis demonstrates that chronically Fe-limited oceanic diatoms will restructure Fe, N, and C metabolism in a distinctive manner following Fe addition when compared to the response of coastal diatom communities that receive inherently more variable Fe inputs. Pseudo-nitzschia and Thalassiosira, two cosmopolitan diatom taxa found at all locations investigated, at times demonstrated divergent transcriptomic responses to changes in Fe status in terms of photosynthetic processes and N metabolism, even under identical environmental conditions.

Potential limitations to our approach include gene expression analyses being conducted on specific diatom genera while the physiological rate process measurements correspond to bulk phytoplankton communities. We therefore assumed the physiological characteristics to be representative of all phytoplankton members present. Furthermore, the metatranscriptomic approach used here consisted of analyzing cumulative expression responses of pooled gene copies; however, distinct gene copies have been shown to vary in their transcriptional response to environmental conditions within a single organism (Bender et al., 2014; Levitan et al., 2015; Rogato et al., 2015). In order to gain further resolution, we recommend laboratory-based studies be performed investigating the direct relationships between nutrient uptake rates and expression of specific gene copies encoding proteins involved in nutrient assimilation in distinct members from each of the genera Pseudo-nitzschia and Thalassiosira.

The findings presented here support the notion that a tremendous degree of genetic diversity is contained within the diatom lineage, and this may strongly influence the abundance and distribution of phytoplankton communities. Since Fe bioavailability to phytoplankton is predicted to change with increasing temperature and acidification of surface seawater (Shi et al., 2010; Sunda, 2010; Capone and Hutchins, 2013; Hutchins and Boyd, 2016), these findings will aid in predicting the consequences of changing ocean conditions on phytoplankton productivity and community growth dynamics.

# AUTHOR CONTRIBUTIONS

AM, BT, and KB designed the study; NC, KE, BT, and AM performed the incubation experiments; NC conducted the metatranscriptomic and physiological analysis; RL provided bioinformatic support; FK and KT obtained photophysiological measurements onboard the R/V Melville; MB and HM quantified biogenic silica; MM provided primary productivity measurements; CT and BT quantified trace metals; WS contributed to iron metabolism interpretations; SB quantified domoic acid; NC and AM wrote the manuscript. All authors contributed to intellectual content and approved the final manuscript.

# FUNDING

Research was funded through National Science Foundation grants OCE-1334935 to AM, OCE-1334632 to BT, OCE-1334387 to MB, OCE-1333929 to KT, and OCE-1259776 to KB, as well as through Discovery NSERC grant 261521-13 to MM.

# ACKNOWLEDGMENTS

We thank J. Roach (UNC), S. Haines (UNC), W. Gong (UNC), M. Kanke (UNC), S. Davies (UNC), and M. Love (UNC) for sequencing analysis advice and guidance. P. Morton (FSU) analyzed the dissolved Fe samples from the Line P cruise (C4 and O5). Dissolved nutrients were analyzed by T. Coale (SIO) onboard the R/V Melville (C1-C3), and by M. Belton (IOS) onboard the CCGS J.P. Tully (C4 & O5). Z. Li (Duke) provided climatological remote sensing images in **Figure 1**. We are grateful to the scientists and crew of the CCGS J.P. Tully (Line-P cruise 2015-09) and the R/V Melville (cruise 1405) for their support and assistance at sea. RNA-Seq data was processed using UNC's Research Computing clusters. Finally, we would like to thank Carolyn Duckham for her help collecting and processing samples for DIC uptake rates at stations C1–C3.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars. 2017.00360/full#supplementary-material

#### REFERENCES


export in a frontal zone within the California Current. J. Geophys. Res. Ocean. 120, 4654–4669. doi: 10.1002/2015JC010829


Iron binding at the cell surface is a critical step. Plant Physiol. 160, 2271–2284. doi: 10.1104/pp.112.204156.


**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 Cohen, Ellis, Lampe, McNair, Twining, Maldonado, Brzezinski, Kuzminov, Thamatrakoln, Till, Bruland, Sunda, Bargu and Marchetti. 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.

# Corrigendum: Diatom Transcriptional and Physiological Responses to Changes in Iron Bioavailability across Ocean Provinces

Natalie R. Cohen<sup>1</sup> , Kelsey A. Ellis <sup>1</sup> , Robert H. Lampe<sup>1</sup> , Heather McNair <sup>2</sup> , Benjamin S. Twining<sup>3</sup> , Maria T. Maldonado<sup>4</sup> , Mark A. Brzezinski <sup>2</sup> , Fedor I. Kuzminov <sup>5</sup> , Kimberlee Thamatrakoln<sup>5</sup> , Claire P. Till 6,7, Kenneth W. Bruland<sup>6</sup> , William G. Sunda<sup>1</sup> , Sibel Bargu<sup>8</sup> and Adrian Marchetti <sup>1</sup> \*

*<sup>1</sup> Department of Marine Sciences, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, <sup>2</sup> The Marine Science Institute and the Department of Ecology Evolution and Marine Biology, University of California, Santa Barbara, Santa Barbara, CA, United States, <sup>3</sup> Bigelow Laboratory for Ocean Sciences, East Boothbay, ME, United States, <sup>4</sup> Department of Earth, Ocean, and Atmospheric Sciences, University of British Columbia, Vancouver, BC, Canada, <sup>5</sup> Department of Marine and Coastal Sciences, Rutgers, The State University of New Jersey, New Brunswick, NJ, United States, <sup>6</sup> Department of Ocean Sciences, University of California, Santa Cruz, Santa Cruz, CA, United States, <sup>7</sup> Chemistry Department, Humboldt State University, Arcata, CA, United States, <sup>8</sup> Department of Oceanography and Coastal Sciences, School of the Coast and Environment, Louisiana State University, Baton Rouge, LA, United States*

Keywords: diatoms, Thalassiosira, Pseudo-nitzschia, iron, metatranscriptomics, California upwelling zone, Northeast Pacific Ocean

#### **A corrigendum on**

#### **Diatom Transcriptional and Physiological Responses to Changes in Iron Bioavailability across Ocean Provinces**

by Cohen, N. R., Ellis, K. A., Lampe, R. H., McNair, H., Twining, B. S., Maldonado, M. T., et al. (2017). Front. Mar. Sci. 4:360. doi: 10.3389/fmars.2017.00360

In the original Acknowledgments of this article, we neglected to thank Carolyn Duckham for her help collecting and processing samples for DIC uptake rates at stations C1-C3. The authors regret this oversight. This error does not change the scientific conclusions of the article in any way. The original article has been updated.

**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 © 2018 Cohen, Ellis, Lampe, McNair, Twining, Maldonado, Brzezinski, Kuzminov, Thamatrakoln, Till, Bruland, Sunda, Bargu and Marchetti. 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.

#### Approved by:

*Marine Science Editorial Office, Frontiers, Switzerland*

> \*Correspondence: *Adrian Marchetti amarchetti@unc.edu*

#### Specialty section:

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

Received: *16 March 2018* Accepted: *19 March 2018* Published: *05 April 2018*

#### Citation:

*Cohen NR, Ellis KA, Lampe RH, McNair H, Twining BS, Maldonado MT, Brzezinski MA, Kuzminov FI, Thamatrakoln K, Till CP, Bruland KW, Sunda WG, Bargu S and Marchetti A (2018) Corrigendum: Diatom Transcriptional and Physiological Responses to Changes in Iron Bioavailability across Ocean Provinces. Front. Mar. Sci. 5:115. doi: 10.3389/fmars.2018.00115*

# Algal Hot Spots in a Changing Arctic Ocean: Sea-Ice Ridges and the Snow-Ice Interface

Mar Fernández-Méndez <sup>1</sup> \*, Lasse M. Olsen<sup>1</sup> , Hanna M. Kauko<sup>1</sup> , Amelie Meyer <sup>1</sup> , Anja Rösel <sup>1</sup> , Ioanna Merkouriadi <sup>1</sup> , Christopher J. Mundy <sup>2</sup> , Jens K. Ehn<sup>2</sup> , A. Malin Johansson<sup>3</sup> , Penelope M. Wagner <sup>4</sup> , Åse Ervik 5,6, Brian K. Sorrell <sup>7</sup> , Pedro Duarte<sup>1</sup> , Anette Wold<sup>1</sup> , Haakon Hop1,8 and Philipp Assmy <sup>1</sup>

<sup>1</sup> Norwegian Polar Institute, Fram Centre, Tromsø, Norway, <sup>2</sup> Centre for Earth Observation Science, University of Manitoba, Winnipeg, MB, Canada, <sup>3</sup> Department of Physics and Technology, UiT The Arctic University of Norway, Tromsø, Norway, <sup>4</sup> Norwegian Ice Service, Norwegian Meteorological Institute, Tromsø, Norway, <sup>5</sup> Sustainable Arctic Marine and Coastal Technology, Centre for Research-based Innovation, Norwegian University of Science and Technology, Trondheim, Norway, <sup>6</sup> The University Centre in Svalbard, Longyearbyen, Norway, <sup>7</sup> Department of Bioscience, Aarhus University, Aarhus, Denmark, <sup>8</sup> Department of Arctic and Marine Biology, Faculty of Biosciences, Fisheries and Economics, UiT The Arctic University of Norway, Tromsø, Norway

#### Edited by:

Letizia Tedesco, Finnish Environment Institute (SYKE), Finland

#### Reviewed by:

Urania Christaki, Université du Littoral Côte d'Opale, France Kevin Arrigo, Stanford University, United States Maria A. Van Leeuwe, University of Groningen, Netherlands

#### \*Correspondence:

Mar Fernández-Méndez mar.fdez.mendez@gmail.com; mar@npolar.no

#### Specialty section:

This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science

Received: 15 July 2017 Accepted: 20 February 2018 Published: 12 March 2018

#### Citation:

Fernández-Méndez M, Olsen LM, Kauko HM, Meyer A, Rösel A, Merkouriadi I, Mundy CJ, Ehn JK, Johansson AM, Wagner PM, Ervik Å, Sorrell BK, Duarte P, Wold A, Hop H and Assmy P (2018) Algal Hot Spots in a Changing Arctic Ocean: Sea-Ice Ridges and the Snow-Ice Interface. Front. Mar. Sci. 5:75. doi: 10.3389/fmars.2018.00075 During the N-ICE2015 drift expedition north-west of Svalbard, we observed the establishment and development of algal communities in first-year ice (FYI) ridges and at the snow-ice interface. Despite some indications of being hot spots for biological activity, ridges are under-studied largely because they are complex structures that are difficult to sample. Snow infiltration communities can grow at the snow-ice interface when flooded. They have been commonly observed in the Antarctic, but rarely in the Arctic, where flooding is less common mainly due to a lower snow-to-ice thickness ratio. Combining biomass measurements and algal community analysis with under-ice irradiance and current measurements as well as light modeling, we comprehensively describe these two algal habitats in an Arctic pack ice environment. High biomass accumulation in ridges was facilitated by complex surfaces for algal deposition and attachment, increased light availability, and protection against strong under-ice currents. Notably, specific locations within the ridges were found to host distinct ice algal communities. The pennate diatoms Nitzschia frigida and Navicula species dominated the underside and inclined walls of submerged ice blocks, while the centric diatom Shionodiscus bioculatus dominated the top surfaces of the submerged ice blocks. Higher light levels than those in and below the sea ice, low mesozooplankton grazing, and physical concentration likely contributed to the high algal biomass at the snow-ice interface. These snow infiltration communities were dominated by Phaeocystis pouchetii and chain-forming pelagic diatoms (Fragilariopsis oceanica and Chaetoceros gelidus). Ridges are likely to form more frequently in a thinner and more dynamic ice pack, while the predicted increase in Arctic precipitation in some regions in combination with the thinning Arctic icescape might lead to larger areas of sea ice with negative freeboard and subsequent flooding during the melt season. Therefore, these two habitats are likely to become increasingly important in the new Arctic with implications for carbon export and transfer in the ice-associated ecosystem.

Keywords: Arctic ecosystem, ice algae, phytoplankton, infiltration communities, sea-ice ridges, community composition, climate change

# INTRODUCTION

Current changes in sea-ice conditions have consequences for algal biomass and growth, with bottom-up cascading effects on the Arctic marine food web (Wassmann et al., 2011). The significant decline in sea-ice extent and thickness during the last 30 years has caused an increase in the light available for phytoplankton (Arrigo and van Dijken, 2011; Bélanger et al., 2013) and, thus, an increase in phytoplankton net annual primary production (Arrigo and van Dijken, 2015). Likewise, there have been several reports of under-ice phytoplankton blooms in the recent years enabled by the increased light transmission through melt ponds (e.g., Mundy et al., 2009; Arrigo et al., 2012) or through leads (Assmy et al., 2017). In contrast, ice algal areal production is probably decreasing on a pan-Arctic scale due to the loss of sea-ice habitat (Dupont, 2012). In addition, biomass standing stocks are low in young ice compared to the disappearing older ice, probably limited by recruitment, adding to the reduction in sea-ice algal areal production (Lange et al., 2017a; Olsen et al., 2017). As the ice edge retreats further north each summer, ice algae will be limited to the stratified deep basins of the Central Arctic with more oligotrophic conditions compared to the more productive shelves (Barber et al., 2015). The trend toward earlier ice melt and later ice formation may furthermore cause a mismatch in the timing between primary and secondary producers, diminishing the amount of carbon and energy transferred up the food chain (Søreide et al., 2010; Leu et al., 2011; Ji et al., 2013).

Diatoms typically dominate both the phytoplankton and the sea-ice spring blooms, while flagellates, dinoflagellates, and picoeukaryotes usually dominate in late summer (Tremblay et al., 2009; Moran et al., 2012; van Leeuwe et al., 2018). Some diatom species, such as Shionodiscus bioculatus (formerly Thalassiosira bioculata) (Alverson et al., 2006) and Fragilariopsis cylindrus, are sea-ice associated and have been observed both in the water column and in the ice (von Quillfeldt, 2000). Other seaice specialists such as Nitzschia frigida and Melosira arctica grow attached to the ice, while Chaetoceros gelidus (formerly Chaetoceros socialis) (Chamnansinp et al., 2013), Fragilariopsis oceanica and the haptophyte P. pouchetii are typically found in the water column (Booth and Horner, 1997). Current estimates of algal biomass and production in the ice-covered Arctic Ocean generally include phytoplankton and less often sea-ice algae (Gosselin et al., 1997; Sakshaug et al., 2004). Only recent studies have quantified the contribution of other sea-ice related environments, such as melt ponds (Mundy et al., 2011; Lee et al., 2012; Fernández-Méndez et al., 2015), and other more elusive forms of algal accumulations under the ice such as floating algal aggregates (Assmy et al., 2013; Fernández-Méndez et al., 2014).

There are few observations of ice algae growing in ridges (Syvertsen, 1991; Hegseth, 1992; Legendre et al., 1992) and at the snow-ice interface in the Arctic (Buck et al., 1998; McMinn and Hegseth, 2004; von Quillfeldt et al., 2009). Ridges are known to be hot spots for biological activity since they act as shelters for ice fauna and ice-associated zooplankton (Hop and Pavlova, 2008; Gradinger et al., 2010) and juvenile polar cod (Gulliksen and Lønne, 1989). Ridges have also been recently identified as locations of high algal biomass using under-water remotely operated vehicles (Lange et al., 2017b). However, due to the sampling challenges that these complex structures pose, algae have only been sampled sporadically. Snow infiltration communities growing at the snow-ice interface, have been widely described for Antarctic pack ice (Horner et al., 1988; Spindler, 1994; Robinson et al., 1997; Kristiansen et al., 1998; Garrison et al., 2003), where they contribute substantially to sea-ice primary production (Arrigo et al., 1997). In the few observations obtained from the Arctic, the dominant species reported are mostly phytoplankton such as P. pouchetii in pack ice north of Svalbard and Svalbard fjords (McMinn and Hegseth, 2004; von Quillfeldt et al., 2009), and unidentified pennate and centric diatoms in Disco Island, Greenland (Buck et al., 1998).

Despite these important observations, algal communities growing in ridges and at the snow-ice interface are understudied in the Arctic. Published studies of these two environments mainly focused on a qualitative assessment of the algal species present (especially in ridges), and the photosynthetic performance of the snow infiltration community in the study by McMinn and Hegseth (2004). During the Norwegian young sea ICE (N-ICE2015) drift expedition, we followed the evolution of these communities over 6 weeks and were able to characterize the physical-chemical environment in which these algal communities thrive, and we explain why these environments are suitable habitats for Arctic microalgae.

The aim of this study is to characterize sea-ice ridges and snow-ice interfaces as potential habitats and refuges for algae in the Arctic Ocean. In particular, we assess the importance of their biomass compared to adjacent environments, we define the light and nutrient regimes that these communities experience, we assess their photosynthetic activity, and we describe the species present. Furthermore, we discuss the role of these environments for hosting algae in the future Arctic Ocean against the background of the ongoing and predicted changes in the Arctic icescape.

# MATERIALS AND METHODS

#### Sampling

All samples were collected during the N-ICE2015 drift expedition that took place between January and June 2015 in ice-covered waters north-west of Svalbard (Granskog et al., 2016). In total four ice floes were occupied and monitored during the expedition. Data presented in this study were obtained during drifts of Floe 3 and 4 (**Figure 1A**). Sea-ice algae present in ridges were sampled during the drift of Floe 3 between 10 May and 3 June 2015. Between 10 and 18 May, scuba divers using a slurp gun (modified 3.5 L Trident <sup>R</sup> suction gun) collected samples from the surface of the submerged ledges on the thin ice side every other day (side labeled with a star in **Figures 1B**, **2**). These samples were used for algal biomass, physiology, and community analysis. Slurp gun sampling can potentially lead to loss of algal biomass, however it can be considered the most appropriate method to sample these surfaceattached algal layers. To use these samples quantitatively, the area sampled on the ledge's surface was measured (0.05 ×

0.54 m) and used to estimate areal biomass. On 28 May, 31 May and 3 June, sea-ice algae at the ridge were sampled by ice coring with a 9-cm diameter ice corer (Mark II coring system, KOVACS Enterprise, Roseburg, USA). Bottom and top 0.1 m of the cores were collected on 28 May and entire cores of submerged sea-ice ledges were collected in three pieces with the ice corer on 30 May and 3 June for chlorophyll (Chl) a measurements and quantitative taxonomic analysis at both sides of the ridge. Melting of the ice cores occurred in the dark without addition of filtered seawater to avoid the addition of nutrients.

Algae growing at the snow-ice interface were sampled on Floe 4 between 9 and 18 June. Snow was removed with a shovel to search for brownish coloration as an indicator for algae at random areas with negative freeboard and high snow accumulation. In a radius of 500 m around the ship, we found and sampled these dense algae accumulations at eight different locations, usually along cracks in the ice (**Figures 1C**, **3**). Samples for qualitative analysis were taken with a snow shovel and melted in clean wide-necked plastic buckets. On 9 June, samples for quantitative analyses were taken using the bottom part of the ice corer and a plastic plate to close the bottom once it was filled with slush.

## Characterization of the Physical Setting: Sea Ice and Snow

The ridge we chose for this study was a typical first-year ice (FYI) ridge (based on the characterization of its physical properties by Ervik et al., under review) that had formed adjacent to a refrozen lead as we started sampling Floe 3 in late April. We were able to follow its progression for a month. The internal ridge structure

The pink square indicates the low biomass side of the ridge. (C) Aerial image of the study area during Floe 4 (image taken on 14 June 2015) and locations of snow-ice

interface sampling. Vasilii Kustov and Sergey Semenov (Arctic and Antarctic Research Institute. St. Petersburg. Russia).

FIGURE 2 | Scheme of a first-year ice ridge based on observations and measurements performed during May 2015. The square and star correspond to the sampling sites indicated on Figure 1B. The main water current below the ice is a simplification from Figure 5C. The transmitted irradiance is depicted in a qualitative way to show the reflections that occur inside ridge cavities where the light might be higher than below the ridge itself. The most abundant algal species at the distinct surfaces of the ledges are depicted in the circles to the right based on Figure 3.

was determined by drilling holes with a 0.051 m auger along a transect perpendicular to the ridge length, as described in Ervik et al. (under review). To calculate the ridge macro-porosity (ratio of voids filled with water or slush to the total thickness of the ice) of the unconsolidated part (rubble), we added up the lengths of all the voids inside the rubble ice and divided by the total lengths of all the drill holes inside the rubble. Six videos of the underwater part of the ridge were recorded with a GoPro Hero black on 25 and 28 May (a compilation of these videos can be found in the Supplementary Material).

Ridge and rubble ice coverage, as well as smooth ice, new ice and open water percentages were assessed from satellite scenes. Five Radarsat-2 scenes of areas located within 10 km of the research vessel's position from 25 May until 15 June were processed and the percentage of deformed ice was estimated (Table S1, Figure S1). Radarsat-2 scenes use the standard frequency (C-band) for operational sea-ice monitoring and have successfully been used to separate deformed, FYI and multiyear ice (MYI) (Casey et al., 2014). The satellite imagery used here are fully polarimetric scenes with a high spatial resolution (5 m). The scenes were radiometrically calibrated using the included metadata calibration information (MacDonald, 2016), and subsequently segmented using the "extended polarimetric feature space" algorithm (Doulgeris and Eltoft, 2010; Doulgeris, 2013). The segmentation algorithm separated each image into distinct categories based on the statistical properties of the texture features. The identification and classification into open water, new and young ice, smooth ice, and ridges and rubble ice followed procedures used by operational ice analysts and documented in MANICE [Canadian Ice Service (CIS) Meteorological Service of Canada, 2005]. The total percentage of each category type was estimated once the segments had been combined into classified areas.

Water currents below the ice close to the ridge were measured with a medium-range vessel-mounted broadband 150 kHz acoustic Doppler current profiler (ADCP; Teledyne RD Instruments, Poway, CA, USA). Profiles were averaged hourly in 8-m vertical bins with the first bin centered at 23 m (Meyer et al., 2017). Current speed and direction at 23 m depth were used to analyze the current dynamics relative to the ridge during Floe 3 based on the ship's navigation data. The 23 m depth current data from the vessel-mounted ADCP were the shallowest current data set available for the study time period and were validated by comparing with near surface (1 m depth) current speed available for part of the time period from Acoustic Doppler Velocimeter instruments (ADV; Sontek Xylem, San Diego, CA, USA).

Snow depth and ice thickness on Floe 4 were determined using an electromagnetic instrument (EM31) in combination with a GPS-snow probe as described in Rösel et al. (2018). Negative freeboard areas that could potentially be flooded through cracks in the ice were estimated based on data from snow and ice thickness transects within a radius of 5 km around the ship (Rösel et al., 2016a,b) and drill hole data (Rösel and King, 2017). Additionally, a snow pit was dug and analyzed on 13 June at the first location (SI1) (**Figure 1C**), where we sampled the snow infiltration communities. Density, temperature, hardness and grain size of the snow were determined at 0.1 m intervals (Gallet et al., 2017).

#### Light Measurements and Calculations

Transmitted irradiance below ridges was measured during Floe 3 with a vLBV300 remotely operated vehicle (ROV) (SeaBotix.Inc, San Diego, CA, USA). The amount of transmitted photosynthetically active radiation (PAR) available below the studied ridge was measured using a cosine-corrected hyperspectral irradiance sensor (HyperOCR, Satlantic, Halifax, Canada) mounted on the upper part of the ROV. The same type of sensor was mounted on the surface of the ice looking upwards to measure incoming irradiance. Simultaneous measurements with both sensors allowed for transmittance estimates. In total, 334 radiation measurements at <5 m depth below the ridge were performed during 7, 18, and 20 May. Moreover, based on the observations by divers and the videos from the ROV's camera (600TVL color), as well as with an underwater camera attached to a pole and deployed through a hole in the ice, we could qualitatively assess the light field inside the ridge.

In addition, we used the following modification of the equation by Light et al. (2008) to calculate light transmitted (PARz) through the ridge with three overlaid ice ledges, separated by voids with water:

$$\begin{split} PAR\_{\rm z} &= (1 - R) \times PAR\_{\rm surface} \times \exp[-K\_{\rm snow} \times Z\_{\rm snow} \\ &- K\_{ic\varepsilon} \times (Z\_{ic\varepsilon\_1} + Z\_{ic\varepsilon\_2} + Z\_{ic\varepsilon\_3}) - K\_{water} \times (Z\_{water\_1} - Z\_{\rm water\_2} - Z\_{\rm water\_1}) \\ &+ Z\_{\rm water\_2}) \end{split}$$

where R is the specular reflection that happens at the surface (5%) (Perovich, 1989), PARsurface is the incoming PAR from a Trios-Sensor located at the weather station on the ice camp (Hudson et al., 2016), Ksnow is the snow light attenuation coefficient for PAR (14.82 m−<sup>1</sup> ), Kice is the ice light attenuation coefficient (0.93 m−<sup>1</sup> ), Kwater is the water light attenuation coefficient (0.1 m−<sup>1</sup> ), and Z is thickness of the three different ledges or the depth of the water voids in between them. The snow attenuation coefficient was calculated from time series of incident and transmitted PAR and the sea-ice light attenuation coefficient was taken from Light et al. (2008). To compare with the ROV under-ice measurements, we calculated the light transmitted at the side of the ridge facing the refrozen lead (marked with a star in **Figures 1B**, **2**) from 23 April to 5 June, using its minimum (0.07 m) and maximum (0.11 m) snow depths measured on 28 May. In addition, to obtain an idea of the spatial variability of light transmitted through the ridge, we calculated PAR transmitted at 1-m intervals where we measured snow depth, ice thickness and water voids on 24 and 31 May.

The amount of light available for the snow infiltration communities was measured with a scalar Mini PAR logger (JFE MKV-L, Japan). In addition, we calculated the transmitted irradiance below 0.2–0.7 m of snow using the measured incoming irradiance and the snow attenuation coefficient mentioned above.

#### Chemical and Biological Analysis

Inorganic nutrients (nitrate, phosphate, and silicic acid) were sampled at 5 m below the ridge and at the snow-ice interface, collected in 20 mL scintillation vials, fixed with 0.2 mL chloroform and stored refrigerated until sample analysis ∼6 months later. Nutrients were measured spectrophotometrically on a modified Scalar auto-analyzer following Bendschneider and Robinson (1952) for nitrate, and Grasshoff (1965), for phosphate and silicic acid. The measurement uncertainty was 10% or less for all nutrients. Ammonium, which can reach very high concentrations in sea ice, was unfortunately not measured in these samples. In order to elucidate nitrogen remineralization in these high algal biomass environments, it should be measured in future studies. Nutrient concentrations in the water column at 5 m depth are available at the Norwegian Polar Data Centre (Assmy et al., 2016).

For chlorophyll a (Chl a) and particulate organic carbon and nitrogen (POC and PON) 10–200 mL of sample (depending on the coloration of the melted sea-ice sample) were filtered through GFF and pre-combusted GFF filters (diameter 25 mm; Whatman, GE Healthcare, Little Chalfont, UK), respectively. Chl a was extracted in 5 mL of 100% methanol at 5◦C in the dark for 12 h and measured fluorometrically using a Turner 10-AU Fluorometer (Turner Designs, San Jose, USA). POC and PON samples were analyzed with continuous-flow mass spectrometry (CF-IMRS) using a Roboprep/tracermass mass spectrometer (Europa Scientific, UK).

To calculate the percentage of algal biomass that each environment was contributing to the total sea-ice biomass we multiplied the percentage of surface that each environment (e.g., ridges and deformed ice, deformed edges next to open water or young ice, flooded FYI; non-flooded FYI or second-year ice (SYI) and young ice) covered by the range of biomass measured in each environment.

To calculate nutrient demand we followed Cota et al. (1987) and used our measured Chl a concentrations, the N:Chl a and Si:Chl a ratios, and the calculated growth rate based on Chl a measurements taken over consecutive days. Furthermore, we calculated the nutrient replenishment rate (mmol m−<sup>2</sup> d −1 ) by multiplying the measured nutrient concentrations in the underice water (transformed from per cubic meter to per square meter) by the measured water current velocity below the ice.

The physiological status of the photosynthetic apparatus of the algae was assessed with Pulse Amplitude Modulation (PAM) fluorometry using a Phyto-PAM Phytoplankton Analyzer (Walz, Eiffeltrich, Germany). Samples from the ridge were carefully collected by divers every 2 days between 10 and 18 May using a slurp gun, and between 28 and 31 May by scraping the surface of the ice core (the top and the bottom) into filtered seawater. Snow-ice infiltration layer samples for PhytoPAM analysis were collected with a clean bucket on the 9, 10, 11, 13, and 14 June. The quantum yield (8PSII) of photosystem II fluorescence was determined on 30-min dark-acclimated samples from the ratio of variable and maximal fluorescence (Fv/Fm). In addition, Rapid Light Curves (RLCs) were performed with 20 sec pulses of actinic light ranging between 1 and 900 µmol photons m−<sup>2</sup> s −1 in 13 steps. The relative photosynthetic electron transport rate (rETR) was calculated as the product of 8PSII, the theoretical absorption of PSII and the scalar irradiance of PAR at each pulse. The RLCs were fitted using the equation of Webb et al. (1974) to yield data from which the initial slope (α), the maximum rETR, and the photoacclimation parameter (Ek) were derived. There was no evidence of photoinhibition in any RLCs, so no photoinhibitory modification was included in the model. Only photosynthetic parameters obtained from the blue excitation channel (470 nm) were used, to optimize the signal-to-noise ratio and due to the strong absorption by Chl c, fucoxanthin and carotenoids in blue light by diatoms, which were the dominant algal group in our samples (Walz, 2003; Johnsen and Sakshaug, 2007). To statistically test for differences in the photosynthetic parameters of the different algal communities in the ridges we used the ANCOVA test for comparison of regression lines; (Sokal and Rohlf, 2012).

An additional approach used to test whether the diatoms found in the ridges and the snow-ice interface were actively growing was the silica stain method (McNair et al., 2015). We added 100 µL of the fluorescent dye 2-(4 pyridyl)-5-((4-(2-dimethylaminoethylaminocarbamoyl) methoxy)phenyl)oxazole (PDMPO) (1 mM PDMPO in dimethylsulphoxide (DMSO) solution; ThermoFisher Scientific, Waltham, MA, USA) to 70 mL of each sample. After incubating in transparent plastic cell culture bottles in situ for 24 h, the samples were observed and photographed under an inverted Nikon TS100 light microscope (Nikon, Tokyo, Japan) on board. We show a selection of these images taken on board in the Supplementary Material to demonstrate the in situ uptake of silicate by the diatoms. Unfortunately, the preservation of these samples was unsuccessful and therefore further quantitative analysis could not be performed.

For algal taxonomy analysis, 190 mL of melted sample were filled into brown glass bottles and fixed with an aldehyde mixture of hexamethylenetetramine-buffered formaldehyde and glutaraldehyde at 0.1 and 1% final concentration, respectively. Quantitative estimates of each species were performed using an inverted Nikon Ti-U light microscope (Nikon TE300 and Ti-S, Tokyo, Japan) using the Utermöhl (1958) method, as described in Olsen et al. (2017). Furthermore, a variant of the Imaging FlowCytobot (IFCB) (Sosik and Olson, 2007) was used to obtain digital micrographs of algae from ridge-surface samples (slurp gun and scrapes) in the nano- and micro-size fraction (Olsen et al., 2017). These images of algae were assigned to taxonomical groups manually using custom software written by S. R. Laney at Woods Hole Oceanographic Institution and were used for quantitative analysis for the slurp gun and scrape samples from the ridge.

Ice fauna samples collected by divers with a suction pump (Lønne, 1988) below the ridge were preserved in 4% hexamethylenetetramine-buffered formaldehyde solution immediately after sampling. Organisms were identified under a Leica M80 stereo-microscope (Leica Microsystems, Wetzlar, Germany), equipped with an ocular micrometer.

## RESULTS

# Sea-Ice Ridge Properties

The ridge chosen for the study was formed during a storm between 26 and 30 April 2015 from FYI next to a refrozen lead, as observed from the vessel. Based on its physical properties we characterized the ridge as a FYI ridge. MYI ridges, which were not the object of this study, are usually more consolidated than FYI ridges and have lower macro-porosity. The percentage of deformed ice (including ridges and rubble ice) in the area studied between 26 and 31 May 2015, was 50.9 ± 3.2% based on classifications of surface types in three 25 × 25 km Radarsat 2 scenes (Table S1). The percentage of deformed edges next to leads was 2.8–7.4%. At the two sides of the ridge, where we cored for biological analysis (**Figure 1B**), we encountered three ledges on top of each other with voids between them. The three ledges at the star sampling point (**Figure 2**), from top to bottom were 1.29, 0.88, and 1.69 m thick on 28 May, and 0.23, 0.80, and 0.55 m on 31 May. The decrease in thickness was probably a combination of melting and spatial variability. In general, across the ridge, from 24 to 31 May, both snow depth and sea-ice thickness decreased (**Figure 4**). The rubble macro-porosity of the unconsolidated submerged part of the ice, which represents the percentage of voids in between the ice ledges, was 25% on 24 May and decreased to 16% on 31 May. On 28 May, snow thickness was 0.13–0.22 m on the thick ice side (square) of the ridge, while it was 0.07–0.11 m on the refrozen lead side (star) (**Figure 2**).

Incoming PAR averaged from 7, 18 and 20 May was 786 ± 21 µmol photons m−<sup>2</sup> s −1 (average and standard deviation). PAR transmitted through the ridge varied between 0.1 and 8.5% of the incoming PAR. The average transmitted PAR below the ridge was 24 ± 10 µmol photons m−<sup>2</sup> s −1 (n = 334) (n is the number of samples), i.e., about 3% of the average incoming PAR, based on ROV measurements at 0–5 m below the ridge. This was higher than light transmitted through the thicker ice (Average 0.37 ± 0.08 µmol photons m−<sup>2</sup> s −1 , n = 44638) and lower than through the thin refrozen lead (Average 114 ± 69 µmol photons m−<sup>2</sup> s −1 , n = 55) measured during the N-ICE2015 expedition (Taskjelle et al., 2016; Kauko et al., 2017; Olsen et al., 2017). However, from the videos we observed that transmitted light was highly variable and patchy inside the ridge structure. Bright spots were observed inside the ridge in between the ledges (see Video in Supplementary Material).

Since light transmission measurements below ridged areas were scarce, we also attempted to model in a simplistic way the PAR transmitted through the ridge based on the snow and ice thickness and based on optical properties (cf. section Light Measurements and Calculations). The PAR transmitted through the thick-ice side of the ridge was lower (average on 24 May: 9 µmol photons m−<sup>2</sup> s −1 ; average on 31 May: 59 µmol photons m−<sup>2</sup> s −1 ) than through the thin-ice side (average on 24 May: 62 µmol photons m−<sup>2</sup> s −1 ; average 31 May: 274 µmol photons m−<sup>2</sup> s −1 ; **Figure 4**). This coincides with higher snow accumulation on the thick side of the ridge compared to the thin ice side. On 28 May, snow depth ranged between 0.07 and 0.11 m at the thin ice side of the ridge, so we calculated the theoretical minimum and maximum light transmitted through that specific spot from 23 April to 3 June to estimate temporal variability according to measured incoming irradiance (**Figure 5A**). The calculated transmitted PAR at one spot, without taking into account changes in snow and ice light attenuation coefficients as the melt season progressed, was generally one order of magnitude lower than the measured PAR with the ROV, except on 7 May when they

estimated light transmitted below the ridge (yellow).

compared well (**Figure 5A**). The ROV measurements covered a wide area below the ridge and included lateral light sources since measuring depth was up to 5 m below the ridge (Katlein et al., 2016). Therefore, when comparing the measurements with the transmitted PAR calculated across the ridge we do encounter similar values, especially toward the thin ice side where the influence of the refrozen lead allowed more light to penetrate. The spatial variability of calculated light transmitted across the ridge (**Figure 4**) indicates that changes in snow depth and ice thickness were the major drivers of light-transmission variability.

Nutrient concentrations in the water column (at 5 m depth) between 28 April and 25 May were 8.4 ± 0.8µM nitrate, 3.4 ± 0.4µM silicic acid and 0.6 ± 0.1µM phosphate (average and standard deviation) (**Figure 5B**). After the development of a Phaeocystis-dominated under-ice bloom in the water column (26 May−2 June) (Assmy et al., 2017), nitrate concentrations were reduced to 2.4 ± 1.2µM and phosphate to 0.4 ± 0.1µM, while silicic acid increased slightly to 4.1 ± 0.1µM (**Figure 5B**) as we drifted into more Atlantic-influenced waters.

Overall currents were weak, averaging 0.1 m s−<sup>1</sup> relative to the ice, and came from various directions during the study period (23 April−5 June). However, over the period from 30 May to 5 June, current speeds larger than 0.2 m s−<sup>1</sup> were observed with a mean relative current speed of 0.3 m s−<sup>1</sup> flowing in a north-east direction (32◦ ) (Figure S2) that crossed the ridge from the thickice side toward the thin refrozen lead side (**Figure 5C**). Thus, the

part of the ridge facing the refrozen lead was on the lee side of the stronger currents (**Figure 2**).

#### Algal Communities in FYI Ridges

Dense accumulations of algae were observed by naked eye on the top and bottom of the ledges during the entire sampling period (10 May to 3 June; **Figure 7** and Video in Supplementary Material). When sampling these surfaces communities, a clear distinction became apparent between the bottom of the ledges and their vertical surfaces, and the top of submerged ledges. The bottom and the vertical wall communities were dominated by the pennate sea-ice diatoms Nitzschia frigida and Navicula species, while the top community was dominated by Shionodiscus bioculatus (**Figure 6A**). Pennate diatoms of the genus Navicula increased their dominance from 10 to 31 May. The fluffy algal layer that accumulated on the top of submerged ledges and was dominated by S. bioculatus could be easily washed off by divers. A more diverse community was revealed in the ice cores taken from the ridge (**Figure 6B**). The three species that dominated the internal ice community on 31 May were F. cylindrus, N. frigida, and Pseudo-nitzschia sp. Three days later, on 3 June, the percentage of dinoflagellate cysts increased from <10% to >25%. On that day, the most abundant diatoms were Pseudo-nitzschia sp. and N. frigida (**Figure 6B**).

(Bot) or top (Top) of the ice core. Samples analyzed with the imaging FlowCytoBot (IFCB). (B) Entire ledges melted. Numbers at the top correspond to the order of the ledges from top to bottom and the dates of sampling are indicated below. Samples analyzed by light microscopy enabling a higher taxonomic resolution.





Samples taken at the ridge and the biogeochemical and photosynthetic parameters measured are summarized in **Table 1**. Chl a concentrations in the slurp gun samples from the beginning of May ranged between 0.3 and 9.9 mg m−<sup>2</sup> . In late May and early June, the volumetric Chl a concentrations, from melting entire cores from the ledges, ranged between 13.8 and 29.4 mg m−<sup>3</sup> (n = 6) at the thin ice side, which correspond to an integrated Chl a stock of 26–74 mg Chl a m−<sup>2</sup> . The thick ice side had lower Chl a concentrations (4–11 mg m−<sup>3</sup> , n = 2) which correspond to 0.4–1.3 mg Chl a m−<sup>2</sup> based on one bottom and one top 10-cm section (**Table 1**); therefore the biomass in the thick ice is probably underestimated. The integrated POC on the thin ice side of the ridge was 1,134–2,247 mg C m−<sup>2</sup> (94–187 mmol C m−<sup>2</sup> ), the PON 154–314 mg N m−<sup>2</sup> (11–22 mmol N m−<sup>2</sup> ), and the biogenic silica 9–77 mg Si m−<sup>2</sup> (0.3–2.7 mmol Si m−<sup>2</sup> ). The C:Chl a weight ratio of the integrated biomass in the three ledges was 35.8 ± 9.6, the C:N molar ratio of the organic material was 8.4 ± 0.5, and the N:Si molar ratio 4.7 ± 2.1 (n = 6) (**Table 1**). The maximum nutrient demand of the integrated ridge community on 31 May was 15.7 mmol N m−<sup>2</sup> d −1 and 38.9 mmol Si m−<sup>2</sup> d <sup>−</sup><sup>1</sup> based on an estimated growth rate of 0.7 d−<sup>1</sup> (derived from Chl a measurements on 31 May and 3 June) and the measured N:Chl a w:w ratio of 4.25 and the Si:Chl a ratio of 2.11.

The photosynthetic acclimation of the diatoms to the prevailing light climate was assessed with photosynthetic parameters obtained from RLCs. The maximum dark-adapted quantum yield (ϕ) of the slurp gun and scrape samples was 0.40 ± 0.16 (n = 9) for Nitzschia-dominated bottoms of the ledge, and 0.42 ± 0.11 (n = 5) for the Shionodiscus-dominated top part of the ledge (**Table 1**). Variability was very high (range: 0.19–0.61), but most samples were photosynthetically healthy with no evidence of chronic photoinhibition in the dark-adapted yield data. In addition, on-board observations of silica stain uptake samples revealed that the N. frigida bottom community and the S. bioculatus surface community were growing and taking up silicate at the time of sampling (Figures S3A–D). The photoacclimation parameter (Ek), calculated from electron transport with the PhytoPAM, was higher but highly variable for Nitzschia-dominated communities (421 ± 295 µmol photons m−<sup>2</sup> s −1 ) and slightly lower with less variability for Shionodiscus-dominated communities (266 ± 86 µmol photons m−<sup>2</sup> s −1 ). No statistically significant differences were detected in the light-response parameters between these two communities (ANCOVA test for comparison of regression lines; Sokal and Rohlf, 2012).

The sympagic amphipod Apherusa glacialis was the most dominant ice fauna species. Other amphipods present were Themisto libellula, Gammarus wilkitzkii, Onisimus glacialis, and Eusirus holmi. Some zooplankton species, such as the copepods Oithona similis, Calanus glacialis and undetermined Harpacticoida were present, although in lower numbers (Table S2).

#### Snow-Ice Interface Properties

When we arrived on Floe 4 on 11 June, the wider surrounding was mainly composed of FYI with a modal ice thickness of 1.0 m and an average snow depth of 0.25 ± 0.17 m on top.


TABLE 2 | Compilation of biogeochemical parameters of the snow-infiltration communities (SI).

By 18 June, the modal ice thickness decreased to 0.8 m due to a strong bottom melting event, while the snow thickness remained in the same range (Rösel et al., 2018). Penetrating swell caused a breakup of the icepack into scattered 100– 200 m pieces on the morning of 19 June. The snow depth at the first snow-ice interface sampled (0.7 m) was thicker than the mean snow depth of 0.32 ± 0.20 m on Floe 4 (Rösel et al., 2018). When the relation between ice thickness and snow thickness exceeded the hydrostatic equilibrium, the thick snow cover pushed the ice below sea level creating areas of negative freeboard. Based on drill hole measurements, 53% of the area of Floe 4 had negative freeboard (Rösel and King, 2017).

According to the snow pit performed on 13 June at SI1 (Figure S4), the top 0.3 m of the snow pack was hard wind slab of 0.5 mm grain size. The bottom 0.5 m consisted of refrozen melt layers of larger grain size (1.0 mm). The snow hardness decreased toward the bottom of the snow pack, close to the slush where the algae had accumulated (Figure S4). The temperature profile across the snow showed values around 0◦C in the upper 0.3 m and <0 ◦C in the lower 0.5 m (−0.1 to −1.2◦C). Compiling the information from the eight locations sampled (**Figure 1C**), the slush where the algae were found had thickness of 0.04– 0.2 m, temperature of −1.2 to −1.7◦C, and bulk salinity of 6.5 to 21.1 (practical salinity unit, henceforth unitless) (average 13.9 ± 4.3, n = 11; **Table 2**) in the melted slush depending on the amount of seawater that had percolated to the snowice interface. Snow infiltration communities were typically found in areas with thick snow (0.2–0.7 m), thin ice (0.4–0.9 m) in an advanced stage of melt, and were usually associated with cracks in the ice. Seawater percolated through the cracks in the ice toward the flooded snow-ice interface. Algae were found along the cracks and spreading ∼0.5 m to either side of the crack (**Figures 7B,C**).

The PAR transmittance through 0.2–0.4 m snow cover was 3–14% of the incoming irradiance based on measurements on 11 June at SI1 using a scalar PAR sensor. When using the average estimated snow light attenuation coefficient of 14.82 m−<sup>1</sup> for Floe

FIGURE 7 | (A) Underwater photograph of FYI ridge sampled on 31 May indicating the most abundant species of the bottom and top of the ledges. Microscopy images of the bottom of the first ledge shows the pennate diatom Nitzschia frigida (RI19) and the top of the second ledge shows the centric diatom Shionodiscus bioculatus (RI20). (B) Photo of the snow infiltration community SI1 found below 0.7 m of snow on 9 June 2015. (C) Photo of the snow-ice interface SI1 with a metric tape in cm scale to give an idea of its thickness.

3, the calculated transmitted PAR was one order of magnitude lower than in situ measurements. This is due to the fact that the scalar PAR sensor collects light from all directions, while

our calculations assume a downwelling light field. On an average sunny day (1,200 µmol photons m−<sup>2</sup> s −1 incoming PAR), algae below 0.2 m of snow would receive 62 µmol photons m−<sup>2</sup> s −1 and 0.03 µmol photons m−<sup>2</sup> s <sup>−</sup><sup>1</sup> below 0.7 m of snow. This corresponds to 0.0025–5% transmitted PAR.

## Algal Communities at the Snow-Ice Interface

Snow infiltration communities were found at eight different spots on Floe 4 between 9 and 18 June (**Figure 1C**). The taxonomic composition of the snow-infiltration communities was very diverse and included both pelagic and ice-associated species. Besides a small percentage of flagellates, ciliates and dinoflagellates (sum of the three groups 3.4 ± 2.8%), the snow infiltration communities were dominated by the haptophyte P. pouchetii (51 ± 31%) and diatoms (42 ± 27%; **Figure 8**). Phaeocystis pouchetii, which was the dominating species of the under-ice phytoplankton bloom taking place at the same time, was present in the snow infiltration communities (8.3 × 10<sup>5</sup> - 8.7 × 10<sup>7</sup> cells L−<sup>1</sup> ) in similar concentrations as in the water column (8.6 × 10<sup>5</sup> -9.9 × 10<sup>7</sup> cells L−<sup>1</sup> ; Assmy et al., 2017). The dominant pelagic diatoms present were F. oceanica, C. gelidus, Pseudo-nitzschia sp. and Thalassiosira spp.; and the main iceassociated diatoms were F. cylindrus, Navicula sp. and Nitzschia sp. (**Figure 7**). However, some species such as F. oceanica and F. cylindrus can be quite abundant in both sea ice and the water column making the pelagic vs. ice-associated distinction difficult. In terms of diatoms, most snow-ice interface communities were dominated by a typical pelagic algal composition except for SI6 that had a more ice-algal composition (**Figure 7**). This sample had very low counts and a higher percentage of resting spores than all the others, indicating a senescent stage. Also, in many of the snow-ice interface communities, a high percentage of the P. pouchetii colonies observed were decaying and contained very few cells (3–70% of the community was P. pouchetii cells in bad shape), indicating that this species infiltrated from the water column but was not performing optimally in its new habitat.

The Chl a concentration in the slush collected at the snow-ice interface ranged three orders of magnitude: from 0.37 to 362 mg m−<sup>3</sup> (Average: 69 ± 115 mg m−<sup>3</sup> , n = 11; **Table 2**). The POC was 552–15,000 mg C m−<sup>3</sup> and the PON 59–2,102 mg N m−<sup>3</sup> (n = 11) (**Table 2**). The average C:N molar ratio of the algal biomass was 9.7 ± 1.8. Despite being in the middle of the productive season, some of the nutrients present in the melted slush, such as

phosphate, had very high concentration. Nitrate ranged between 0.4 and 2.2µM, phosphate between 0.2 and 4.9µM, and silicic acid between 0.7 and 11.6µM (n = 11) (**Table 2**). The maximum nitrate demand of the snow-ice interface community was 17.8 mmol N m−<sup>2</sup> d −1 at a growth rate of 0.62 d−<sup>1</sup> (calculated with Chl a values from 9 and 10 June in SI1) and a N:Chl a ratio of 5.7.

On-board observations of silica stain uptake samples revealed that the diatoms were growing and taking up silicic acid at the time of sampling (Figures S3E, F), while light microscopy analysis revealed a high percentage of decaying P. pouchetii cells. This variability in the physiological status of the algae at the snow-ice interface was reflected in the photosynthetic activity measurements performed with the PhytoPAM. The 8PSII, at SI1 during 5 days (9, 10, 11, 13, and 14 June), ranged between 0.22 and 0.46 (n = 5) indicating that only part of the snowice infiltration community was healthy. The low salinities at SI1 (6–18) could be responsible for the decaying P. pouchetii cells. The photoadaptation parameter (Ek) ranged between 156 and 453 µmol photons m−<sup>2</sup> s −1 . When comparing the snow-ice interface community with the ridge communities, we found that ridge communities had significantly higher light saturation level (ANCOVA test for homogeneity of regression curves; Sokal and Rohlf, 2012), implying that these communities were acclimated to higher light intensities than the snow infiltration community (**Figure 9**).

#### DISCUSSION AND CONCLUSIONS

#### Contribution of FYI Ridges and Snow-Ice Interfaces to Arctic Algal Biomass and Sampling Challenges

Algal accumulations in complex structures such as ridges and in hidden layers at the snow-ice interface are understudied in the Arctic Ocean and thus have not been accounted for in seaice algal biomass estimates. In this study, we have estimated the contribution of ridge and snow infiltration communities to the total ice algal biomass for the first time (**Table 3**) based on RadarSat-2 satellite scene ice-type classification (Figure S1), in situ negative freeboard measurements, and the measured Chl a in each sea-ice environment.

Ridges and rubble ice could contribute 36–96% of the total sea-ice biomass, assuming that all of this area would sustain the same amount of biomass as the thin ice side of the ridge we sampled (**Table 3**). In reality, the percent contribution was probably lower since not all ridges are FYI ridges close to a refrozen lead. Indeed, in our study region, only 2.8–7.4% were deformed edges next to open water or young ice (**Table 3**). If only this particular type of ridge would host algal biomass as we observed in the thin ice side of our study ridge (26–74 mg Chl a m−<sup>2</sup> ) and the rest of the ridges and deformed areas would only host as much biomass as we observed on thick ice side of the ridge (0.4–1.3 mg Chl a m−<sup>2</sup> ), their contribution would be 34–75% of the total sea-ice biomass (**Table 3**). Nevertheless, compared to other sea-ice environments (FYI, new, and young ice), ridges and deformed ice areas, can account for most of the sea-ice related biomass. This is in agreement with large-scale under-ice ROV surveys that point toward ridges as relevant for algal biomass accumulation (Lange et al., 2017b). It is therefore critical that ridges are examined more closely and included in biomass and productivity estimates for Arctic sea ice.

Snow infiltration communities, which in the Antarctic can be responsible for most of the ice-associated production (Arrigo et al., 1997) and biomass (0.5–30 mg Chl a m−<sup>2</sup> ; Arrigo and Thomas, 2004), seem to have a smaller contribution in the Arctic. Assuming a minimum thickness of the slush layer of 0.04 m, the integrated Chl a was 0.01–14 mg m−<sup>2</sup> . On a larger scale, if all areas with negative freeboard would be inhabited by these communities, algal standing stocks on flooded sea ice could potentially reach 0.1–3.4 mg Chl a m−<sup>2</sup> and contribute 9–32% to the total sea-ice integrated Chl a (**Table 3**). The minimum percent contribution of snow infiltration communities (9%) to total ice algal standing stocks was tenfold higher than the minimum contribution of ice algal biomass in level FYI and SYI (0.9%). Thus, including snow infiltration communities in sea-ice biomass and productivity estimates is relevant, especially during the late productive season when the ice starts melting and bottom sea-ice production decreases (Leu et al., 2015). However, these communities were usually only found along cracks in the ice and not in all flooded areas. Unfortunately, there is currently no way to quantify the percentage of the flooded area covered by cracks in the ice, so the estimate provided is just a potential maximum of the real contribution.

Ridge and snow infiltration algal communities have not been extensively studied in the Arctic, likely due to the difficulties in sampling and detecting them. Ridges are complex ice structures that are challenging to sample using the regular ice-core drilling techniques (Timco and Burden, 1997; Gradinger et al., 2010; Lange et al., 2017a). In our study, the ridge algal community was very loosely attached to the surfaces of the ledge and is therefore, partially lost when sampling the entire core by drilling. Slurp gun samples collected by divers, in otherwise inaccessible cavities and structures of the ridges, seem to be better suited for determining the dominant algal species concentrated at the surface of the submerged ledges compared to ice coring. The latter will likely result in the loss of loosely attached surface communities, such as the fluffy layer of S. bioculatus, during core retrieval. Coring of the entire ridge on the other hand will provide information on the internal ice community structure and biomass. Thus, a combination of ice coring and slurp gun sampling or ice coring from below by divers would be the best approach to assess both qualitatively and quantitatively the algal community in ridges.

ROVs can be used to study the optical properties and derive algal biomass from light transmission. However, ROVs are likely to get entangled in complicated under-ice structures and therefore, measurements are usually taken several meters below the ridge (Lange et al., 2016), integrating light from a broad area under the ridge. Smaller ROVs with better maneuverability might be a solution to map the spatial variability in light penetration and algal biomass inside the ridge structure, but divers are needed to obtain measurements inside specific structures. In addition, specific modeling approaches need to be developed to describe the complex light regime inside the ridge structure, since simple 1D vertical models, like the one we used, fail to reproduce the observed complex light field inside the ridge.

The challenge in the case of snow infiltration communities is to detect them since they are covered by snow and therefore not readily visible from the surface, except at the edge of ice floes (von Quillfeldt et al., 2009). In addition, upscaling of the potential habitat suitable for snow infiltration communities requires a good knowledge of the percentage of the ice floe that has negative freeboard. In this study, we based our estimates on in situ observations from several kilometer long transects with EM31 (Rösel et al., 2016a), the snow probe (Rösel et al., 2016b), and drill holes (Rösel and King, 2017). Satellites cannot detect infiltration layers at the snow-ice interface (Ackley et al., 2008). Detecting potential zones of surface flooding is challenging due to the difficulties in differentiating wet melting snow from surface flooding (Onstott, 1992). It is therefore necessary to be either in person in the field or have autonomous instruments such as Ice Mass Balance buoys deployed on the ice to qualitatively observe rapid sea-ice melt events. However, knowing the potentially flooded area does not give any information on where the snow infiltration communities grow. Our observations indicate that they concentrate along cracks in the ice, and these are difficult to detect when covered with snow, and therefore, difficult to upscale. Despite the potential local importance of snow infiltration communities, their upscaling is challenging and this study is just a first attempt to estimate their contribution to ice-associated algal biomass that needs to be further refined.

# The Role of Ridges and the Snow-Ice Interface as Algal Safe Havens: Irradiance, Nutrients, and Grazing Pressure

The high algal biomass encountered in ridges and at the snowice interface during the 2015 productive season in the high Arctic indicates that these two environments provide shelter


and favorable conditions for algal accumulation and potentially growth.

Newly formed FYI ridges with complex ice structures offer plenty of cavities and surfaces for attachment and deposition (16– 25% voids). The complex structure of sea ice piled up in ridges creates an extensive habitat for sea-ice algae which exceeds other sea-ice related environment in terms of its total surface area, with the exception of Antarctic platelet ice. Indeed, other studies have reported higher macro-porosity values of 30–35% (Høyland, 2007; Strub-Klein and Sudom, 2012). In addition, the lee side of ridges provide algal communities protection from underice water currents, particularly for those algae that are loosely attached to ice surfaces such as the communities dominated by S. bioculatus. Therefore, higher biomass concentrations are expected on the lee side or hydrodynamic shadow of ridges. This effect has been suggested to explain accumulation of diatoms (Melnikov and Bondarchuk, 1987; Krembs et al., 2002), algal aggregates (Katlein et al., 2015b), as well as ice fauna (Hop and Pavlova, 2008; Kiko et al., 2017) in sheltered areas of ridges. Indeed, ridges are hot spots for accumulation of sea-ice fauna (Gradinger et al., 2010) since the organisms can graze on the abundantsurface-attached sea-ice algae. For example, A. glacialis, which was the most abundant ice-associated amphipod in this study, has been shown to actively feed on ice algae, which can be a main contribution to its diet (Werner, 2000; Brown et al., 2017).

The snow-ice interface on the contrary provides shelter from grazing since only small ciliates were observed grazing on the snow infiltration community (**Figure 8**). In addition, the low salinity in the slush (6–21) would reduce the grazing activity of potential grazers that could reach this layer. This lack of strong metazoan grazing pressure in the infiltration community environment could have favored the accumulation of algal biomass. Processes of physical concentration of algal biomass in the slush could also be responsible for the high accumulation of biomass. Since the highest algal biomass accumulations occurred within half a meter around cracks in the ice, we hypothesize that infiltrated communities concentrated in these areas, trapped in the porous snow-ice slush as water percolates to the rest of the floe.

The accumulation of snow on the side of ridges (Chapters 3 and 4 in Thomas, 2017) has likely led to the assumption that light transmission through ridges is very low. However, according to our observations, inside the complex ridge structure there are cavities that appear as bright areas inside the ridge. Bright areas were present especially in ridges associated with leads and thin ice (**Figure 2** and Videos in the Supplementary Material). Furthermore, cracks at the sides of ridges (Katlein et al., 2015a), as well as often snow-free portions of high points in a ridge due to wind erosion (Sturm and Massom, 2010) have been suggested to transmit more light than adjacent level thick ice (Lange et al., 2017a). In addition, the side with less snow and close to the thin ice received more light and could therefore support higher algal growth rates, assuming light limitation at the thick-ice side (**Table 1**).

Snow infiltration communities received more light (PAR transmittance 3–14%) than ridge algal communities (PAR transmittance 0.06–8.5%) according to in situ measurements depending on the snow depth. This transmittance values were similar to those measured below ridges by an ROV in the Central Arctic (up to 5%) (Lange et al., 2017b), and lower than the PAR transmittance in the thin ice next to the ridge (5–40%) (Kauko et al., 2017). However, in situ measurements below the snow might be affected by lateral spreading of radiation and light scatter when removing part of the snow cover to introduce the sensor. Calculated PAR transmittance at the snow-ice interface (0.0025–5%) is generally lower than at the ridges (0.12–71%, range from transects in **Figure 4**) especially at the thin ice side. In some cases, the snow-ice interface received one order of magnitude more light than the water column below thick ice (Olsen et al., 2017). This implies that the snow-ice interface, when flooded, might provide an advantage for infiltrated pelagic diatoms that were growing at low rates in the water column at that time (Assmy et al., 2017).

The other key factor for algal growth is nutrient availability. The algal communities growing on the surfaces of submerged ledges in ridges have direct access to the nutrients in the sea water (**Figure 2**), while the ones in the snow-ice interface are dependent on the nutrients available a priori in the snow-ice layer and those percolating upwards from the water column (**Figure 3**). The observed currents below the ice crossed the ridge from the thickice side to the thin-ice side most of the time, especially toward the end of May, when stronger currents were observed (Figure S2). Before we drifted into the under-ice phytoplankton bloom (Assmy et al., 2017), the currents could have provided a constant flux of nutrients to the ridge surface-attached communities. Diatoms are able to store nutrients intracellularly without using them for growth immediately (Kamp et al., 2011; Fernández-Méndez et al., 2015). Based on our nutrient and current measurements, before 25 May (pre-bloom) one centimeter water layer moving below the ice provided 1.56 × 10<sup>3</sup> -7.78 × 10<sup>3</sup> mmol N m−<sup>2</sup> d −1 and 5.18× 10<sup>3</sup> -2.59 × 10<sup>3</sup> mmol Si m−<sup>2</sup> d −1 , which is two orders of magnitude more than the calculated nutrient demand for these communities (15.7 mmol N m−<sup>2</sup> d −1 and 38.9 mmol Si m−<sup>2</sup> d −1 ) using the method explained in Cota et al. (1987) and our own measured ratios and growth rates. This calculation suggests that the algae fixed to the ridge surfaces are flushed with enough nutrients to support their growth demands. Nevertheless, the currents and nutrient uptake dynamics inside the ridge and at the ice-water interface would need to be resolved better in order to assess the reality of the nutrient supply and limitations. The high C:N ratio (8.4 ± 0.5, n = 6) of the integrated biomass in the entire ledges might be due to the higher fraction of dead cells and/or detritus inside the ice as compared to the surface layers.

Nutrient concentrations in the melted slush at the snow-ice interface were highly variable, yet did reach surprisingly high concentrations, especially phosphate (up to 5µM) and silicic acid (up to 12µM). Nitrate was lower probably due to active consumption by P. pouchetii in the water column (Assmy et al., 2017). The high phosphate concentrations compared to the water column could be due to leakage of nutrients previously stored inside the algal cells (Needoba and Harrison, 2004; Kamp et al., 2011), active remineralization by bacteria (Arrigo and Thomas, 2004; Cowie et al., 2014) or atmospheric deposition with snow precipitation (Nomura et al., 2010). A re-supply of nutrients can eventually come from the infiltrated surrounding seawater. At the time when the snow-ice infiltration communities were observed, nitrate concentrations in the water column were relatively low due to uptake by the under-ice P. pouchetii bloom (**Figure 4B**). However, if the ice had been previously flooded, nutrients from a different water mass could have been trapped in this layer. During the winter months of the N-ICE expedition, snow-ice formation was observed in February-March in ice floes of the same area and similar conditions (Granskog et al., 2017; Merkouriadi et al., 2017). Nutrient concentrations in the snow/slush sampled in March 2015 reached values up to 17µM nitrate, 1µM phosphate and 4µM silicate. This amount of nitrate could yield 42 mg Chl a m−<sup>3</sup> , which is one order of magnitude less than the maximum Chl a concentrations observed in the slush layer (362 mg Chl a m−<sup>3</sup> SI1, **Table 2**). This indicates that the winter pre-formed nutrients are insufficient to explain the high biomass observed at the snow-ice interface.

In addition, the fact that we found several species of pelagic diatoms growing in the snow-ice interface points toward a flooding of the ice and establishment of the infiltration community in a different water mass with a more diatomdominated phytoplankton community than the one we observed. Backtracking of Floe 4 (Olsen et al., 2017, **Figure 1**) indicates that the floe was closer to the shelf break some weeks earlier and these waters might have hosted a different phytoplankton community than the P. pouchetii dominated community observed on the Yermak Plateau. Indeed, the presence of abundant pelagic diatoms in surface waters on 8 June (Assmy et al., 2017) could explain the presence of pelagic diatoms in the snow-ice interface. On the other hand, the haptophyte P. pouchetii, despite being present in high cell abundance, was not performing well, indicated by the high amount of disintegrated cells and colonies observed under the microscope (**Figure 8**). These dead cells might be the reason for the high C:N ratio.

## Distinct Algal Communities Occupy Different Ridge Surfaces

One interesting aspect of understudied ridge environments is that they seem to favor specific algal communities. Inside ridges, two clearly distinct communities were observed at the bottom and at the top of the submerged ledges (**Figure 7A**). A mixture of sea-ice pennate diatoms dominated by N. frigida and Navicula species at the bottom of the ledges is in accordance with previous observations of FYI and MYI, in which these species are dominating the bottom of the sea ice (Syvertsen, 1991; Melnikov et al., 2002). Sea-ice pennate diatoms excrete extracellular polymeric substances that enable them to attach inside brine channels at the under-side of the ice (Krembs et al., 2000, 2011; Bowman, 2013). On the contrary, the centric diatom S. bioculatus seems to have a clear advantage for colonizing the top of ledges (von Quillfeldt et al., 2009) as a fluffy algal layer since this species is not able to actively attach to the ice. This fluffy layer can be easily washed off by strong currents which agrees with previous observations of Arctic ridge communities (Hegseth, 1992; Ambrose et al., 2005). Furthermore, their presence supports the protective role of interior ridge cavities from currents (**Figure 2**).

The difference in the photoacclimation parameter (Ek) between the Nitzschia (421 µmol photons m−<sup>2</sup> s <sup>−</sup><sup>1</sup> on average) and the Shionodiscus-dominated communities (266 µmol photons m−<sup>2</sup> s <sup>−</sup><sup>1</sup> on average) might indicate that different parts of the submerged ledges receive on average different light intensities that favor different species that are able to acclimate to those light conditions. For example, S. bioculatus is more light sensitive and better shade adapted since it usually performs poorly under high light environments such as melt ponds (Assmy et al., 2013). Small-scale light measurements inside ridges and a spatially resolved light transmission model for complex under-ice structures are needed to further confirm this hypothesis.

# Diatoms vs. Phaeocystis at the Snow-Ice Interface

The snow-ice interface had no distinct biotopes within the slush layer. Nevertheless it is interesting to compare the species that accumulated in the infiltration layer with the ones in the water column, which in this study was the source of the snow infiltration community. In the literature there are examples of snow-ice interface layers dominated by Phaeocystis (McMinn and Hegseth, 2004), by diatoms (Buck et al., 1998), or by a mixture of both (Kristiansen et al., 1998). During our study, we encountered five snow infiltration communities dominated by P. pouchetii (SI1, SI2, SI5, SI7, and SI8), and three dominated by diatoms (SI3, SI4, and SI6), although both groups were present in all of them. Differences in Phaeocystis vs. diatom dominance could reflect differences in phytoplankton composition in the source waters when infiltration occurred through the cracks in the ice or the time since flooding occurred at a particular site, the latter on the scale of community succession. Phaeocystis pouchetii was the most abundant species based on cell numbers in the water column at the time of sampling (Assmy et al., 2017), which is consistent with its presence in the infiltration community. This species is supposed to be very plastic since it can adapt its photosynthetic efficiency to the rapidly changing light regime (Palmisano et al., 1986; Cota et al., 1994; McMinn and Hegseth, 2004). However, during a side experiment, in which we removed the snow on top of SI1 and sampled it 24 h later, we could observe a decrease in the healthy cell numbers of Phaeocystis and an increase in poor-quality cell numbers, with no significant change in the diatom composition (Figure S5). This indicates that P. pouchetii could not deal with the rapid increase in irradiance and that diatom frustules are more resistant to decay. The average E<sup>k</sup> of SI1 was 331 ± 125 µmol photons m−<sup>2</sup> s −1 (n = 5) and the measured Ed(PAR) below 0.2–0.7 m of snow ranged between 1 and 162 µmol photons m−<sup>2</sup> s −1 . The difference between E<sup>k</sup> and E<sup>d</sup> indicates that the cells are adapted to a higher light intensity than what they were experiencing in the snow-ice interface at the time we measured light intensity below the snow. In general, the higher light intensities experienced in the snow-ice interface compared to the water column (<1% of incoming irradiance; Assmy et al., 2017), as well as the low salinities at the snowice interface (6.5–21), might have negatively affected part of the P. pouchetii population, while not having as deleterious impact on the diatom portion of the community due to their rigid silica frustules. Similar findings have been observed in ice melt processing studies with respect to flagellate versus diatom species (Buck et al., 1998; Garrison et al., 2003). Pelagic diatoms such as F. oceanica, C. gelidus, and Pseudo-nitzschia sp. which were not abundant in the water column at the time of sampling but might have infiltrated previously, managed to rapidly accumulate in the snow-ice interface (**Figure 8** and Figure S3).

#### Future Predictions and Implications

With the ongoing changes in the Arctic icescape due to anthropogenic climate change, a shift in community composition and productivity of sea-ice algae is expected (Dupont, 2012; Fernández-Méndez et al., 2015; Hardge et al., 2016; Olsen et al., 2017). During May-June 2015, the percentage of ridge and deformed ice cover was very high (46–51%), in agreement with recent airborne surveys indicating that ridged ice can make up a substantial fraction of the pack ice (Haas et al., 2010). For example, in Fram Strait from 1990 to 2011 ridges contributed 66% of the mean thickness of sea ice (Hansen et al., 2014). In the coming decades, as the ice gets thinner and more dynamic due to increased temperatures and wind (Spreen et al., 2011; Renner et al., 2014), an increase in FYI pressure ridge formation is expected (Wadhams and Toberg, 2012). As we have shown in this study, FYI ridges close to refrozen leads can host high biomass of healthy algal communities and could therefore play an important role in the Arctic icescape's future productivity. We encourage future studies to focus on pressure ridges despite the sampling challenges, since they are an important and under-quantified part of the Arctic icescape.

The mean snow thickness observed on FYI on Floe 4 was 0.32 ± 0.20 m (Rösel et al., 2018), which is in the same range given in the Warren-Climatology based on observations from snow on thick MYI (snow of 0.33 m; Warren et al., 1999). Seaice and snow thicknesses have changed toward a thinner, FYIdominated ice cover that had less time to collect snow than older ice (Gallet et al., 2017; Merkouriadi et al., 2017). The combination of thin and rapidly melting sea ice and a relatively thick snow cover, led to negative freeboard and flooding of approximately half of Floe 4. This situation might become more frequent in the future, as sea-ice thickness continues to decrease (Maslanik et al., 2007; Stroeve et al., 2012), while precipitation falling on sea ice has been predicted to increase north of Greenland and in the Eurasian basin of the high Arctic where the remaining ice will reside (Bintanja and Selten, 2014). In addition, in the Atlantic sector, the influence of an increasingly warm Atlantic water inflow will contribute to faster ice melt from below (Polyakov et al., 2017). Thus, the contribution of snow to seaice mass balance could increase (Granskog et al., 2017), with flooding events in early spring (Granskog et al., 2017; Provost et al., 2017). These conditions favor the accumulation of algae at the snow-ice interface. These snow-infiltration communities have been frequently observed in the Antarctic, where the ratio of snow-to-sea ice thickness is high. We hypothesize that this "Antarctification" of the Arctic icescape will lead to more frequent accumulation of sea-ice algae at the snow-ice interface, especially in the Atlantic sector, and that snow infiltration communities might play a similarly important role in sea-ice related productivity in the future Arctic, as in the Antarctic (Arrigo et al., 1997).

The consequences of more algae accumulating in these two environments are still unknown, but we can hypothesize that ridges will become hot spots of biomass that will fuel the iceassociated food chain, since they will be accessible for grazers (Gradinger et al., 2010) and their carbon will be transferred to upper trophic levels (Falk-Petersen et al., 2009). On the contrary, snow infiltration communities will remain largely inaccessible for larger grazers during the productive season, although some grazers have been observed at the ice surface in Antarctic sea ice (Schnack-Schiel et al., 2001), and will likely sink when the ice melts, strengthening the sympagic-benthic coupling (Søreide et al., 2013) if they are not being decomposed and remineralized by bacteria. Moreover, the different algal species accumulating in these environments will influence how much carbon is exported to the seafloor, given that diatoms are more efficient carbon exporters than P. pouchetii (Reigstad and Wassmann, 2007). In terms of timing, while snow infiltration communities seem to appear only at the end of the productive season linked to ice melt, ridge communities are likely important year-round but particularly during the summer melt season when most ice algal biomass is lost from level sea ice. Thus, pressure ridges might act as refuges for the ice-associated flora and fauna during times of rapid melt and as an algal seed bank for newly formed ice.

The key points of this study are:


We conclude that both, ridges and the snow-ice interface are important and understudied environments in the Arctic ecosystem. This study provides a comprehensive description of these two environments and, thus, can be used as a baseline for more extensive studies in the future. An assessment of the role of FYI ridges and snow infiltration communities for Arctic sea-ice biomass and productivity will become more important in the future with the ongoing trends of sea-ice thinning and increase in precipitation.

#### AUTHOR CONTRIBUTIONS

All coauthors contributed to data collection, analysis, and interpretation. The conception and design of the study was led by PA, PD, HH, CM, and MF-M. Writing of the manuscript was done by MF-M with input from all coauthors.

#### FUNDING

This study was supported by the Centre of Ice, Climate and Ecosystems at the Norwegian Polar Institute through the N-ICE2015 project. PA, PD, LO, HK, CM, BS, and HH were supported by the Research Council of Norway (project Boom or Bust no. 244646). MF-M, PA, CM, AR, and HH were supported by the Ministry of Foreign Affairs, Norway, through the ID Arctic project. AJ was funded by the Norwegian Research Council (NFR) through the NORRUSS program (NFR project no. 233896) and we were thankful for using resources developed

# REFERENCES


under CIRFA (NFR number 237906). The ROV purchase was funded by the Canada Foundation for Innovation (CFI) grant to CM and JE.

#### ACKNOWLEDGMENTS

We would like to thank the captain and crew of the RV Lance and the polar bear guards for making the sampling possible. We acknowledge the work of Max König and George Roth with **Figure 1A** map and Vasilii Kustov for the AUV aerial pictures of Floes 3 and 4 in **Figures 1B,C**. Radarsat-2 data was provided by NSC/KSAT under the Norwegian-Canadian Radarsat agreement 2015. We recognize the efforts of Max König (NPI) and Thomas Kræmer (UiT) for making the co-located satellite image acquisitions possible.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars. 2018.00075/full#supplementary-material

The video for this article can be found online at: https:// figshare.com/s/d31c7742e889e31c6b32

a rapidly changing Arctic Marginal Ice Zone. Progr. Oceanogr. 139, 122–150. doi: 10.1016/j.pocean.2015.09.003


**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 © 2018 Fernández-Méndez, Olsen, Kauko, Meyer, Rösel, Merkouriadi, Mundy, Ehn, Johansson, Wagner, Ervik, Sorrell, Duarte, Wold, Hop and Assmy. 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.

# Phytoplankton Realized Niches Track Changing Oceanic Conditions at a Long-Term Coastal Station off Sydney Australia

Penelope A. Ajani <sup>1</sup> \*, Niall McGinty <sup>2</sup> , Zoe V. Finkel 3† and Andrew J. Irwin2†

*<sup>1</sup> Climate Change Cluster (C3), University of Technology, Sydney, NSW, Australia, <sup>2</sup> Department of Mathematics and Computer Science, Mount Allison University, Sackville, NB, Canada, <sup>3</sup> Environmental Science Program, Mount Allison University, Sackville, NB, Canada*

#### Edited by:

*Kristian Spilling, Finnish Environment Institute (SYKE), Finland*

#### Reviewed by:

*Andrew M. Fischer, University of Tasmania, Australia Alena S. Gsell, Netherlands Institute of Ecology (NIOO-KNAW), Netherlands*

#### \*Correspondence:

*Penelope A. Ajani Penelope.Ajani@uts.edu.au*

#### †Present Address:

*Zoe V. Finkel, Department of Oceanography, Dalhousie University, Halifax, NS, Canada Andrew J. Irwin, Department of Mathematics and Statistics, Dalhousie University, Halifax, NS, Canada*

#### Specialty section:

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

Received: *30 April 2018* Accepted: *25 July 2018* Published: *17 August 2018*

#### Citation:

*Ajani PA, McGinty N, Finkel ZV and Irwin AJ (2018) Phytoplankton Realized Niches Track Changing Oceanic Conditions at a Long-Term Coastal Station off Sydney Australia. Front. Mar. Sci. 5:285. doi: 10.3389/fmars.2018.00285* Phytoplankton dynamics are closely linked to the ocean-climate system with evidence that changing ocean conditions are substantially altering phytoplankton biogeography, abundance and phenology. Using phytoplankton community composition and environmental data spanning 1965 to 2013 from a long-term Pacific Ocean coastal station offshore from Sydney, Australia (Port Hacking 100 m), we used the Maximum Entropy Modelling framework (MaxEnt) to test whether phytoplankton realized niches are fixed or shift in response to changing environmental conditions. The mean niches of phytoplankton closely tracked changes in mean temperature, while the mean salinity and mixed layer depth realized niches were consistently at the extreme range of available conditions. Prior studies had shown a fixed niche for nitrate in some phytoplankton species at a site where nitrate concentration was decreasing and potentially limiting; however, at Port Hacking nitrate and silicate niches increased more rapidly than environmental conditions, apparently in response to periodic occurrences of elevated nutrient concentrations. This study provides further evidence that climate change model projections cannot assume fixed realized niches of biotic communities, whilst highlighting the importance of sustained ocean measurements from the southern hemisphere to enhance our understanding of global ocean trends.

#### Keywords: Port Hacking, climate change, MaxEnt, nitrate, species distribution models

#### INTRODUCTION

Warming of the Earth's ocean and atmosphere due to anthropogenic CO<sup>2</sup> emissions has seen a global average increase in surface air temperature of 0.85◦C over the past century (IPCC, 2014), with the upper ocean trapping the majority of the anthropogenic heating (IPCC, 2013). Atmospheric warming is associated with a decrease in pH and acidification of ocean waters (Rost et al., 2008; Beaufort et al., 2011). It is in these upper, sunlit waters of the global ocean that phytoplankton flourish, producing ∼ 45 Gt a−<sup>1</sup> of organic carbon (Falkowski et al., 1998; Field et al., 1998). Phytoplankton are a critical food source for higher trophic levels, sustaining marine food webs, which culminate in important fish stocks (Falkowski et al., 2004; Doney, 2006; Richardson and Poloczanska, 2008).

The flow-on effects that changes in climate and phytoplankton communities may have on ocean food webs and global biogeochemical cycles are poorly understood yet potentially profound and include the potential for harmful algal bloom intensification (Edwards and Richardson, 2004; Gobler et al., 2017). Moreover, establishing links between climate change and trends in the structure of phytoplankton assemblages is challenging, as phytoplankton have been shown to exhibit ordersof-magnitude variability over seasonal, inter-annual and interdecadal time scales (Zingone et al., 2010).

With the potential for rapid dispersal, high reproduction rates and unexpected genetic structuring (Koester et al., 2013), both the fundamental niche of phytoplankton species and the ecological space or realized niche occupied in a particular community (Hutchinson, 1957) may be adaptable to changing ocean conditions. The set of all abiotic and biotic conditions in which a species can persist is called the fundamental niche and is usually determined experimentally in controlled conditions. Observational studies of natural communities can be used to document the realized niche where a species actually occurs, which can be thought of as a modification of the fundamental niche resulting from competition and biotic interactions. Various niche characteristics (e.g., niche position, breadth, overlap, plasticity, and conservatism) have been estimated under varying single- or multi-stressor conditions both in the laboratory (Collins et al., 2014; Boyd et al., 2015; Thomas et al., 2016; Ji et al., 2017) and the field (Irwin et al., 2012; Brun et al., 2015). However, understanding the organismal tolerance and plasticity that will drive evolutionary change remains uncertain, with model projections of biotic communities under climate change still assuming fixed realized niches (Chivers et al., 2017).

Recent approaches have used many species with different trait values to seed marine ecosystem models to examine emergent biogeography of microbial communities (Follows et al., 2007). Field based observations have been used to compare the changes in biogeography of different functional groups of phytoplankton and zooplankton to the velocity of climate change (isotherm movement) (Chivers et al., 2017). Combined with species distribution modeling (SDM), field observations have been used to model phytoplankton biogeography between historical and projected future ocean conditions (Barton et al., 2016) and to track species over time at a single location (Irwin et al., 2015). In the latter example, modeled phytoplankton data from a longterm coastal station CARIACO (Carbon Retention in a Colored Ocean) off the coast of Venezuela, revealed that species niches were not stable over decadal time periods, but were able to exhibit some adaptive capacity to changes in environmental conditions. Many questions remain, however, as we assess the importance and generality of this phenomenon. Little is known about how quickly species respond to environmental changes, if some species or functional groups are particularly flexible or resistant to changes, and what is the potential for contrasting effects of individual environmental drivers or directional changes. More time series are needed to resolve these open questions.

The Port Hacking 100 m (at a depth of 100 m) coastal monitoring station (hereafter PH100m), located on the east coast of Australia, is one of the longest established coastal stations in the Southern Hemisphere. Located 5 km from Sydney, this station has been the focus of many short-term phytoplankton and hydrological investigations since its inception in 1954, with its disparate phytoplankton datasets only recently assembled (Ajani et al., 2016). In 2009, this station became one of nine National Reference Stations located around Australia maintained by the Integrated Marine Observing System (IMOS) to monitor long-term changes in Australia's ocean (Lynch et al., 2014). PH100m is located within a very complex oceanographic setting, dominated by the East Australian Current (EAC), which originates in the Coral Sea to the north and brings warm, oligotrophic waters into more temperate latitudes (Ridgway and Dunn, 2003). The EAC has strengthened and moved poleward (Ridgway and Hill, 2012; Wu et al., 2012) with long-term increases in temperature, salinity, and nitrate and a decline in silicate recorded at this station over the past 60 years (Thompson et al., 2009).

While no concomitant shift in total phytoplankton abundance has been reported at this station over the past 60 years, species composition over the past decade has seen a decline in dinoflagellates compared to diatoms toward the present. There is also an emerging dominance of two tropical species Trichodesmium erythraeum (cyanobacterium) and Bacteriastrum spp. (diatom) at this station (Ajani et al., 2014a,b). This shift in composition, however, has only been reported over the most recent decade, when water temperatures declined amidst a longterm warming signal.

With this in mind, we studied the realized niches of phytoplankton species at PH100m to address the following questions. Do realized niches change in response to environmental change, or are they primarily conserved resulting in a restructuring of communities and changes in biogeography as environmental conditions change? The changes in physical and chemical conditions at PH100m allowed us to test for changes in realized niches over time, so we refined this question by examining how rapidly realized niches change. Another refinement recognizes that the realized niche for some environmental variables may be much more plastic than for others, so we investigate if niches change at different rates for different variables. Many environmental conditions and changes in conditions are correlated, so we investigate if these correlations are reflected in the realized niches and their changes. Finally, we investigate if species respond to changes in the mean environmental conditions, or if more detailed information about the distribution of environmental conditions is required to anticipate changes to realized niches.

#### MATERIALS AND METHODS

Five disparate phytoplankton sampling campaigns were conducted at the PH100m coastal station (34◦ 7'3.36"E, 151◦ 13'5.52"S, **Figure 1**) over the period from 1965 to 2013 (approximately ∼20 years of sampling over a ∼50 year period). These were coded OSL (1965–1966), HALLE (1978–1979), AJANI97 (1997–1998) AJANI98 (1998–2009), and NRS (2009– 2013) and their sampling frequencies, methodologies and circle.

references are summarized in **Table 1**. In brief, phytoplankton samples were collected approximately weekly to monthly using either discrete bottle samples or phytoplankton mesh nets (20 or 37µm) and preserved before microscopic examination. Cells were identified to the lowest possible taxon using light microscopy and where identification to species level was not possible, cells were assigned to genus level only (e.g., Chaetoceros spp., Thalassiosira spp. etc.). For comparability across sampling campaigns, certain taxa were pooled to genus level. Simultaneous meaurements of several environmental parameters including temperature (◦C), salinity, oxidized nitrogen (µmol L−<sup>1</sup> ; nitrate and nitrite, hereafter "nitrate"), and dissolved reactive silicate (µmol L−<sup>1</sup> ; hereafter "silicate") occurred during these sampling campaigns (refer to references in **Table 1** for all environmental collection and analytical methodologies).

All phytoplankton abundance data were pooled across 0–50 m except for AJANI97 and AJANI98 which were 0–100 m and converted to simple records of species presence, while all environmental parameters were averaged across the upper 50 m. To assess if the phytoplankton were tracking changes in thermal stratification at this location, we also calculated the mixed-layer depth (MLD) over the sampling duration. This was defined as the depth where the temperature was more than 0.5◦C lower than the surface temperature (Levitus, 1982).

Differences in sampling methodology and frequency can introduce a sampling bias and effect model performance when performing niche models (Kramer-Schadt et al., 2013). Although this concern is largely focused on spatial distribution modeling, similar problems could arise from temporal models collected at a single station. Temporal filtering of weekly data to monthly reduced the number of viable species to model, but for those retained we found significant positive correlations (r = 0.78 to 0.93) between the filtered and unfiltered species in terms of how each species track their niche. Similarly, a comparison between the two principal modes of sample collection (net hauls and discrete bottle samples) during 2009, found that while cell counts were significantly lower in the bottle samples, over 72% of the species were collected by both methods and the five most abundant species were the same. We therefore did not correct for changes in sampling frequency or method. For sampling frequency in particular, the spatial filtering would have resulted in a significant loss of modeled species and our tests suggest there would be no change to the overall conclusions.

#### Tracking Changes in Phytoplankton Niches

To investigate whether phytoplankton niches have adapted to changing environmental conditions on a decadal scale at PH100m, we divided the combined dataset into three distinct periods representing different thermal regimes.The first period, P1, represented four years of stable colder temperatures before the rapid warming during the mid 1990s, and included data from the years 1964, 1965, 1978, and 1979. The second period, P2, which included data from 1997 to 2004, showed no abrubt step-wise changes in environmental conditions but did exhibit a significant decline in ocean temperatures. The final period, P3, showed evidence of renewed warming and included data collected from 2005 to 2013. We chose the end of year midpoint as the division for the two periods separating the data into an initial post El Niño cooling phase (P2) and a cooler period with renewed warming (P3).

Secondly, we investigated how quickly phytoplankton track changes environmental conditions by observing the rate at which the niche changed on a rolling annual basis. We limited this analysis to the years between 1997 and 2013 due to the large gaps present between the earlier four years of data. We calculated a running mean niche of each species using a 4-year moving window which provided a possible total of 14 niche measurements (i.e., 1997–2001, 1998–2002, . . . , 2010–2013). For both analyses, species that were found to occur at least 15 times in at least one period (P1, P2, and P3), or within one 4-year window, were retained for subsequent niche modeling.

#### Statistical Analysis

We used the Maximum Entropy Modelling framework (MaxEnt, Phillips and Dudik, 2008) to estimate the logistic probability of finding a particular species in univariate or multivariate niche space by comparing the environmental conditions where a taxon is present with the measured background environment of all sampling events. MaxEnt has become one of the most popular algorithms used to model species distributions. It has consistently been found in comparative studies to give robust estimates and to perform as one of the best algorithms available (Elith and Leathwick, 2009). Our focus is on using MaxEnt to describe the realized niches of phytoplankton and to test for changes over time in these niches. Alternative modeling frameworks could be used


TABLE 1 | Phytoplankton sampling campaign codes, frequency, duration, methodology and reference from PH100m coastal station (IMOS = Integrated Marine Observing System; NRS = National Reference Station).

*P1, P2, and P3 refer to the 3 periods of analyses used in this manuscript; P3 starts in 2005 (see methods for additional detail).*

ranging from regression to machine learning. MaxEnt is both highly flexible, with weak a priori assumptions about the response of each species to each condition, and highly interpretable in contrast with some "black-box" machine learning techniques. For each MaxEnt model run, a total of 100 bootstrap re-sampling runs were performed. Threshold and hinge responses were disabled as our focus is primarily on the mean niche conditions for each species and to reduce the likelihood of overfitting of the model (Elith et al., 2011). Model performance was evaluated by examining the area under the curve (AUC) of the receiving operator characteristic, where values approaching one suggest a higher probability that a model will correctly identify a species presence. Using the AUC as a single measure for evaluating model performance can be prone to several biases (Yackulic et al., 2013). Nevertheless, by raising the minimum AUC for inclusion of a model to 0.7, we successfully removed the most poorly fitted models that may have biased our results. We used the permutation importance to measure the contribution of each variable to the total AUC on a percentage scale by permuting the values of one variable at a time at random and observing the decrease in model performance. A large decrease in the AUC indicates that the variable is very important in the characterisation of the species' niche.

To investigate whether phytoplankton niches have adapted to changing conditions on a decadal scale, we focused on each environmental variable individually and used the probability response curves to calculate the mean realized niche of each modeled species for each period (P1, P2, P3) or 4-year moving window during the period 1997–2013. For the decadal analysis, the background data were sampled from all observations. For the moving window approach, the background data were sampled from the years included in the moving window. For both analyses, only species that were found to occur at least 15 times in at least one period (P1, P2, and P3), or within one 4-year window were examined. The 100 bootstrap samples provided an estimate of 95% confidence intervals of the mean niche for each species. We restricted our analysis to environmental conditions that were common to all periods and removed extreme values present within individual periods of the time series. Had they been included these extreme values could have had an effect on the species niche calculation, where a difference in niches would be found largely due to the change in available background conditions.

We performed several tests to examine the effect of the changing environment on the realized niches. For each environmental variables, we calculated the community average mean niche for all species within a period as well as the background environmental conditions. We hypothesize that the magnitude and direction of the changes in the niche and environmental differences will be similar. We also hypothesized that the degree of niche change is dependent upon the initial niche and its distance from the mean environmental conditions. To test this, we first examined species that were found in two or more periods and used a linear regression model to examine the pair-wise relationship between the mean niche of the earliest period and the change in mean niches between the two periods. We then examined the rate at which changes to a species niche occurred. We computed the correlations between the running mean of the background environmental conditions and the running mean of the mean niche for all species that were modeled in at least 5 of the 14 time steps. To test for a delay in the response of the community to environmental change, we computed these correlations on simultaneous measures and with a lag of 1–4 years between the environmental conditions and the estimated mean niches. Since some of the changes in niches did not seem to be directly proportional to changes in the mean environmental conditions, we compared the distance between the mean niche and mean environmental conditions to the standard deviation of the environmental conditions in each 4-year moving window.

#### RESULTS

#### Phytoplankton Niche Modelling

A total of 34 species were present in at least two periods with a mean AUC of 0.7 or greater (**Supplementary Table 1**). The variation in mean AUC was low with a standard deviation of < 0.1 across all three periods (P1: 0.71 ± 0.089; P2: 0.81 ± 0.063; P3: 0.72 ± 0.061). Four environmental variables (temperature, salinity, mixed layer depth, and nitrate concentration) allowed comparisons across all three periods while comparisons for silicate concentration were only possible between the latter time periods P2–P3 (**Table 2**). A total of 16 species were present in both periods P1 and P2 and 31 species were present in periods P2 and P3 (**Supplementary Tables 2**, **3**). We found changes in the community average mean niches and the mean environmental conditions were always in the same direction for temperature, MLD and nitrate concentration between P1 and P2 and between P2 and P3. Changes in the community average mean niches exceeded changes in the environmental conditions for temperature, nitrate and silicate concentration (**Table 2**). Salinity and mixed layer depth changes were very small or not significantly different from the environment and mean niche between both pairs of periods. A few niche changes nominally in the opposing direction to changes in the environment all corresponded to small, non-significant changes in environmental conditions. Changes in the mean niches for each species were in the same direction as the environmental change for almost all species (**Supplementary Tables 2**, **3**). For each species, the magnitude of niche change was negatively associated with their initial niches, indicating that species with niches in the initial period (P1 or P2) farthest from new conditions in the second period (P2 or P3) exhibited the largest change in their mean niche (**Figure 2**). We interpreted this as a signal of direct pressure on the realized niche due to changes in environmental conditions.

The most important variables determining a species' niche varied between the three periods examined (**Table 3**). In P1, temperature was the most important variable for all species overall (with a mean permutation importance of 36%), with nitrate concentration (21%) being marginally more important than the two remaining variables (silicate was not available for this period). For P2 temperature was the third most important variable (18%) having been supplanted by the added variable silicate concentration (32%) as most important and nitrate concentration (20%) as the second most important. Period P3

TABLE 2 | Mean environmental conditions (top) and niches (bottom), 95% confidence interval on the mean, and change in means between two periods for temperature, mixed layer depth (MLD), salinity, nitrate and silicate concentrations over time (P1: 1965–1979, P2: 1997–2004, and P3: 2005–2009) at the Port Hacking monitoring station.


*Differences between periods shown in bold are statistically significant (P* < *0.05).*

showed an increase in the importance of salinity (22%) and a slight decrease in importance for both nutrient variables (**Table 3**). Overall, all the predictors played an important role in determining the realized niches of species on average, with no clear signal in which one variable dominated the information characterizing the niches.

#### Tracking Change in Phytoplankton Niches

A total of 33 species appeared in suitable numbers of samples to model the niche at least once during the period 1997– 2013 (**Supplementary Table 4**). There were strong positive correlations found between changes in environmental conditions and each species niche and the average niches of all species (**Figure 3**). We found no evidence of any lagged responses, meaning that the changes in mean niche likely occurred within one year of the corresponding change in the environment. The time-series analysis revealed three different patterns in the differences between the mean environment and mean realized niches and how they changed over time. For temperature, the mean niche was approximately equal to the mean temperature in the environment, changing at a similar rate and in the same direction. For salinity and mixed layer depth, the mean niches were at extreme values of the distributions of environmental conditions, corresponding to low (relatively fresh) salinity and large (relatively deep) mixed layer depths. Changes in environmental conditions and mean niches were relatively small for both of these variables (summarized in **Table 2**, **Supplementary Table 3**). A third pattern was observed for nitrate and silicate concentrations. Mean niches increased more rapidly than corresponding changes in the mean environmental conditions from 2000 to 2007. The changes arrested in the second half of the study from 2007 to 2013. Despite the increases in silicate niches, there was a steady decrease in the mean silicate concentrations between 1997 and 2013. We computed the difference between the community average mean niche and the mean environmental conditions for each variable. This separation between niches and environmental conditions was close to 0 for temperature, approximately constant for salinity and mixed layer depth, and increased rapidly for nitrate and silicate concentration. Since the mean environmental conditions was not predictive of this difference, we compared the difference to the standard deviation of the environmental conditions (**Figure 4**). Nitrate and silicate concentrations both showed strongly positive correlations demonstrating that changes in the distribution of environmental conditions (represented here by changes in the standard deviation, but likely due the increasing frequency of higher nutrient concentrations) led to the changes


TABLE 3 | The relative importance (Mean, %) and 95% confidence interval (CI) of each of the five environmental variables within each of the three periods (P1: 1964, 1965, 1978, 1979; P2: 1997–2004; P3:2005–2013) averaged over all species analyzed.

in the mean niches. No such relationship was found for the other environmental variables (dotted lines in **Figure 4**) Each variable is analyzed in a separate panel of **Figure 4** and the final panel **F** overlays all the variables, standardized to zero mean and unit variance in each axis to facilitate comparison across variables. The community was dominated by diatoms so we attempted to determine if the silicate effect was stronger in diatoms compared to other functional types, but the small number of species of other groups made a careful taxonomic analysis inconclusive.

## Correlations in the Environment and Niche Interpretations

Correlations among environmental variables are ubiquitous in the ocean and always have the potential to complicate the analysis of observational data. Our focus has been on univariate species distribution models and the mean realized niche rather than on a multivariate model which might be selected to get the most explanatory power to be used for prediction. Univariate niches are generally easier to interpret (Elith et al., 2011; Irwin et al., 2012). The correlations among our predictors are generally small in this time series, and only significantly different from 0 for three pairs of variables: temperature and nitrate concentration, silicate and nitrate concentration, and a very small correlation between silicate concentration and salinity (**Table 4**). The sign of these correlations agrees with the relative changes in mean realized niches, but the relative changes in temperature and nitrate niches relative to changes in the mean environmental conditions are very different, indicating that there is an independent signal observed in changes in the temperature and nitrate realized niches.

# DISCUSSION

The Port Hacking coastal monitoring station (PH100m) is one of the longest running ocean time series in the Southern Hemisphere (**Figure 1**). The hydrography of the region exhibits complexity at several time scales (Hallegraeff and Jeffrey, 1993; Ajani et al., 2001, 2016; Pritchard et al., 2003). Over the last 60 years there has been a long-term increase in temperature (0.75◦C century−<sup>1</sup> ), salinity (0.23 century−<sup>1</sup> ), and nitrate concentration (0.56 µmol L−<sup>1</sup> century−<sup>1</sup> ) and decline in silicate concentration (−1.97 µmol L−<sup>1</sup> century−<sup>1</sup> ) (Thompson et al., 2009). Superimposed on these multi-decadal trends are seasonal, annual, and decadal variations. In recent years (1997– 2013) there has been a decline in salinity, likely due to an increase in rainfall, a modest decline in sea surface temperatures, a decline in mixed layer depth, an increase in annual average nitrate concentrations and a decline in silicate concentrations. Climate predictions for Australia include warmer ocean temperatures and more intense rainfall events across the nation, although annualaverage rainfall is projected to decline (www.csiro.au/state-ofthe-climate). Australian climate patterns are also influenced by the long-term increasing trend in global air and ocean temperatures (http://www.bom.gov.au). It has been hypothesized that the decline in silicate and other environmental conditions will lead to changes phytoplankton community composition at this site (Ajani et al., 2014a,b). Phytoplankton are rapid and effective indicators of changes in the oceanic environment (Richardson and Schoeman, 2004). Here we determine the capacity of the realized niche of phytoplankton at this coastal site to track changes in their environment conditions since 1965, with an emphasis on the years from 1997 through 2013.

Microbes may have a high capacity to adapt to climate change through selection on standing diversity and de novo mutation, although there may be limits on the ability of species to adapt to multiple stressors (Collins, 2013). In the laboratory, evolutionary change has been demonstrated in numerous microbial species including Escherichia coli, Trichodesmium, Chlamydomonas, and Emiliania, often in less than a thousand generations, indicating that marine phytoplankton may be able to adapt to climate change nearly as rapidly as it occurs (Frank and Slatkin, 1990; Collins and Bell, 2004; Benner et al., 2013; Collins, 2013; Hutchins et al., 2015; Walworth et al., 2016). Despite this evidence, many researchers assume that the physiological traits and niches of phytoplankton are fixed as this facilitates projections of the effects of climate change and there remain few studies documenting changes in phytoplankton niches over time, especially in the field (Thomas et al., 2004, 2012; Flombaum et al., 2013; Barton et al., 2016).

The fundamental niche is the full set of environmental conditions under which a species can persist, and for phytoplankton, this is typically defined in the laboratory. In the field, species are influenced by the interaction of multiple environmental conditions and simultaneous, competition and other biotic interactions, all of which are rarely examined in the lab and as a result, their realized niche is generally a limited subset of their fundamental niche. At a tropical site

FIGURE 3 | Mean realized niches change over time in response to changes in mean environmental conditions and their distributions. Running mean niches are compared to environmental conditions in panels (A–E). Monthly environmental conditions are shown (black line, filled circles) for (A) temperature (◦C), (B) mixed layer depth (m), (C) salinity, (D) nitrate *(Continued)* FIGURE 3 | concentration (µmol L−<sup>1</sup> ), and (E) silicate concentration (µmol L −1 ). The 4-year running mean of the mean realized niche for species found in at least 4 of the 15 periods between 1997 and 2013 (dashed lines) and a loess smoother (black, thick line) with a shaded region 1 standard error wide (red) are overlaid on these panels.

in coastal Venezuelan waters (Station CARIACO) over a 15 year period (1995–2011), we observed changes in the realized niches of 67 phytoplankton species in response to warming of about 1.3◦C and a decrease in nitrate concentration of about 0.7 µmol L−<sup>1</sup> (Irwin et al., 2015). While changes in the realized temperature niche closely tracked changes in average conditions in the environment, we observed two distinct responses in the nitrate niches. Some species tracked the decrease in nitrate concentration while others retained an essentially fixed realized niche. We do not know if these results are typical of changes in phytoplankton at other sites, how rapidly the changes occurred, or the reason for the differences we observed in temperature and nitrate niche changes.

The evidence from the Port Hacking coastal station is that realized niches of phytoplankton track changes in environmental conditions (**Figure 2**, **Table 2**). We observed three patterns of changes in the realized environmental niches that we believe to be a result of mismatch between the fundamental niches and available environmental conditions. For temperature, the mean niche, averaged over all species, is very close to the mean environmental conditions and the mean niche closely tracks changes in the environment. The changes in mean niche have the same sign but are slightly larger than changes in ocean temperature, whether we examine changes from one period to the next (**Table 2**) or changes in the running-mean estimate over time (**Figure 3A**). This is expected, since the fundamental temperature niche for many species is generally wide compared to the ranges of temperatures observed at PH100m (Boyd et al., 2013; Brun et al., 2015). In addition, ocean currents can be expected to have exposed the drifting phytoplankton communities arriving at PH100m to large temperature fluctuations, which will have helped retain species with wide temperature niches (Doblin and van Sebille, 2016), although in some cases currents may promote local adaptations and barriers to gene flow (Rynearson and Armbrust, 2005). By contrast, mean niches are biased relative to the environment for salinity and mixed layer depth, with salinity niches being smaller (fresher) by about 0.25 psu compared to the environment and mixed layer depth niches being larger (deeper) by about 30 m. Mean niches for both these variables are generally at an extreme edge of the distribution of the environmental conditions, indicating an increased probability of finding phytoplankton species in general at one end of the distribution of these variables. Changes in environmental conditions between periods (**Table 2**) are not significant, and changes in mean niche are near zero but significant (for salinity) or not significant (for mixed layer depth). As with temperature, inter-annual variability in these environmental conditions is much larger than the changes over the time-series in mean niche (**Figures 3B,C**). More dramatic changes in mean salinity and mixed layer depth could be expected

FIGURE 4 | The difference between the community average of mean realized niches and the mean environmental conditions for each 4-year window as a function of the standard deviation of environmental conditions. For nitrate and silicate concentration (D, E) increasing variability in the environment led to larger differences between mean environmental conditions and mean realized *(Continued)*

FIGURE 4 | niches (linear regressions, black line, *p* < 0.05). The relationship was non-significant (dashed line, *p* > 0.05) for temperature, mixed layer depth, and salinity (A–C) revealing close tracking of the mean or an extreme value of the environmental conditions for these variables.

TABLE 4 | The correlation matrix between the five environmental variables for all observations between 1997 and 2013. Significant (*P* < 0.05) correlations are in bold.


to change the corresponding niches, but the ability of species to persist at the edge of the distribution of these variables allowed the realized niches to remain largely unaffected over the study period. A third pattern is observed for nitrate and silicate concentrations. Relatively small changes in mean concentrations in the ocean (**Table 2**, **Figures 3D,E**) appear to have resulted in considerably larger changes in the mean niches, and for silicate concentration, the change is in the opposite direction (although the environmental change was not significantly different from zero). These three different kinds of changes in realized niches may be related to the mismatch between the fundamental niche and the available environmental conditions at this site.

Unlike Station CARIACO, where many phytoplankton species displayed a fixed niche for nitrate, the realized niche for nitrate concentration at PH100m tracks an increase in nitrate concentration at the site. Also somewhat surprisingly, for several variables, most notably nitrate and silicate concentration, the mean realized niche over all species, at times, increases more rapidly than the mean concentration in the environment (**Figure 3**). Differences in the concentration and temporal dynamics in nitrate concentrations between the two locations may account for the different phenomena at the two sites. The average monthly nitrate concentration at Station CARIACO was <1 µmol L−<sup>1</sup> , indicating that nitrate may be the limiting resource for many phytoplankton species at this location, whilst the higher concentrations observed at PH100m (∼2 µmol L−<sup>1</sup> ) may not be limiting for many species. Diatoms are known to thrive under variable conditions at PH100m, often taking advantage of nutrient pulses (Hallegraeff and Reid, 1986; Ajani et al., 2014a,b). Increasing nitrate loading in the coastal ecosystem offshore from Sydney is predicted to continue in future years, with more frequent upwelling events anticipated and an increase in anthropogenic nutrient loading via river discharges and ocean outfalls (Pritchard et al., 2003). How phytoplankton will react to this further increase in nitrate concentration will require further investigation over longer time scales and highlights the importance of Australia's National Reference Stations to monitor long-term changes in Australia's ocean.

Because of mismatch between fundamental and realized niche, the consequences of environmental changes on some of the realized niches are not sufficiently well described by changes in the mean conditions. Nitrate and silicate concentrations exhibit increasing frequency of relatively high concentrations over the 1997–2013 time series, starting most noticeably around 2005 (**Figures 3D,E**). The standard deviation of nutrient concentrations, measured over a calendar year, captures this increase and can be used to explain why mean realized niches increase more rapidly than mean environmental conditions (**Table 2**, **Figures 4**, **Supplementary Tables 3**, **4**). The difference between mean realized niche and mean environmental condition is correlated with the standard deviation of the environmental conditions; all measured on a calendar year for nitrate and silicate concentrations (**Figure 4**). The corresponding relationship for temperature, salinity, and mixed layer depth was not significant.

The realized niches of phytoplankton changed rapidly in response to changes in environmental conditions at Port Hacking. We evaluated the lag-times associated with the mean realized environmental niches and environmental conditions and found the correlation was highest with no lag, demonstrating that there was no evidence of a lag even as small as 1 year between environmental change and corresponding changes in the realized niche. Analyses between non-overlapping periods in this study (**Table 2**, **Figure 2**) and a previous study at Station CARIACO showed changes in realized niches between periods but did not address the question of how quickly realized niches changed. This rapid change in realized niche suggests that niche tracking arises from rapid processes such as gene frequency change in the population caused by ecological selection or immigration of new ecotypes. We do not anticipate that physiological plasticity is responsible for these changes in realized niche, except possibly for temperature, since the interannual variation in environmental conditions is much larger than longer-term changes that we are emphasizing. If plasticity was responsible, then the niches would likely be much broader and uniform across species, encompassing the full range of environmental conditions (Irwin et al., 2015). Moreover, since the realized niche for temperature is very close to the mean conditions and tracks the mean very closely, it is possible that some of the species at Port Hacking have broad temperature niches relative to the range of environmental conditions.

Our analysis of phytoplankton realized niches at Port Hacking has demonstrated that realized niches for the species observed at this site approximately track changes in environmental conditions and that these changes happen with a time lag of less than 1 year. This reinforces results observed at Station CARIACO which showed that phytoplankton realized niches adapt to changing ocean conditions. Our analysis at PH100m provides a second, independent test of this idea. The longer time record at Port Hacking included some changes in environmental conditions in opposing directions compared to those observed at Station CARIACO, and some reversals, for example an increase in temperature (and niche) between 1965 and 79 (P1) and 1997–2004 (P2) followed by a decrease between 1997 and 2004 (P2) and 2005–2013 (P3). Our results also extend the previous analysis by illustrating that changes in mean conditions may not always be sufficient to explain the changes in the mean niche. Episodic increases in macronutrient concentrations were sufficient at Port Hacking to enable disproportionate increases in the corresponding realized niches, as phytoplankton exploited these new conditions. Furthermore, the East Australian Current is intensifying (poleward extension of approximately 350 km) and continuing to undergo significant warming (2.28◦C/century) (Ridgway and Hill, 2012). This latitudinal shift and warming trend is predicted to cause shifts in phytoplankton abundance, distribution and composition along the east Australian coast with the emergence of tropical species into more temperate waters already documented (Hallegraeff et al., 2012; Ajani et al., 2014a,b). This spatial restructuring of the plankton may in turn cause changes in biotic interactions (predation, competition), with potential impacts on biogeochemical cycling, higher trophic levels, and biodiversity (Chivers et al., 2017). Taken together, these studies and the details of rates and magnitude of changes in realized niche, emphasize the need to expect changes in phytoplankton niches when designing ecosystem models used to project biotic responses to climate change. This study provides further evidence that climate change model projections cannot assume fixed realized niches of biotic communities, and highlights the importance of sustained ocean measurements from the southern hemisphere (as well as the northern hemisphere) to enhance our understanding of global ocean trends. Moreover, future work to test the proposed mechanisms for the observed niche flexibility should combine population genetics and microbial experimental evolution to allow for a mechanistic understanding how changes in realized niches can be predicted and thereby taken into account in climate change projections.

# AUTHOR CONTRIBUTIONS

PA, ZF, and AI are responsible for the conceptualisation of this work. PA, NM, and AI performed the analysis. All authors (PA, NM, ZF, and AI) are responsible for the writing, reviewing, and editing of this manuscript.

# ACKNOWLEDGMENTS

We acknowledge all contributors, past and present to the PH100m time-series dataset especially Professor Gustaaf Hallegraeff, Ms Alex Coughlan, Professor Anthony Richardson, Dr Tim Ingleton, and Dr Andrew Allen. National Reference Station data are sourced from the Integrated Marine Observing System (IMOS). IMOS is a national collaborative research infrastructure, supported by the Australian Government. Finally, PA would like to thank the University of Technology Sydney Chancellor's Postdoctoral Fellowship scheme for funding; ZF and AI thank the Natural Science and Engineering Research Council (NSERC) Canada for funding. This work was supported by grants from the Simons Foundation (549935, AI; 549937 ZF).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars. 2018.00285/full#supplementary-material

# 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 © 2018 Ajani, McGinty, Finkel and Irwin. 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(s) 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.

# Shifting Diatom—Dinoflagellate Dominance During Spring Bloom in the Baltic Sea and its Potential Effects on Biogeochemical Cycling

Kristian Spilling1,2 \*, Kalle Olli 3,4, Jouni Lehtoranta<sup>1</sup> , Anke Kremp<sup>1</sup> , Letizia Tedesco<sup>1</sup> , Tobias Tamelander <sup>2</sup> , Riina Klais <sup>3</sup> , Heikki Peltonen<sup>1</sup> and Timo Tamminen<sup>1</sup>

<sup>1</sup> Finnish Environment Institute (SYKE), Helsinki, Finland, <sup>2</sup> Tvärminne Zoological Station, Faculty of Biological and Environmental Sciences, University of Helsinki, Hanko, Finland, <sup>3</sup> Institute of Ecology and Earth Sciences, University of Tartu, Tartu, Estonia, <sup>4</sup> Institute of Agricultural and Environmental Sciences, Estonian University of Life Sciences, Tartu, Estonia

#### Edited by:

Michael Arthur St. John, Technical University of Denmark, Denmark

#### Reviewed by:

Maren Voss, Leibniz Institute for Baltic Sea Research (LG), Germany Bernd Wemheuer, University of New South Wales, Australia

\*Correspondence: Kristian Spilling kristian.spilling@environment.fi

#### Specialty section:

This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science

Received: 30 April 2018 Accepted: 27 August 2018 Published: 19 September 2018

#### Citation:

Spilling K, Olli K, Lehtoranta J, Kremp A, Tedesco L, Tamelander T, Klais R, Peltonen H and Tamminen T (2018) Shifting Diatom—Dinoflagellate Dominance During Spring Bloom in the Baltic Sea and its Potential Effects on Biogeochemical Cycling. Front. Mar. Sci. 5:327. doi: 10.3389/fmars.2018.00327 The Baltic Sea is affected by a range of human induced environmental pressures such as eutrophication. Here we synthesize the ongoing shift from diatom dominance toward more dinoflagellates in parts of the Baltic Sea during the spring bloom and its potential effects on biogeochemical cycling of key elements (e.g., C, N, and P). The spring bloom is the period with the highest annual primary production and sinking of organic matter to the sediment. The fate of this organic matter is a key driver for material fluxes, affecting ecosystem functioning and eutrophication feedback loops. The dominant diatoms and dinoflagellates appear to be functionally surrogates as both groups are able to effectively exhaust the wintertime accumulation of inorganic nutrients and produce bloom level biomass that contribute to vertical export of organic matter. However, the groups have very different sedimentation patterns, and the seafloor has variable potential to mineralize the settled biomass in the different sub-basins. While diatoms sink quickly out of the euphotic zone, dinoflagellates sink as inert resting cysts, or lyse in the water column contributing to slowly settling phyto-detritus. The dominance by either phytoplankton group thus directly affects both the summertime nutrient pools of the water column and the input of organic matter to the sediment but to contrasting directions. The proliferation of dinoflagellates with high encystment efficiency could increase sediment retention and burial of organic matter, alleviating the eutrophication problem and improve the environmental status of the Baltic Sea.

Keywords: eutrophication, pelagic-benthic coupling, ecosystem functioning, community composition, plankton sedimentation, carbon sink

# INTRODUCTION

Global change is causing drastic changes to the lowest levels of the marine food web (Halpern et al., 2008; Duarte, 2014), with evidence of shifting community composition of primary producers in e.g., the North Atlantic (Leterme et al., 2005), the North Sea (Hinder et al., 2012), parts of the Mediterranean Sea (Mercado et al., 2007) and the Baltic Sea (Klais et al., 2011). The main drivers/pressures for this change are warming of the surface water, changes in stratification, eutrophication, ocean acidification, overfishing, loss of biodiversity, spreading of non-indigenous species and increasing UV exposure (e.g., Hallegraeff, 2010; Duarte, 2014).

Different species of phytoplankton have different traits (Litchman and Klausmeier, 2008), most notably size and shape, growth rate, life history, and behavior such as motility that together determine their ecological niche and preferred environmental conditions, and phytoplankton is a major driver for global carbon fixation and biogeochemical cycles. State-ofthe-art biogeochemical models typically have several functional groups of phytoplankton, but as the models become more advanced we need empirical studies to disentangle what effect the observed shifts in phytoplankton communities have on ecosystem functioning (Fennel and Neumann, 2014).

The Baltic Sea is among the shelf seas projected to change most rapidly, due to its close interaction with land and reduced alkalinity (Niiranen et al., 2013; BACCI, 2015). Eutrophication is one of the main threats to the Baltic Sea ecosystem (e.g., HELCOM, 2009), but also ocean acidification (Havenhand, 2012; Omstedt et al., 2014), global warming (Belkin, 2009; Meier et al., 2014) and pollutants (HELCOM, 2010) greatly affect this ecosystem and pose challenges for effective management of natural habitats (Elmgren et al., 2015).

Long-term eutrophication in the Baltic Sea has led to accumulation of phosphorus (P), particularly in the sediment, to an extent that internal loading off-sets reduction in nutrient loading from the catchment (Gustafsson et al., 2012). Spread of hypoxia in the bottom water and sediments directly affect the cycling of the main elements (Conley et al., 2011; Carstensen et al., 2014), which in turn affects the community of primary producers when bottom water is transported to the surface during seasonal turnover or upwelling (Cloern, 2001).

Global warming is projected to reduce the sea ice coverage, ice thickness and increase water temperature in the Baltic Sea (Thomas et al., 2017). Expected increase in precipitation in the Northern Baltic catchment will affect freshwater inflow and nutrient run-off (Andersson et al., 2015). Changes to both temperature and freshwater inflow have the potential to change stratification of water layers, with direct implications for vertical transport of O<sup>2</sup> and for planktonic life forms, in particular during the build-up of stratification in spring (Stipa, 2004). Input of freshwater will also influence the concentration of dissolved organic matter with implication for light dependent phytoplankton (Andersson et al., 2018). At present, freshwater induced stratification is important for the initial start of the spring bloom, but thermal stratification may become more important in the future (Hordoir and Meier, 2012). Warming of the Baltic Sea has already caused temporal shifts in the phytoplankton distribution during the highly productive spring, with earlier and more prolonged spring bloom (Groetsch et al., 2016; Kahru et al., 2016). In addition, longterm monitoring data suggest that the phytoplankton community is changing during spring in some areas of the Baltic Sea from diatom to dinoflagellates dominance (Klais et al., 2011), as a consequence of the ongoing climate change (Klais et al., 2013).

We have some understanding of how ongoing environmental changes affect the phytoplankton community, but much less is known how phytoplankton community composition feeds back into ecosystem functioning. Using the Baltic Sea as a case study, we review the present knowledge of the ongoing diatom dinoflagellate shift, and provide a synthesis of the existing knowledge of how the phytoplankton community composition directly affects ecosystem functioning through species-specific sedimentation, life cycle changes and yields of resting stages. We argue that the diatom to dinoflagellate shift has the potential to induce substantial changes in pelagic and benthic ecosystem functioning, and we provide a conceptual model of how this change could affect biogeochemical cycling of key elements. The ratio of diatoms to dinoflagellates was recently suggested to be a new environmental indicator, with a high ratio being associated with good environmental status as diatoms have historically dominated the spring bloom (Wasmund et al., 2017). However, a clear understanding of how this shift may affect the environment is still missing, and we argue that this shift could increase burial of organic matter, which would alleviate the eutrophication problem and improve the environmental status of the Baltic Sea.

# IMPORTANCE OF THE SPRING BLOOM FOR BIOGEOCHEMICAL CYCLES IN THE BALTIC SEA

Long term Chl-a data from western Gulf of Finland indicates a doubling in the biomass peak during spring from early 1970s to mid-1980s; the trend has somewhat reversed after the early 1990s (Raateoja et al., 2005), but it is clear that much of the increased algal production resulting from eutrophication takes place during spring. In the most eutrophied parts of the Baltic Sea such as the Gulf of Finland (**Figure 1**), 40–60% of annual carbon fixation takes place during the spring bloom. This covers only 3–5 weeks of the year, and a large fraction of this fixed carbon sinks to the seafloor (Lignell et al., 1993; Heiskanen, 1998).

The onset of the spring bloom is related to salinity stratification and warming of the surface water (Stipa, 2004). It typically starts in the Southern Baltic Sea in February/March and moves northwards like a mosaic of patches (Kahru and Nõmmann, 1990; Kahru et al., 1990). It reaches the Gulf of Finland in April and the Bothnian Sea and Bothnian Bay in May/June, but in the Bothnian Bay the biomass amplitude is much lower than for most other parts of the Baltic Sea. Two recent papers examining time series of the spring bloom timing point toward an earlier start and a longer bloom with lower biomass amplitude (Groetsch et al., 2016; Kahru et al., 2016) in agreement with modeling scenarios (Thomas et al., 2017). This was attributed to the general warming of the climate where the periods with warm water had expanded temporally over the main Baltic basins (Kahru et al., 2016). Kahru et al. (2016) also reported that the summer community of phytoplankton is becoming more abundant than the spring community based on satellite observations of Chl a. However, the Chl a determination from remote sensing is difficult to estimate (Darecki and Stramski, 2004), and recent direct measurements of Chl a, covering large parts of the Baltic Sea, demonstrate that the Chl a concentration during spring is still much higher than during summer in the main basins of the Baltic Sea (Simis et al., 2017).

The spring bloom in the Baltic Sea (with the exception of Bothnian Bay that is P-limited) is terminated when inorganic nitrogen has been depleted, i.e., the community is N-limited (Wulff et al., 2001; Tamminen and Andersen, 2007). The cascading effect of increased spring blooms during the last decades has led to a regime-shift-like development, i.e., nonlinear change, when inorganic P pools of the Baltic Proper and Gulf of Finland have remained at elevated levels (>0.1 umol L −1 ) after the spring bloom (Spilling, 2007b; Raateoja et al., 2011). Low, hardly detectable phosphate concentrations were as a rule observed prior to mid-1990s, although N limitation of the spring bloom was demonstrated even then (Tamminen, 1995). Similar changes have also been reported in the Bothnian Sea (Lundberg et al., 2009; Rolff and Elfwing, 2015). The loading of P from the catchment in these areas has decreased after mid-1990s, seemingly a paradox, but the reason for this increase in P concentrations is attributed to the release of P from the sediment and exchange between sub-basins, which masks the realized reductions in the P loading (Pitkänen et al., 2001; Kiirikki et al., 2006; Stigebrandt et al., 2014; Lehtoranta et al., 2017). The sediment release of P might be much larger than loading from land in parts of the Baltic Sea (Conley et al., 2002), and the observed regime shift in P availability has seemingly affected both the magnitude and source of the P loading.

Diatoms is the only major phytoplankton group to take up substantial quantities of dissolved silicate (DSi) in addition to N and P. DSi is used to build their cell walls in the form of biogenic silicate (BSi) (Martin-Jézéquel et al., 2000). The DSi originates mainly from natural processes such as weathering of rock, and human activities such as damming of rivers have reduced the natural supply DSi to the Baltic Sea (Humborg et al., 2008). In addition, eutrophication has been suggested to affect the DSi concentration; if the elevated input of N and P increases the biomass of fast growing diatoms, the flux of BSi to the seafloor could increase where it is potentially buried (Schelske and Stoermer, 1971; Conley et al., 1993). However, in the Gulf of Finland the release of DSi from the bottom and back to the water column forms a large part of the pelagic DSi pool (Tallberg et al., 2017).

The concentration of DSi is important as it may influence the competition between diatoms and others phytoplankton groups such as dinoflagellates. For example, the increased N:DSi ratio associated with eutrophication favors non-siliceous phytoplankton (Officer and Ryther, 1980). A molar N:DSi ratio >2 (Gilpin et al., 2004), or absolute concentrations of <2 µmol DSi L−<sup>1</sup> (Egge and Aksnes, 1992) have been suggested to favor flagellates over diatoms. However, some diatoms can acclimate to DSi stress and sustain high growth rates despite low DSi concentrations (Olsen and Paasche, 1986; Brzezinski et al., 1990).

In the Baltic Sea, there has been a long term decrease in DSi concentration that has leveled off after the 2000s (Papush and Danielsson, 2006). There is a difference between the subbasins in terms of silicate budget and most of the accumulation of BSi takes place in the Gulfs of Bothnia, Finland and Riga (Papush et al., 2009), but the spring bloom diatoms are at present not DSi limited (Wasmund et al., 2013), perhaps with the exception of Gulf of Riga (Olli et al., 2008). Shifting nutrient stoichiometry could shift the competitive balance to non-siliceous phytoplankton in the future (Danielsson et al., 2008). However, many of the dominant Baltic Sea diatoms seem to be less silicified in terms of the N:Si ratio than the Redfield ratio (Spilling et al., 2010) and consequently less sensitive to low DSi concentration. An exception are their resting spores that typically have thicker frustules than the vegetative forms, and the spore formation could be affected by low DSi concentration at the end of the spring bloom (Kremp et al., 2008).

# Changing Phytoplankton Community

The phytoplankton community during the highly productive spring is dominated by diatoms and dinoflagellates in most of the Baltic Sea (Niemi, 1975; Heiskanen, 1993; Wasmund et al., 1998; Höglander et al., 2004; Tamelander and Heiskanen, 2004; Jaanus et al., 2006). Diatoms are generally very successful during periods of high new production (i.e., production based on accumulated nitrate), like the spring bloom, as their higher growth rate enables them to outcompete e.g., dinoflagellates (Reynolds, 2006). This is also the case for the spring phytoplankton community in the Baltic Sea, where the dinoflagellates are not able to achieve growth rates comparable to diatoms under controlled lab conditions (Spilling and Markager, 2008). Rather, the recruitment, affecting abundances before the spring bloom start, seem to govern the success of dinoflagellates during spring bloom (Kremp et al., 2008). Additionally, certain sequences of weather conditions, particularly stratification periods at the onset of the spring bloom have been related to subsequent dinoflagellate dominance (Heiskanen, 1998; Höglander et al., 2004; Klais et al., 2013). Stratification beneath the ice may serve as a platform for dinoflagellates to build up high biomass already during winter (Spilling, 2007a). In addition, some dinoflagellates are mixotrophs (uptake or feeding on organic components), which may supplement C fixation during light limited conditions (Carlsson and Graneli, 1998; Legrand and Carlsson, 1998; Rintala et al., 2007).

In spite of varying dominance of diatoms and dinoflagellates during the spring bloom period, both of the groups are able to effectively exhaust the wintertime accumulation of inorganic N, and produce bloom-level biomasses, and they thus appear to be functional surrogates (Kremp et al., 2008). However, the stoichiometry of the sinking material varies depending on the species composition, as diatom-dominated communities have a higher C: N: P ratio than mixed or dinoflagellate-dominated communities (Spilling et al., 2014). Changes in C: N: P ratio may have consequences for food quality and remineralization in the pelagic and benthic systems of the Baltic Sea. Grazing pressure is relatively low during the spring bloom (Lignell et al., 1993), which is the reason why much of the biomass sink to the seafloor. However, the planktonic grazer communities can be affected as changes in spring bloom composition influence the quality of food available for the emerging copepod populations (Vehmaa et al., 2011), and increasing temperature may boost heterotrophic activity and reduce the export of phytoplankton to the seafloor in the future (Tamelander et al., 2017).

The phytoplankton community composition may also affect the structure of the sinking biomass that may influence the benthic biogeochemistry. Diatoms are worldwide known to be important vehicles for transporting fixed carbon from the atmosphere to great depths (Doney, 1997; Smetacek, 1998). In the relatively shallow Baltic Sea, a large fraction of this carbon sinks to the seafloor. The majority of vegetative dinoflagellate cells lyse, in contrast to diatoms, leading to remineralization in the pelagic zone or forming of slow sinking phytodetritus (Heiskanen, 1998; Tamelander and Heiskanen, 2004). Dinoflagellates can also constitute a major fraction of the sedimentation flux, with sharp peaks during mass encystment, depending on the species and their respective encystment strategies. Built for longterm survival, the cysts that do settle, do not contribute to the benthic oxygen demand (Spilling and Lindström, 2008), although they have a similar carbon content (roughly 50% of the dry weight). Diatoms also have resting stages called spores, and there is a large variation in the amount of resting spores being produced after the bloom depending on the species, ranging from virtually none to most of the cells forming spores (e.g., Kuosa et al., 1997). The different life history traits have also been suggested to impact sedimentation of phytoplankton biomass quantitatively, with higher vertical export from diatom dominated blooms (Heiskanen, 1998). However, vertical export estimates of phytoplankton and total particulate organic carbon (POC) export vary greatly among years characterized by diatom or dinoflagellate dominance (Heiskanen, 1993, 1995; Tallberg and Heiskanen, 1998; Tamelander and Heiskanen, 2004), with no consistent trend. Conceivably, both intact cells (diatoms, spores, and cysts) and detritus (mainly from vegetative dinoflagellate cells) may settle but the share of resting stages will depend on environmental effects on cyst/spore formation and deposition, and on the life cycle strategy of the prevailing species (Kremp et al., 2009; Warns et al., 2012).

Consequently, the phytoplankton composition during the spring bloom may affect both the summertime nutrient pools of the water column, and the input of labile organic matter to the bottom sediments. In the case of diatom dominance, drawdown of nutrients to the bottom is efficient and leaves impoverished nutrient stocks for summertime regenerated production in the euphotic zone. During dinoflagellate dominance, more organic material may be available for remineralization in the productive surface layer, supporting recycling through the microbial loop. However, a complicating factor, which may counteract this assumption, is the active release of dissolved organic matter that is more prevalent during diatom dominance (Camarena-Gómez et al., 2018). This can be done for several reasons (e.g., Thornton, 2014), for example stimulating bacterial growth, which may provide some benefits in return for phytoplankton like the production of B12 vitamin (Kazamia et al., 2012), or be a way to dissipate excess light energy (Zlotnik and Dubinsky, 1989).

# CHANGING SPECIES COMPOSITION–A FUNCTIONAL SHIFT IN THE MAKING?

In the last few decades, the proportion of dinoflagellates in the spring bloom biomass has increased in the northern Baltic Sea, most notably in the Gulf of Finland. For example, monitoring data from the waters off the city of Helsinki (see Olli et al., 2011 for description of data), reveal that in spite of substantial variation, the proportion of dinoflagellates has increased from 10 to 20% in the 1970s up to around 80% by the turn of the century (**Figure 2**). The increase of the dinoflagellate proportion has been, with varying strength, a common phenomenon in large parts of the Gulf of Finland, including the north-western part of the Gulf (e.g., archipelago off the Hanko peninsula), but also in the eastern parts of the Gulf (Klais et al., 2011). Overall, the previously diatom dominated spring blooms in the eutrophied Gulf of Finland have incrementally shifted to dinoflagellate dominated blooms in just three decades, but with large inter-annual variation. The inter-annual variation is partly climate driven (Klais et al., 2013), but the reason for the decrease of dinoflagellates in the 1990s outside Helsinki (**Figure 2**) is not yet resolved, suggesting other variables affecting the population dynamics e.g., the timing of recruitment is important for dinoflagellate development (Kremp et al., 2008).

In spite of the general increase in the dinoflagellate proportion in the Gulf of Finland, opposing decadal scale trends have occurred in other sub-basins. In the central Baltic Proper, where the proportion of dinoflagellates was ca. 80–90% for many decades, a slow decrease have been observed (Wasmund et al., 1998). Wasmund and Uhlig (2003) reported a weak increase of spring bloom dinoflagellate biomass in the southern Baltic Sea, but gave no data on proportions. In the eutrophied

Gulf of Riga, strong spring blooms of heavily silicified diatom species exhausted the dissolved silicate stocks in mid 1990s (Olli et al., 2008), giving rise to a short period of high dinoflagellate proportions, in the otherwise diatom dominated basin.

year. The contour line shows the 75% probability level.

#### Dinoflagellates

The dinoflagellate spring bloom community includes several species that can reach high biomasses depending on conditions and location. Single-celled, medium-sized, oval-shaped dinoflagellates regularly reach high abundances in the northern basins and the central Baltic Sea. These were previously considered as one species, Scrippsiella hangoei, but high resolution scanning electron microscopy and rDNA analyses revealed that this entity is represented by at least three different species belonging to different dinoflagellate orders: Biecheleria baltica (syn Woloszynskia halophila), Apocalathium malmogiense (syn S. hangoei), and Gymnodinium corollarium (Larsen et al., 1995; Kremp et al., 2005; Moestrup et al., 2009). Records of their specific resting cysts suggest that B. baltica is mainly found in the northern basins, particularly the Gulf of Finland (Olli and Trunov, 2010), while G. corollarium has its main distribution in the central Baltic Sea (Sundström et al., 2009). A. malmogiense cysts are rare in Baltic surface sediments, suggesting that this species is of a minor importance in the spring dinoflagellate community (Kremp et al., 2018). In addition to the three single-celled species, the chain forming, arctic Peridiniella catenata can reach high abundances in some years during spring and it has a wide distribution in the Baltic Sea. If B. baltica prevails in the Gulf of Finland, the shift to dinoflagellate dominated spring blooms can be ascribed to the proliferation of this species alone (**Figure 3**). To the best our knowledge, B. baltica has not been found in any other water body and its origin remains thus enigmatic, but it seems to have a center of expansion in the western Gulf of Finland, from where it has in recent decades spread east- and westwards. Due to the potential of B. baltica to completely take over the spring bloom in just a few decades, its unique life history traits, recruitment preferences, and biogeochemical consequences of dominance are of great interest (see below). Despite the conspicuous expansion of B. baltica in the northern Baltic Sea, the species is virtually absent in the neighboring Gulf of Riga (Olli and Heiskanen, 1999), suggesting that local hydrography can act as an efficient dispersal barrier.

The major spring bloom dinoflagellates differ from each other in several functional traits, which may affect biogeochemical pathways, e.g., different life cycle strategies may affect sedimentation rate of the population. Formation and subsequent sedimentation of cysts now known to be produced by B. baltica at the end of the spring bloom can contribute up to 45% of the total vertical POC flux (Heiskanen, 1993) Notable cyst fluxes have also been documented for G. corollarium from the Bornholm and Gotland basins (Sundström et al., 2009). For P. catenata, it has been established that most of the biomass produced by this species during spring disintegrates in the water column at the end of the bloom (Heiskanen and Kononen, 1994), and only a minor part of the population undergoes encystment (Kremp, 2000; Spilling et al., 2006; Olli and Trunov, 2010).

Species-specific life cycle strategies, involving fundamentally different mechanisms leading to the respective transitions, likely affect biogeochemical processes. High cyst yields are triggered by different environmental factors and physiological processes (Kremp et al., 2009). Depending on the dominant species, specific temperature and inorganic nutrient dynamics during spring will induce different life cycle responses, with different outcomes regarding vertical fluxes of cyst bound carbon. This has been demonstrated by simulations of numerical models that explicitly consider the regulation of life cycle transitions (Warns et al., 2013). Species-specific differences also exist in germination efficiencies of cysts, which may affect the fate of cyst-bound carbon in the sediment. Of the B. baltica cysts produced, only 10–20% will germinate under optimal conditions the following spring (Kremp, 2000), which implies that B. baltica blooms could support long-term organic carbon storage in the sediments.

The behavioral adaptations of the vernal bloom influence the biogeochemical cycling within the water column. At the late phase of the spring bloom, the vertically migrating dinoflagellate cells (Olli et al., 1998) descend below the euphotic layer (30– 40 m depth), where cyst formation takes place (Spilling et al., 2006). If only a few percent of the cells encyst at most, the bulk of the biomass disintegrates and contributes to the DOM and slowly sinking phytodetrital pools in mid-depths of the water column. This resource, re-mineralized nutrients and organic

carbon substrate, will only have restricted availability to the organisms in the euphotic layer, as the rapidly establishing thermocline effectively segregates the water column. The single cell dinoflagellates, i.e., Biecheleria complex, was not found to migrate vertically in the coastal Gulf of Finland (Olli et al., 1998), but was shown to do so in the open waters of the north-western Baltic Proper, as a result of mineral nutrient depletion of the surface layer (Höglander et al., 2004). Thus the behavioral adaptations are not only species-specific, but probably also reflect regional variations in hydrodynamics and nutrient availability.

#### Diatoms

The diatom community tends to be more diverse during bloom conditions, compared to dinoflagellates, and functions more as a guild of complementing species (Smayda and Reynolds, 2003). This is also the case in the Baltic Sea where there are more diatom species present than dinoflagellates, but only a few species occur throughout the Baltic Sea during spring, making up a large fraction of the diatom biomass. Some of these diatoms are also present during winter in connection to the ice (Ikävalko and Thomsen, 1997), and can also be found in the Arctic, e.g., Achnanthes taeniata and Melosira arctica. These two and Thalassiosira levanderi are generally abundant in the initial phase of the spring bloom. Other abundant diatoms during the main bloom phase are Chaetoceros wighamii, C. holsaticus, Diatoma tenuis, T. baltica, and Skeletonema marinoi, where the latter typically become relatively more dominating after the peak of the spring bloom in the Gulf of Finland (Spilling, 2007b). Although some of these species occur also outside the Baltic Sea, the strong salinity gradient seems to be an effective barrier for gene exchange (Sjöqvist et al., 2015).

Diatom spores are less studied than dinoflagellate cysts in the Baltic Sea, but they seem to share many of the same characteristics. The fraction of the population that goes through the life-cycle change to form spores is highly species-specific and affected by environmental conditions (Heiskanen and Kononen, 1994; Kuosa et al., 1997). For example, A. taeniata, C. holsaticus, and M. arctica may settle to the seafloor primarily as spores, but the spore formation is affected e.g., by the availability of DSi, and the spore formation could be lower under less favorable conditions (Kuosa et al., 1997; Heiskanen, 1998). The functional role of diatom spores is similar to dinoflagellate cysts, i.e., long term survival in the sediment, and would thus have very much the same traits in terms of how they affect biogeochemical cycles. They can be buried and can be a useful tool for studies of paleoenvironmental changes (Witak et al., 2011). Interestingly, some spores (and also dinoflagellate cysts) are able to survive for up to a century in the sediment, enabling genetic comparison with present day strains (Härnström et al., 2011; Kremp et al., 2018). To our knowledge, there are no good estimates of how much carbon and nutrients are buried in the sediments in the form of diatom spores annually, but it is clear that the community composition is one of the key components determining this.

# LINKING SEDIMENTATION OF PHYTOPLANKTON TO MICROBIAL MINERALIZATION PROCESSES

The main driver of sediment nutrient cycling is organic matter serving the energy source for heterotrophic microorganisms (Berner, 1974; Froelich et al., 1979). In addition to the organic matter, the availability of electron acceptors (O2, NO3, Mn, and Fe oxides and SO4) create the premises for the pathways of organic carbon mineralization. In the Baltic Sea, availability of organic matter and electron acceptors vary in the bottoms of the different sub-basins. For example, the sediments of the Gulf of Bothnia are organic poor compared to the Gulf of Riga, the Baltic Proper and the Gulf of Finland (Lehtoranta et al., 2008; Aigars et al., 2015; Egger et al., 2015). The Bothnian Bay is oxic down to the seafloor and O<sup>2</sup> may penetrate centimeters into the sediment (Slomp et al., 2013). The sediment of the Gulf of Riga is oxic as well, but the penetration depth varies considerably throughout the year (Aigars et al., 2015). In the oxic areas, the bottoms can use the entire set of electron acceptors to mineralize settled phytoplankton.

In contrast to the Gulf of Bothnia, the basins of the Baltic Proper suffer from long-term hypoxia and anoxia, and the O<sup>2</sup> penetrates into the sediment only in the shallow slopes with oxic water and after major Baltic inflow events (Almroth-Rosell et al., 2015; Mohrholz et al., 2015). Further north, at the entrance of the Gulf of Finland, anoxic periods vary inter-annually and seasonally, being most common in summer (Lehtoranta et al., 2017). The pore-water concentrations of dissolved Fe are low in the sediments of this region (Lehtoranta and Heiskanen, 2003; Jilbert et al., 2011), but, the concentration of sulfate is twice as high as in the Bothnian Bay. So in the Baltic Proper and the Gulf of Finland microbes have reduced access to O2, NO3, and Fe(III), but a continuous contact to abundant sulfate.

To summarize, the ability of the benthic systems to mineralize the settled phytoplankton, their cysts, spores, and detritus is subbasin specific in the Baltic Sea. The marked changes in the pelagic ecosystem or in the hydrodynamics have the potential to induce sub-basin specific shifts in the mineralization pathways which regulate the cycles of C, N, P, Fe, Mn, and S in the sediments.

## Settling Spring Bloom Biomass and Mineralization of N and P

Nitrate (NO3) can be used as a nutrient in the pelagic zone and as an oxidant in mineralization and in oxidation of reduced substances such as Fe2+, HS−, H2S, and CH<sup>4</sup> leading to formation of N<sup>2</sup> gas in sediments. In the oxic sediments of the Baltic Sea, the coupled nitrification-denitrification takes place right below the interface between oxic and anoxic layers, and is considered to be the major pathway for denitrification in the Baltic Sea (Hietanen and Kuparinen, 2008). The organic N is mineralized first to ammonium (NH4) and then oxidized in the presence of O<sup>2</sup> to nitrite (NO2) and subsequently to NO<sup>3</sup> which feeds the denitrification process. This process forms a link between settled organic matter, nitrogen, and respiration because organic C is oxidized by NO<sup>3</sup> to CO<sup>2</sup> and nitrate is denitrified to N<sup>2</sup> gas.

Denitrification is often seen as a negative feedback mechanism for eutrophication processes as it removes N from the system. Oxic conditions are needed to maintain the stock of NO<sup>3</sup> for the coupled nitrification-denitrification, but anoxic condition is obligatory for denitrification. Vahtera et al. (2007a) suggested that the increase in size of hypoxic area in the Baltic Sea is linked to the decrease in storage of N through removal of N by denitrification. However, stratification and excess organic matter may deplete O2, which inhibits the formation of NO<sup>3</sup> and may lead to accumulation of NH<sup>4</sup> and decrease the N removal through denitrification in sediments (Jäntti and Hietanen, 2012). There is an alternative pathway for N removal through ammonium oxidation coupled to nitrite reduction, also leading to the production of gaseous forms of N (mainly N<sup>2</sup> but also NO and N2O) (Kuypers et al., 2003). This N removing mechanism, anammox, requires one oxidized N species (NO<sup>−</sup> 2 ) for every N<sup>2</sup> molecule made, which is half the requirement of denitrification. Anammox bacteria grow relatively slowly and the lack of NO<sup>−</sup> 2 will also stop the anammox process. However, N removal might in this case take place further up in the water column in the Baltic Proper (Dalsgaard et al., 2013), but this is a process that is not very pronounced in other sub basins such as the Gulf of Finland (Jäntti and Hietanen, 2012).

The cycling of sediment P is intimately linked to the mineralization of the settling phytoplankton and detritus. Phosphate itself is not a redox-sensitive substance, but there is commonly a strong, negative correlation between O<sup>2</sup> and phosphate concentrations. This relationship can largely be explained with the coupling of P to the Fe cycling (Einsele, 1938). Based on literature and the tests with estuarine sediments from the Gulf of Finland, it is evident that the mineralization through Fe(III) oxides and Fe reduction by H2S (or HS−) may result in markedly different outcomes regarding cycling of P in different sub-basins (Lehtoranta et al., 2009).

In the Bothnian Bay, the concentrations of dissolved Fe2<sup>+</sup> and P in pore water indicate that the microbial Fe reduction may maintain the coupled cycling of Fe and P in the sediments (Blomqvist et al., 2004; Slomp et al., 2013). The Bothnian Bay, therefore, has a great potential to capture the P mineralized from settled phytoplankton and detritus to newly forming Fe(III) oxides in the oxic zone of the sediment. Also the burial of the reduced Fe phosphate mineral vivianite (Fe3(PO4) . 2 8H2O), may enhance the P binding in the methanogenic zone of the Gulf of Bothnia (Slomp et al., 2013; Egger et al., 2015). In contrast to the Bothnian Bay, the formation of solid Fe sulfide minerals may block the cycling of Fe in the Baltic Proper and the Gulf of Finland (Lehtoranta et al., 2008). In these areas the dominance of sulfate reduction creates sulfidic conditions and the P mineralized from phytoplankton and Fe(III) oxides may freely escape from the sediment to water without sequestration by Fe. From a water quality perspective, the formation of Fe sulfides rapidly deteriorates environmental conditions as the ability of the sediment to sorb P into Fe is largely lost. This is the situation in the Baltic Proper and the Gulf of Finland where the sediments ability to capture P originating from settling algae has been reduced or lost.

It is evident that the amount of settled phytoplankton and detritus has direct consequences for the cycling of N and P in the Baltic Sea ecosystem. What is less apparent is the qualitative aspect of the settling biomass and what electron acceptor may be used in the mineralization process. Spilling and Lindström (2008) added different cultures of spring bloom algae to a natural sediment, and found that they produced very different responses in the sediment, depending on species-specific life cycle change. In short, a dinoflagellate that produced cysts did not contribute to oxygen depletion, NH<sup>4</sup> formation and P release. The vegetative stages, in contrast, (of both a diatom and another dinoflagellate) quickly turned the sediment anoxic with the formation of Fe sulfides, NH<sup>4</sup> and release of P. This highlights the qualitative aspect of the settling material, i.e., not all of the phytoplankton settled is easily mineralized, and this is to our knowledge not taken into account in any models of biogeochemical cycling.

Resting stages, being cysts or spores, are built to survive for a long time in the sediment. The qualitative aspect of the settling biomass cannot be divided into dinoflagellates or diatoms as such, but is linked to life cycle changes. However, the amount of resting stages made after the spring bloom is species-specific, and dinoflagellates tend to produce more carbon bound to resting stages than diatoms. In particular B. baltica and G. corollarium go through mass encystment and because their germination success is low the following year, they are effective vehicles for carbon burial in the sediment. At least B. baltica has increased its biomass substantially over the past decades, and if this continues a larger share of the biomass produced during spring could be buried permanently at the seafloor.

# Sediment Thresholds Controlling Nutrient Dynamics

With increasing transport of phytoplankton to the seafloor, there will generally be successive, non-linear thresholds that change the mineralization pathways and geochemical conditions in the sediment. The first threshold takes place when stratification decreases the transport rate of O<sup>2</sup> downwards, or there is an excess of organic C input enhancing the oxygen consumption. Both cases lead to anoxia first in the sediment and then in overlying water. Under these conditions, the coupled nitrification-denitrification is blocked and the NO<sup>3</sup> flux from water to sediment starts to control the denitrification, which is the case in the Gulf of Riga (Aigars et al., 2015). Without oxidation of NH<sup>4</sup> only few micromoles of labile carbon served by phytoplankton is required to deplete NO<sup>3</sup> as the near-bottom water concentration of NO<sup>3</sup> varies commonly from 2 to 10 µmol L −1 in the Gulf of Finland. Eventually the anoxic conditions result in the accumulation of NH<sup>4</sup> in the near-bottom water. Then the oxidation of released NH<sup>4</sup> to NO<sup>3</sup> may occur only in the chemocline between oxic and anoxic water layers and the formed NO<sup>3</sup> may be denitrified with organic matter and sulfides present in anoxic water as in the Gotland Basin (Dalsgaard et al., 2013). We have modified the conceptual model of Vahtera et al. (2007a) to include the threshold regarding the cycling of N (**Figure 4**). Although part of the N removal may occur above the seafloor (Dalsgaard et al., 2013), the potential termination of N-removal could form a positive feedback loop for eutrophication.

After depletion of NO<sup>3</sup> and in the presence of labile organic C in the sediment, the system is pushed toward a second threshold when microbe-driven Fe(III) reduction results in and simultaneous release of Fe and P to the near-bottom water. When Fe is diffused or transported during vertical mixing or upwelling events to an oxic environment, the newly formed Fe(III) oxides have the capacity to resorb P originating from the settled phytoplankton.

The concentration of poorly crystalline Fe(III) is suggested to control the Fe reduction rate. Four moles of Fe is needed to oxidize one mole of organic carbon and without re-oxidation by oxygen, NO3, and Mn oxides, the pool of bioavailable Fe(III) oxides can be consumed within days in organic rich marine sediments, but large regional differences are evident (Rysgaard et al., 2001; Canfield et al., 2005). The depletion of bioavailable Fe(III) oxides leads to a third threshold, where sulfate reduction is the dominant mineralization pathway for settled phytoplankton biomass. The passing of the third threshold forms enough sulfides capable to precipitate with reactive Fe and block its cycling. Under the sulfidic conditions, the sediment and water column has poor ability to bind P to Fe, which results in a large release of Fe bound P and organic P and a minor release of Fe to the overlaying water. Thus, in the Baltic Sea the large pulse of P in the sulfidic conditions is explained by the dissolution of Fe(III) oxide bound P and the mineralization of fresh organic P that may be released within a short period of time compared to "old" organic P, which may take years to remineralization in sediments (Ahlgren et al., 2006). Additionally, the high amount of labile C and abundance of SO<sup>4</sup> maintains sulfate reduction and leads to accumulation of free sulfides in the water column as found in the bottom water of the Baltic Proper and occasionally in the Gulf of Finland. The oxidation of these sulfides in the chemocline consumes O2, NO3, Mn(IV), and Fe(III) present in the water column, delaying the recovery of the sedimentary system of the Baltic Sea. Typically, there is too little Fe to bind P in the water and the ability of the sediment to resorb P to Fe(III) oxides depends on the re-oxidation of the reduced Fe(II). The oxidation of Fe by O<sup>2</sup> in the water takes place at a very thin surface layer, but may be extended with the help of benthic animals that burrows into the sediment, such as Marenzelleria spp., an alien species that has spread to large bottom areas in the Baltic (Norkko et al., 2012; Maximov et al., 2015).

Modeling of biogeochemical fluxes relies on the information of sediment processes and our understanding of these sediment thresholds as eutrophication increases the amount of sinking phytoplankton. We have concentrated on solving how easily degraded organic matter consume the various electron acceptors in the sediment, and how this affects material fluxes through the sediment-water interface. However, the qualitative aspect of the settling material as described above should be taken into consideration (Stolpovsky et al., 2015), in particular under the scenario with changing composition of the sinking organic matter. Another important biological process to consider is bioturbation by benthic animals, which will affect sediment processes e.g., by increasing oxygen penetration depth into the sediment (Norkko et al., 2012). Understanding the complex interplay between settling phytoplankton biomass, its quality (labile, semi-labile, or refractory) and biological activity in the sediment is the target of ongoing modeling efforts (e.g., Stolpovsky et al., 2018). In the Baltic Sea, an important step would be to better predict the quantity of the settling organic matter that will trigger the three sediment thresholds described above, taking the qualitative aspect of the biomass and bioturbation into consideration.

# PELAGIC NUTRIENT DYNAMICS—FROM SPRING TO SUMMER

One of the highly visible effects of the eutrophication problem in the Baltic Sea is the increased biomass of diazotrophic cyanobacteria during summer (Elmgren et al., 2015). The prevailing N limitation at the end of the spring bloom leads to a low N:P ratio during summer, which favors N-fixing cyanobacteria during warm, calm summer months (Niemi, 1979; Wasmund et al., 2005). The cyanobacterial blooms directly affect the eutrophication problem, as their fixation of N is comparable in magnitude with anthropogenic loading of N in parts of the Baltic Sea (Savchuk, 2005; Wasmund et al., 2005). In addition to the cyanobacterial biomass, the part of the newly fixed N is released and may be taken up by other components in the plankton food web (Adam et al., 2016), and this may start right after the spring bloom when the water is still relatively cold (10◦C) (Svedén et al., 2015). The cyanobacterial input of N might have a self-enforcing nature by increasing the biomass load reaching the sediment, aggravating oxygen consumption and increasing release of P (Tamminen and Andersen, 2007; Vahtera et al., 2007a). However, we argue that the species composition

of the spring bloom could affect this: with the shift toward a more dinoflagellate-dominated community (Wasmund and Uhlig, 2003; Klais et al., 2011), the sharp flux of diatom biomass to the sediment would diminish. This could have ecosystem-wide consequences as the organic material transport to the sediment affects both nitrogen and phosphorus cycles (Vahtera et al., 2007a). To some extent, the sedimentation pattern could be prolonged with more recycling in the euphotic zone (Tamelander et al., 2017), but at present we miss an understanding of how the observed shift in the spring phytoplankton community has on sedimentation and material fluxes on a large scale.

The pool of bioavailable P at the onset of summer is critical for the development of the cyanobacterial blooms later in the season. This has increased over large areas since the mid-1990s. In the Gulf of Finland and Baltic Proper the concentration may be as high as >0.5 µmol PO<sup>4</sup> L −1 after N sources have been depleted (**Figure 5**). In the Bothnian Sea the development is the same but with lower remaining P pool and Bothnian Bay is P limited and has an N surplus.

The two main cyanobacterial species: Aphanizomenon flosaquae and Nodularia spumigena have different P uptake strategies. A. flos-aquae requires inorganic phosphate and can take up and store excess P for later growth (Larsson et al., 2001; Kangro et al., 2007), whereas N. spumigena has higher affinity for P uptake and relies more on recycled P (Hagström et al., 2001; Vahtera et al., 2007b). Under stable conditions there is also a spatial separation between the main summer cyanobacteria where A. flos-aquae typically remains lower in the water column whereas N. spumigena is found at the surface (Hajdu et al., 2007). A. flos-aquae is the main diazotroph in the Northern Baltic Proper, and in addition to N. spumigena, also Dolichospermum spp. contributes to the N-fixation (Klawonn et al., 2016). It is clear that cyanobacteria is benefitting from the oversupply of P (Raateoja et al., 2011), but several open questions remain. The dominant cyanobacterial species grow relatively slowly, in particular in water temperatures below 15◦C (but A. flos-aquae is less temperature sensitive than N. spumigena), and there is typically a relatively short time window between the draw-down of bioavailable P and cyanobacteria growth, and it is not enough to satisfy the cyanobacterial need (Walve and Larsson, 2007; Nausch et al., 2008), so there must be other components in the plankton community that takes up the excess P. These could be other primary producers or heterotrophic bacteria (Nausch and Nausch, 2004). The prevailing N-limitation, limits growth of other phytoplankton and consequently inhibits their uptake of excess P. Heterotrophic bacteria is probably not limited by N, but might be limited by carbon (Lignell et al., 1992), at least their P uptake intensifies when they have a sufficient carbon source (Tamminen, 1989). To our knowledge, it is not clear what plankton groups takes up the excess P, but it does seem to accumulate in the particulate fraction and settle before the onset of dense cyanobacterial blooms (Nausch et al., 2008). It is evident that other P sources must exist for the cyanobacterial blooms (Raateoja et al., 2011), most likely through upwelling (Wasmund et al., 2012).

FIGURE 5 | Nutrient concentrations (µmol L−<sup>1</sup> ) at the surface (3 m depth) following the spring bloom. The samples were taken in May 2015 (A,B) and April 2016 (C,D) during the CFLUX15 and CFLUX16 research cruise, respectively, on board R/V Aranda. The maps have been modified from Spilling (2015) and Spilling (2016).

# Role of Dissolved Organic Matter in Pelagic Nutrient Cycling

Due to the close connection with land with substantial freshwater input, the Baltic Sea has a large pool of dissolved organic matter (DOM) originating from outside the system, i.e., allochthonous DOM. In addition, there are several pathways from primary production into the DOM pool, such as direct release of exudates, lysis of cells, and sloppy feeding by grazers, i.e., autochthonous DOM. DOM is a potential source of carbon (DOC), nitrogen (DON) and phosphorus (DOP) for bacterial production, which may lead to remineralization of N and P back to inorganic forms available for phytoplankton. It may also provide organic matter to higher trophic levels through the microbial loop, although this is quite inefficient in trophic transfer (Anderson and Ducklow, 2001). Part of the allochthonous DOM, approximately 10% of DOC (Asmala et al., 2013), and ∼30% of DON and 75% of DOP (Stepanauskas et al., 2002) is taken up by bacteria. Photochemical degradation of allochthonous DOM can also support plankton production (Vähätalo et al., 2011), and in total, approximately 50% of the riverine DOM is removed (Seidel et al., 2017). The source of DOM can also be important as it may affect the phytoplankton community composition by favoring certain species (Jurgensone and Aigars, 2012). In the future, a warmer climate might lead to more precipitation and increased input of allochthonous DOM, in particular in the Northern Baltic Sea (Andersson et al., 2015), which might reduce light-dependent primary production and increase the importance of pelagic, heterotrophic bacteria (Wikner and Andersson, 2012; Andersson et al., 2018).

The input of autochthonous DOC from primary producers ranges from 2 to 50% of the fixed carbon in estuaries, averaging around 12% (Baines and Pace, 1991). In the Baltic Sea, there is an increase of DOM during the productive season (Hoikkala et al., 2012), calculated to be 20–200 µmol DOC L −1 (Hoikkala et al., 2015). Only a part of the DOC from lysing cells is easily available for bacterial degradation, and in their review of different aquatic habitats, Søndergaard and Middelboe (1995) reported an average of 19% of the total dissolved organic carbon (DOC) pool to be labile in marine waters. However, higher percentages of 25– 55% have been found in the Baltic Sea (Hoikkala et al., 2016), and Hoikkala et al. (2015) presented a conceptual model, where ∼50% of DOC produced by phytoplankton is rapidly degraded by bacteria. As the spring bloom depletes inorganic N sources in most of the sub basins of the Baltic Sea, autochthonous DON is an important component for primary and bacterial production (Korth et al., 2012). Later in the productive season N-fixing cyanobacteria is generally P limited, and DOP is potentially important. However, Nausch et al. (2008) studied the P pools in Baltic Proper and found a stable DOP pool suggesting that production and consumption was equal, and DOP was not a major P source for cyanobacteria.

After a dinoflagellate-dominated spring bloom, more DOC can be expected to be released in the pelagic zone, based on sedimentation studies (Heiskanen, 1998). However, diatoms are generally high DOC producers, and they can excrete organic compounds leading to even higher DOC concentrations than dinoflagellate-dominated blooms (Spilling et al., 2014). There are species-specific differences in DOC excretion, and at least a part of the heterotrophic bacterial community is affected by the phytoplankton community composition (Bunse et al., 2016), affecting also the responses in bacterial production (Camarena-Gómez et al., 2018). There is thus a potential link between the phytoplankton community composition during spring and the early summer DOM pool, with the potential to affect the cyanobacterial blooms during summer. This link has so far not received much attention, and it is currently not taken into account in model predictions of P availability and cascading effects on cyanobacterial bloom development.

# A CONCEPTUAL MODEL

In the relatively shallow coastal and shelf ecosystems, the pelagicbenthic coupling is more important for ecosystem functioning than in the deep oceans. Changes in the phytoplankton communities may consequently lead to spatial and temporal differences in the quantity and quality of organic matter input to the seafloor, which in turn is the main driver for biogeochemical cycling of electron acceptors and nutrients. The conceptual model presented in **Figure 6** summarizes how different biogeochemical pathways may be altered by the phytoplankton community composition in the Baltic Sea, as described above. In particular, we represent the potential feedback mechanisms of changing plankton community in a typical site in the Gulf of Finland, characterized by seasonal sea ice and muddy, soft and oxic sediments inhabited by the two dominant macrozoobenthos species of soft bottom seabed, i.e., Marenzelleria spp. and Macoma baltica, which differentiate mostly by their different feeding habits, efficiency of bioturbation activities and location within sediments.

During harsh winters with a long ice season and thick ice cover, the sea-ice algal bloom typically dominated by diatoms is followed by a pelagic spring bloom of diatoms of similar species (e.g., Chaetoceros spp. T. baltica, and A. taeniata (**Figures 6A,B**). The bloom may be loosely grazed or degraded in the water column (up to 10%, Lignell et al., 1993), but most of it sinks to the seafloor. Under oxic sediment conditions, this large pool of organic matter may be efficiently consumed by the macrozoobenthos. With ample food supply, the macrozoobenthos community might increase and improve the oxygen conditions in the sediments through bioturbation (**Figure 6A**). However, if the amount of settling material exceeds the threshold of being efficiently consumed by the bottom communities, the result is aggravated oxygen consumption leading to permanent or seasonal hypoxia (Norkko et al., 2012). This, in turn, may cause the local fauna to perish with obvious loss of biodiversity, and may lead to an efflux of P from the anoxic sediments, further enhancing the summer bloom of cyanobacteria (**Figure 6B**), and fuelling the vicious cycle of Vahtera et al. (2007a) (**Figure 4**).

During mild winters, with a thin ice cover and short duration, winter mixing conditions in the water column may enhance the germination of dinoflagellates cysts (**Figure 6C**). Depending on snow thickness combined with photoperiod, the sea-ice diatom bloom may be enhanced or reduced compared to that of harsh winter (Tedesco et al., 2017). In the case of favorable light conditions (e.g., snow-free or little snow on the ice cover), the sea-ice diatom bloom may be coupled to an under ice bloom of germinated dinoflagellates after the sea ice has completely melted. If the spring bloom would be mostly composed of P. catenata, this would be mostly degraded in the water column, becoming part of the phyto-detritus available to the macrozoobenthos, in particular to Marenzelleria. On oxic sediments and at high abundances, Marenzelleria may help keep the deep sediments oxygenated and the sinking organic matter remineralized without passing any of the sediment thresholds (**Figure 6C**). However, there might be situations when the sinking material is not efficiently consumed with the risk of seasonal hypoxia/anoxia. In the case of a spring bloom dominated by B. baltica, about 40–50% of the sinking POC would be composed of not easily degradable cysts (**Figure 6D**). On oxic sediments, the remaining part would be available as a source of food for the macrofauna. Of this non-degradable carbon pool, only 10–20% of the cysts may germinate in the following winter (Kremp and Anderson, 2000), thus 40–45% of this spring production would be buried in the sediment. This would represent an effective long-term C sink in the sediments (**Figure 6D**), which has the potential to improve the environmental status of the Baltic Sea.

Marine ecosystem models are currently essential tools for exploring the responses of ecosystems for management actions and for exploitation, but these models can only be useful if the ecological processes are correctly described and the essential components are included. In order to develop and improve this approach, understanding the transfer of nutrients between pelagic and benthic systems from winter through spring and

primary driver for the onset of the spring bloom is the availability of light, which is mainly governed by the mixing/stabilization of the upper parts of the water column. In the Baltic Sea large outflow of freshwater during spring creates patches of stratified water where the bloom initiates. After the spring bloom, diatoms tend to sink quickly to the seafloor where the biomass is remineralized. The sediment processes will depend on the amount of settling material and eutrophication have shifted the pristine state (A) towards a deteriorated environmental state with release of nutrients from the sediments (B). Dinoflagellates on the other hand tend to lyse before reaching the sediment (C) with the exception of resting cysts (D). These resting stages (some species of diatoms also produce large quantities of resting stages with the same function) are built for long term survival and are not remineralized in the sediment on a seasonal basis. Some of the resting stages are buried in the seafloor, providing a biological sink for organic nutrients. The size of this sink largely depends on the species composition of phytoplankton during the productive spring bloom.

into summer is needed. There are open questions related to the role of DOM in transfer of nutrients (quantity and quality), and the qualitative aspect of the sinking material, where changes in functional traits in the spring bloom community is important. The potential feedback mechanisms through changing plankton community composition have so far received relatively little attention, although community structure is known to play a major role for biogeochemical fluxes. The inherent difficulty in resolving causal connections from field observations calls for more dedicated experiments addressing the plankton community composition and the settling of phytoplankton in various forms under different environmental change scenarios.

Finally, as the Baltic Sea is facing increasing temperature associated with shorter ice season and thinner sea ice (Thomas et al., 2017), there are indications that climate change will weaken the magnitude, but prolong the duration of the spring bloom (Groetsch et al., 2016), which would directly affect the timing and flux of organic matter to the seafloor (Tamelander et al., 2017). Earlier stratification could benefit dinoflagellates, and the observed shift toward more dinoflagellate dominance in Gulf of Finland could continue and expand spatially. This would affect both the quantity and quality of the settling material and its mineralization at the seafloor, with less seabed enrichment in labile organic matter due to a more efficient pelagic microbial loop (**Figure 6C**). Dominance of species with high encystment rates and low germination success may enhance long-term burial of organic carbon as more resting stages are settling out of the water column (**Figure 6D**). This would functions as a biological sink and could be an underlying explanation for part of the unresolved N-burial (Jäntti and Hietanen, 2012), but this has not to our knowledge been tested before. Increasing

burial would diminish available nutrients in the ecosystem and would together with the realized reduction in nutrient loading counteract the observed spreading of the eutrophication. The increasing contribution of cyst-producing dinoflagellates could thus contribute to reversing eutrophication in the Baltic Sea.

#### AUTHOR CONTRIBUTIONS

The original idea for this manuscript came from a project involving KS, AK, and TiT. KO carried out data analysis on the Helsinki data (**Figures 2**, **3**). KO, AK, and RK formulated the phytoplankton species section and its development, and RK made **Figure 1** modified by KS. JL wrote the sediment process section and helped formulate the modified vicious cycle (**Figure 4**). LT, HP, KS, and TiT developed the conceptual model and LT made **Figure 5**. ToT reviewed and wrote about the trends in sedimentation. KS wrote the

#### REFERENCES


manuscript with contributions and/or comments from all co-authors.

#### FUNDING

This study was supported by the Academy of Finland (decisions no 259164 and 263376) and Estonian Research Council (no 1575P, KO). Further funding came from Walter and Andrée de Nottbeck Foundation (KS and ToT) and the Swedish Cultural Heritage (ToT). LT acknowledges support also from COCOA— Nutrient cocktail in coastal zones of the Baltic Sea—EU BONUS project.

#### ACKNOWLEDGMENTS

We are thankful to the City of Helsinki Environment Center (Finland) for providing the phytoplankton monitoring data.


northern Baltic. Mar. Ecol. Prog. Ser. 94, 239–252. doi: 10.3354/meps0 94239


of salinity and eutrophication. Mar. Ecol. Prog. Ser. 340, 121–138. doi: 10.3354/meps340121


**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 © 2018 Spilling, Olli, Lehtoranta, Kremp, Tedesco, Tamelander, Klais, Peltonen and Tamminen. 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(s) 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.

# Drivers of Cyanobacterial Blooms in a Hypertrophic Lagoon

Marco Bartoli1,2 \*, Mindaugas Zilius<sup>2</sup> , Mariano Bresciani<sup>3</sup> , Diana Vaiciute<sup>2</sup> , Irma Vybernaite-Lubiene<sup>2</sup> , Jolita Petkuviene<sup>2</sup> , Gianmarco Giordani1,2, Darius Daunys<sup>2</sup> , Tomas Ruginis<sup>2</sup> , Sara Benelli<sup>4</sup> , Claudia Giardino<sup>3</sup> , Paul A. Bukaveckas<sup>5</sup> , Petras Zemlys<sup>2</sup> , Evelina Griniene<sup>2</sup> , Zita R. Gasiunaite<sup>2</sup> , Jurate Lesutiene<sup>2</sup> , Renata Pilkaityte˙ <sup>2</sup> and Arturas Baziukas-Razinkovas<sup>2</sup>

<sup>1</sup> Department of Chemistry, Life Sciences and Environmental Sustainability, University of Parma, Parma, Italy, <sup>2</sup> Marine Research Institute, University of Klaipeda, Klaipeda, Lithuania, <sup>3</sup> Optical Remote Sensing Group, CNR-IREA, Milan, Italy, <sup>4</sup> Department of Life Sciences and Biotechnology, University of Ferrara, Ferrara, Italy, <sup>5</sup> Department of Biology and Center for Environmental Studies, Virginia Commonwealth University, Richmond, VA, United States

#### Edited by:

Riina Klais, University of Tartu, Estonia

#### Reviewed by:

Sai Elangovan S, National Institute of Oceanography (CSIR), India Franziska Wemheuer, University of New South Wales, Australia

> \*Correspondence: Marco Bartoli marco.bartoli@unipr.it

#### Specialty section:

This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science

Received: 13 May 2018 Accepted: 30 October 2018 Published: 15 November 2018

#### Citation:

Bartoli M, Zilius M, Bresciani M, Vaiciute D, Vybernaite-Lubiene I, Petkuviene J, Giordani G, Daunys D, Ruginis T, Benelli S, Giardino C, Bukaveckas PA, Zemlys P, Griniene E, Gasiunaite ZR, Lesutiene J, Pilkaityte R and ˙ Baziukas-Razinkovas A (2018) Drivers of Cyanobacterial Blooms in a Hypertrophic Lagoon. Front. Mar. Sci. 5:434. doi: 10.3389/fmars.2018.00434 The Curonian Lagoon is Europe's largest lagoon and one of the most seriously impacted by harmful blooms of cyanobacteria. Intensive studies over the past 20 years have allowed us to identify the major drivers determining the composition and spatial extent of hyperblooms in this system. We summarize and discuss the main outcomes of these studies and provide an updated, conceptual scheme of the multiple interactions between climatic and hydrologic factors, and their influence on internal and external processes that promote cyanobacterial blooms. Retrospective analysis of remote sensed images demonstrated the variability of blooms in terms of timing, extension and intensity, suggesting that they occur only under specific circumstances. Monthly analysis of nutrient loads and stoichiometry from the principal tributary (Nemunas River) revealed large interannual differences in the delivery of key elements, but summer months were always characterized by a strong dissolved inorganic N (and Si) limitation, that depresses diatoms and favors the dominance of cyanobacteria. Cyanobacteria blooms occurred during high water temperatures, long water residence time and low-wind conditions. The blooms induce transient (night-time) hypoxia, which stimulates the release of iron-bound P, producing a positive feedback for blooms of N-fixing cyanobacteria. Consumermediated nutrient recycling by dreissenid mussels, chironomid larvae, cyprinids and large bird colonies, may also affect P availability, but their role as drivers of cyanobacteria blooms is understudied.

Keywords: Curonian Lagoon, nitrogen, phosphorus, silica, fluxes, stoichiometry, remote sensing, cyanobacteria

# INTRODUCTION

Human activities impact biogeochemical cycles, biological communities and ecosystem functioning of inland and coastal waters on a global scale (Bernot and Dodds, 2005; Muhlolland et al., 2008; Paerl, 2009; Han and Allan, 2012). Estuaries and lagoons have become enriched with nutrients due to wastewater discharge, aquaculture, and agriculture (Galloway et al., 2008; Paerl, 2009). Excess nutrients result in blooms, where algal biomass accumulates and exceeds the mineralization capacity of the heterotrophic community (Valiela et al., 1997). In fresh-brackish waters, algal

blooms may include one or more types of harmful cyanobacteria, resulting in the presence of cyanotoxins. The development of hypoxic or anoxic conditions can lead to die-offs of fish and benthic organisms (Norkko and Bonsdorff, 1996; Ye et al., 2011). In addition, changes in food web structure brought about by invasive species may accelerate eutrophication by reducing grazing pressure and allowing the proliferation of algae, including toxic forms (Carpenter et al., 1998; Rabalais et al., 2002). The alteration of nutrient stoichiometry (sensu Redfield) and changes in climate (e.g., warmer temperatures, precipitation timing and intensity) have also received attention as potential drivers of harmful algal blooms (Cloern, 2001; Yunev et al., 2007; Moore et al., 2008; Howarth et al., 2011). In this review, we analyze the drivers of algal blooms in the Curonian Lagoon, a hypereutrophic freshwater estuary. We discuss the relevance of nutrient loads and their stoichiometry on algal blooms, we analyze how algal blooms affect the ecosystem functioning (e.g., nutrient mass balances) and provide a mechanistic interpretation for positive feedbacks promoting the dominance of cyanobacteria.

#### THE CURONIAN LAGOON: GENERAL FEATURES OF A HYPERTROPHIC FRESHWATER ESTUARY

The Curonian Lagoon is a large (surface area = 1500 km<sup>2</sup> ), shallow (mean depth = 3.5 m) waterbody located along the south-eastern portion of the Baltic Sea (**Figure 1**). The Curonian Spit (a UNESCO heritage site) divides the lagoon from the Baltic Sea. The main source of water and nutrients is the Nemunas River, although the lagoon also exchanges water with the Baltic Sea via the narrow Klaipeda Strait (Vybernaite-Lubiene et al., 2017). Exchange of water through the strait is episodic; during wind-driven forcing events, the salinity of the lagoon rises to ∼7. The principal tributary (Nemunas River) bisects the lagoon such that the northern lagoon is subject to greater fluvial (and marine) influence, whereas the southern portion of the lagoon is more lentic, and has a longer water residence time (Umgiesser et al., 2016). The lagoon has a relatively small hydrologic loading factor (ratio of watershed area to surface area), which makes this system similar to a flowthrough reactor, and provides an opportunity for mass balance studies (Bresciani et al., 2012; Zilius et al., 2014; Vybernaite-Lubiene et al., 2017). Prior work by our multidisciplinary and multinational team has included assessment of nutrient loads from the Nemunas watershed, application of hydraulic models to simulate water circulation in response to changing discharge and wind conditions, investigation of seasonal dynamics of biogeochemical cycles, and use of satellite remote sensing to monitor phytoplankton blooms (Vaici ˇ ute et al., 2015 ˙ ; Petkuviene et al., 2016; Umgiesser et al., 2016; Vybernaite-Lubiene et al., 2017, 2018). Our work at this site has also benefitted from longterm monitoring carried out by the Marine Research Department of the Lithuanian Ministry of Environment.

Data arising from these efforts have helped to guide management of the Curonian Lagoon via a number of national and international programs (HELCOM Baltic Sea Action Plan, European Water Framework Directive, various habitat and bird conservation initiatives). Despite the intensive studies carried out in the lagoon, there remains the question whether and how cyanobacteria blooms can be mitigated. These blooms extend over large areas of the lagoon and negatively impact ecosystem functions, including tourism and recreational activities, as well as local fisheries (Giardino et al., 2010; Belykh et al., 2013; Šulcius ˇ et al., 2015). The use of science for informing management decisions is dependent upon the interpretation and integration of available data, which is the focus of this paper.

## SEASONAL SUCCESSION OF PHYTOPLANKTON

Detailed studies of plankton communities in the Curonian Lagoon have examined seasonal patterns, species interactions, production of cyanotoxins and the role of phytoplankton in food web energetics (e.g., Pilkaityte and Razinkovas, 2006 ˙ , Razinkovas, 2007; Lesutiene et al., 2014 ˙ ; Bukaveckas et al., 2017). Diatoms dominate the spring phytoplankton community, after which, following a short clear-water phase, cyanobacteria biomass increases (Gasiunaite et al., 2005 ˙ ; Pilkaityte and Razinkovas, ˙ 2007). Fresh-brackish species dominate the phytoplankton community of the Curonian Lagoon. Stephanodiscus hantzschii, Diatoma tenuis, Aulacoseira islandica, Asterionella formosa are the dominant diatom species during spring while the N-fixing cyanobacteria Aphanizomenon flosaquae, Dolichospermum affine, D. flosaquae, as well as other cyanobacteria such as Microcystis aeruginosa, M. wesenbergii, M. viridis, and Planktothrix agardhii contribute to the summer biomass peak (Pilkaityte and ˙ Razinkovas, 2007; Gasiunaite et al., 2008 ˙ ). According to longterm monitoring data (2001–2012), monthly average chlorophyll a (chl-a) concentrations reach 47 ± 14 mg m−<sup>3</sup> during the spring diatom bloom and 96 ± 56 mg m−<sup>3</sup> during the summer bloom (Marine Research Department of the Lithuanian Ministry of Environment).

### WIND EFFECTS ON ALGAL BLOOM DEVELOPMENT

Algal blooms in the Curonian Lagoon have been tracked since the 1930's via synoptic sampling (Schmidt-Ries, 1940). More recently, satellite remote sensing has substantially improved our ability to track the spatial and temporal dynamics of bloom events and draw links to local weather conditions. The first attempt to map algal blooms in the Curonian Lagoon utilized the MEdium Resolution Imaging Spectrometer (MERIS) on board the Envisat-1 satellite (Giardino et al., 2010; Bresciani et al., 2012; Vaici ˇ ute et al., 2015 ˙ ). The combination of high spatial resolution (300 m) and short revisit time (2–3 days) greatly enhanced our ability to map chl-a. More recently, the Operational Land Imager (OLI, on board Landsat-8) and Multispectral Instrument (MSI, on board Sentinel-2A/B) have further enhanced spatial resolution (10–30 m) and allowed us

to investigate the patchy distribution of cyanobacteria blooms (INFORM, 2017).

Lagoon during a summer cyanobacteria bloom.

Results based on a large number of images from 2004 to 2016 revealed temporal variability and small-scale spatial patchiness of chl-a (Bresciani et al., 2012, 2014; Vaici ˇ ute et al., 2015 ˙ ). The southern part of the lagoon exhibited high chl-a (up to 500 mg m−<sup>3</sup> ) while the northern areas were characterized by lower values (∼50 mg m−<sup>3</sup> ) (Bresciani et al., 2012; **Figure 2**). Differences between the northern and southern portions of the lagoon were documented by earlier studies (Olenina, 1998; Krevs et al., 2007). However, the use of satellite images allowed us to identify hot spots of chl-a (up to 400 mg m−<sup>3</sup> ) and the presence of surface scums (e.g., Bresciani et al., 2014; **Figure 2**). Highest concentrations were coincident with prevailing wind conditions, suggesting that wind speed and direction was a significant driver for spatial distribution of positively buoyant cyanobacteria (Bresciani et al., 2014).

Wind speed affects not only the spatial distribution of cyanobacteria, but also influences water column mixing. Wind speeds less than 2 m s−<sup>1</sup> are common and allow for the development of transient (daytime) thermal gradients within the water column (Zilius et al., 2014). The lagoon, though shallow, is relatively turbid and it is thought that stagnant conditions associated with low wind allow positively buoyant cyanobacteria to obtain favorable, near-surface light conditions. The combination of remote sensing, in situ biogeochemical studies, and local meteorological data allowed us to investigate these linkages over large spatial scales. Measurements of benthic and pelagic oxygen metabolism along with spatial patterns of MERIS-derived chl-a showed that 60–95% of the area of the lagoon was vulnerable to transient hypoxia when blooms coincide with calm conditions (Zilius et al., 2014).

# HYDRODYNAMIC FACTORS AS DRIVERS OF BLOOMS

Freshwater inputs to the lagoon are dominated by the Nemunas River, which has an annual average discharge of 21.8 km<sup>3</sup> , and accounts for 96% of total inputs (Jakimavicius and Kriau ˇ ci ˇ uniene, ˙ 2013). The Nemunas River discharges into the central part of lagoon, dividing the system in a northern and a southern region that differ in water renewal time (Umgiesser et al., 2016). The northern part of the lagoon is characterized by strong riverine influence and short renewal time ( < 80 days), which result in limited accumulation of suspended matter (Ferrarin et al., 2008; Remeikaite-Nikien ˙ e et al., 2016 ˙ ). The southern part of the lagoon has a longer water residence time ( > 190 days) with minimal fluvial influence. The latest efforts to analyze the water exchange within the Curonian Lagoon (Umgiesser et al., 2016) revealed different seasonal patterns of residence time, primarily driven by changes in hydrographic forcing by the Nemunas River. During elevated spring discharge, the entire lagoon is strongly flushed by Nemunas River. During summer, river discharge decreases, resulting in increased water residence time, particularly in the southern lagoon. Summer stagnation has implications for water temperature, stratification, nutrient availability and stoichiometry, and phytoplankton abundance and composition. In summer, wind forcing appears to be the most important factor influencing water column mixing and

exchange between the southern and the northern part of the lagoon (Umgiesser et al., 2016).

#### NUTRIENT LOADS AND THEIR ECOLOGICAL STOICHIOMETRY

A study coupling the Curonian Lagoon with its watershed was started in 2012 to better characterize the timing of nutrient inputs and their stoichiometry. From 2012, on at least a monthly basis and more frequent (weekly) during high discharge periods, discharge and water chemistry (including all dissolved and particulate forms of N, Si, and P ) were monitored near the inflow of the Nemunas River to the lagoon (Vybernaite-Lubiene et al., 2017; **Figure 1**). Discharge and nutrient concentrations displayed strong seasonality. Nitrate and reactive Si concentrations decreased by two orders of magnitude from spring to summer (e.g., NO<sup>3</sup> <sup>−</sup> from > 300 µM to < 1, SiO<sup>2</sup> from > 200 to < 1 µM) while reactive P concentrations showed comparatively smaller changes (from 0.2 to 4 µM). These seasonal patterns resulted in reduced DIN:DSi and DIN:DIP ratios, which shift the lagoon from an excess of N and Si in colder months, to P excess (DIN:DIP < 16) in warmer months (**Figure 3**). These findings support the hypothesis that cyanobacterial blooms are favored during summer by the limited

N and Si supply via riverine inputs (Pilkaityte and Razinkovas, ˙ 2007).

Our analyses of nutrient loads showed that recent (2012– 2016) N export from the Nemunas River basin is similar to historical data (1986–2002), whereas P loads have declined by

nearly 60 % as a result of sewage treatment plant improvements (Vybernaite-Lubiene et al., 2018). Despite reductions in P loads, the lagoon remains imbalanced with an excess of P relative to N, thereby favoring the growth of N-fixing cyanobacteria. Further P reductions are needed to promote- limitation or co- limitation and thereby diminish the dominance of cyanobacteria. It is also important to stress that despite strong N limitation in the lagoon, hyperblooms of cyanobacteria do not occur every summer due to the influence of other factors that regulate blooms.

#### SEDIMENTS AND BENTHIC PROCESSES AS DRIVERS OF BLOOMS

The distribution of sediment types in the Curonian Lagoon is determined by hydrodynamic factors and by contributions from autochthonous and allochthonous materials (Pustelnikovas, 1994; Ferrarin et al., 2008). Curonian Lagoon sediments include a broad spectrum of deposits, from sand-dominated in the northern (riverine-influenced) sector to silt-dominated in the southern (more lentic) area (Trimonis et al., 2003). Declines in external (riverine) loads during the transition from spring to summer enhances the importance of internal recycling from the benthic compartment as a nutrient source for pelagic primary production. Studies of sedimentary processes revealed a shift in dominant microbial processes and benthic fluxes from spring to summer (Zilius et al., 2012, 2014; Petkuviene et al., 2016). For example, net N<sup>2</sup> production suggests the dominance of denitrification over N-fixation during spring; however N<sup>2</sup> fluxes are reversed during summer, suggesting net N import to the benthic compartment (Zilius et al., 2018).

Since 2009, oxygen penetration depth, total and diffusive sedimentary oxygen demand, pore water chemical environments, sedimentary pools and benthic fluxes were measured or calculated at sites representative of dominant sedimentary environments including littoral, pelagic transitional and confined zones (Zilius et al., 2012; **Figure 1**). In this turbid system, benthic photosynthesis was measurable only in shallow littoral illuminated sediments (∼1 m depth) representing a minor fraction (5%) of the total lagoon surface (Benelli et al., 2018). Here, benthic algae oxidize the upper sediment layer and efficiently retain nutrients, thereby impeding regeneration to the water column (Zilius et al., 2012; Benelli et al., 2018). Deeper sites were always heterotrophic and their seasonal oxygen metabolism and nutrient regeneration was driven by water temperature and phytoplankton blooms; recently settled fresh phytoplankton resulted in significantly higher oxygen uptake, limited oxygen penetration in sediments ( < 1 mm), and high rates of anaerobic to aerobic metabolism (Zilius et al., 2012, 2016).

The mechanisms underlying P release from sediments were analyzed in detail, as they contribute to lower inorganic DIN:DIP ratio in the water column and favor cyanobacteria (Zilius et al., 2014, 2015, 2016; Petkuviene et al., 2016). The distribution of sedimentary pools of P, Fe, Mn and S in the Curonian Lagoon was related to riverine influence; sandy sediments adjacent the Nemunas delta were oxidized and have a large geochemical buffer capacity against the effects of anoxia, with limited accumulation of free sulfide (Petkuviene et al., 2016). Muddy areas along the western and southern portion of the lagoon had chemically reduced sediments where the reduction of iron may result in large P release (Petkuviene et al., 2016). In manipulative experimental studies, simulated deposition of phytoplankton material, primarily composed of cyanobacteria, resulted in an increase of dissimilative nitrate reduction over denitrification and large methane production, but with limited reactive P release (Zilius et al., 2016). Experimental manipulations of intact cores, targeting short-term effects of anoxia, revealed that pools of detritial Ca bound P ( > 70 % of inorganic P) and oxidized Fe and Mn, prevent or buffer redox-dependent reactive P release from sediments (Zilius et al., 2015). However, seasonal measurements of reactive P fluxes at sandy and muddy areas revealed large summer P release at muddy sites coinciding with the occurrence of cyanobacterial blooms, and the onset of hypoxia and anoxia in the water column (Zilius et al., 2014; Petkuviene et al., 2016). These events occurred under specific conditions during prolonged stable weather, with no wind and high water temperature. Benthic P release occurred when the oxidized pools of metals within sediments were exhausted and contributed to the imbalanced stoichiometry by further lowering the DIN:DIP ratio. P regeneration from sediment, despite occurring over short period, had a significant effect on the lagoon P budget, resulting in a large export of P (Petkuviene et al., 2016).

## TOP-DOWN DRIVERS OF BLOOMS

#### Zooplankton

The shift from diatom- to cyanobacteria- dominated phytoplankton communities was accompanied by a decline in relative zooplankton grazing. During the spring diatom bloom, maximum consumption by zooplankton corresponded to 34% of NPP (324 µgCL−1d −1 ), whereas during the summer cyanobacteria bloom grazing decreased to 8 % of NNP (470 µgCL−1d −1 ) (**Figure 4**). A similar pattern was observed in the southern part of the lagoon where zooplankton grazing declined from 60% of phytoplankton production during spring to 4 % in summer (Semenova and Aleksandrov, 2009). Despite reduced grazing rates, stable isotope studies show that cyanobacteria blooms support secondary production in a diverse group of benthic and pelagic consumers within the lagoon (Lesutiene et al., 2014 ˙ ). Our studies also show that cyanotoxins (microcystin) are found in tissues of fish and shellfish, indicating that cyanobacteria production supports higher trophic levels in this system (Bukaveckas et al., 2017).

Grazing by zooplankton may be an important driver of cyanobacteria bloom development. During winter, ciliate growth is limited by low biomass of phytoplankton. In the early spring, when small-sized phytoplankton are dominant, the ciliate assemblage was dominated by small naked oligotrichs and prostomatids. After the late spring diatom bloom, the ciliate assemblage shifted to medium sized tintinnids, which feed on the same nano-fraction of phytoplankton or/and heterotrophic flagellates as ciliates. The summer/autumn phase was characterized by increased taxonomical and functional

diversity of ciliates indicating exploitation of a wide size range of food. Small sized naked oligotrichs (Strobilidium spp.) and peritrichs (Vorticella spp.) (mainly bacterivorous ciliates) dominated in summer, indicating a shift from algal food to bacteria (Griniene, 2013; Grinien ˙ e et al., 2016 ˙ ). The shift from large Daphnia to small-bodied Chydorus sphaericus coincides with the dominance of cyanobacteria (Gasiunaite and ˙ Razinkovas, 2004; **Figure 5**). Chydorus graze on smaller algae and therefore give an advantage to large cyanobacteria (Gasiunaite˙ and Olenina, 1997). In addition, the presence of large filamentous colonies and toxic strains may foster the dominance of bloom forming cyanobacteria (Pilkaityte and Razinkovas, 2007 ˙ ).

#### Macrofauna

Excluding the littoral zone, sediments of the Curonian Lagoon host few macrofauna species due to high organic content and poor oxygen conditions (Zettler and Daunys, 2007). Among them, oligochaetes, chironomid larvae and freshwater mussels, including native unionids and invasive dreissenids, are dominant groups (Daunys, 2001). Chironomid larvae and mussel aggregations may, due to their high densities, influence phytoplankton composition and abundance (Dame et al., 1980; Officer et al., 1982; Gili and Coma, 1998). We discuss here if and under which circumstances macrofauna may favor the onset of cyanobacterial blooms in the Curonian Lagoon.

In the lagoon, periods of short water residence time may impede efficient removal of particulate matter by suspension feeding. In spring (average residence time 7 days) only 10 % of particulate matter was removed by zebra mussels, while in summer (average residence time up to 15 days), the proportion of particulate matter removed increased to 30% (Daunys et al., 2006). Chironomid larvae and mussels may exert a top-down control of pelagic primary production but

they may simultaneously excrete large amount of nutrients. Their activities also enhance the organic matter content of sediment via biodeposition, stimulating microbial activity and re-mineralization (Caraco et al., 1997; Stief, 2013; Ruginis et al., 2014; Benelli et al., 2017). It is unclear whether the net effect of phytoplankton removal via grazing is offset by nutrient regeneration via excretion and whether these processes have a specific benefit to cyanobacteria. Dreissena polymorpha was intensively studied due to its top-down control on phytoplankton and the possible management of its biomass to reverse eutrophication. However, such top-down control on pelagic primary production resulted to be site-specific and context-dependent (e.g., in shallow, well-mixed environments with low nutrient background more than in deep, stratified ecosystems with high nutrient inputs) (Conroy et al., 2005; Caraco et al., 2006). Furthermore, dreissenids excrete large amounts of reactive P and different authors have suggested that these mussels may change nutrient stoichiometry, via P mobilization and by enhancing N removal via denitrification (Zhang et al., 2008; Ruginis et al., 2014). The inability of zebra mussels to graze on larger forms of cyanobacteria may provide a competitive advantage over other algae, which, in combination with increased rates of reactive P re-cycling, enhances the potential for cyanobacteria blooms. These aspects need further study, but suggest that the presence of dreissenids mussels on the Curonian Lagoon may exacerbate the effects of nutrient loading, and favor increased dominance by cyanobacteria.

#### Birds

The Curonian Lagoon hosts a large bird community, including tufted ducks and common pochards with 24,500–54,700 and 1,800–41,000 individuals, respectively (Stanevicius et al., 2009 ˇ ),

goosanders (Žydelis, 2001), cormorants, with more than 10,000 breading birds (Švažas et al., 2011; Dagys and Zarankaitë, 2013), mallards, geese (3,000–6,500 ind/day) and little and blackheaded gulls (1,000–1,500 ind./day). High densities of water birds are vectors of seeds, invertebrates, bacteria and phytoplankton (Tobiessen and Wheat, 2000), and also contribute to nutrient loads (Manny et al., 1994; Hahn et al., 2007; Green and Elmberg, 2014; Han et al., 2017).

During nesting, breeding and roosting periods, water birds enrich the water with guano (Klimaszyk et al., 2015). In enclosed aquatic ecosystems bird feces may contribute 50–69%, 27– 40%, and 70–75% of total C, N, and P loads, respectively (Manny et al., 1994; Post et al., 1998; Boros et al., 2008; Gwiazda et al., 2014). Bird feces have low N:P, implying that water bird excretions may strengthen N limitation and promote cyanobacteria blooms (Rönicke et al., 2008; Han et al., 2017). Birds also have indirect effects on nutrient cycling by removing macrophytes, invertebrates and fish. Herbivorous birds, by removing plants, remove those elements that trap nutrients in the benthic compartment, provide shelter for zooplankton and allow sediment oxidation via radial oxygen loss. While grazing on macrophytes, birds resuspend sediments and mobilize pore water nutrients (Klimaszyk et al., 2015; Klimaszyk and Rzymski, 2016). Furthermore, a large fraction of macrophyte-associated P is released to the water column in the reactive form and is readily available to phytoplankton. The mechanisms that regulate P mobility in sediments are redox-dependent such that the removal of roots and macrofauna, together with particle resuspension, has the potential to mobilize sediment sources of P. In the Curonian Lagoon herbivorous birds represent the second largest water bird group, peaking in spring and distributed throughout the Nemunas River deltaic area and the littoral zone.

Benthivorous birds, feeding on macrofauna, produce an effect on the benthic system similar to that of fishes, removing animals that may keep the sediment oxidized and resuspending sediments and nutrients (Werner et al., 2005; Sánchez et al., 2006; Rodríguez-Pérez et al., 2007; Matuszak et al., 2014). Piscivorous birds convert fish-associated P into reactive P (Putys and Zarankaite, 2010 ˙ ). Large colonies of cormorants have their peak activity during summer. The large bird community in the Curonian Lagoon may therefore affect by various direct and indirect mechanisms the cycling of nutrients and that of P in particular.

#### Fish

Benthivorous fish, including carp, roach, bream and perch, represent the dominant fish component in the Curonian Lagoon (Cline et al., 1994; Persson and Svensson, 2006; Lithuanian Environmental Protection Agency (EPA), 2008; Adámek and Maršálek, 2013). Fish may produce both top-down (e.g., removal of grazers and competitors) and bottom-up effects (nutrient mobilization) that favor eutrophic conditions and cyanobacterial blooms (Shormann and Cotner, 1997; Roozen et al., 2007). The diet of benthivorous fish in the Curonian Lagoon includes mussels, chironomidae larvae, detritus, zooplankton and plants (Bubinas and Ložys, 2000). Benthivorous fish may impact the water quality, leading to nutrient accumulation and algal growth, by suspending the sediments and by feeding on filterfeeding zooplankton, burrowing macrofauna and macrophytes (Zambrano and Hinojosa, 1999; Williams et al., 2002; Parkos et al., 2003). Sediment resuspension by the benthic fish community increases water turbidity, limits light penetration and rooted macrophytes and favors P release and cyanobacteria growth (Hellström, 1991; Breukelaar et al., 1994). By removing invertebrates from the sediments, benthivorous fishes mobilize nutrients from the pore water (Tarvainen et al., 2002; Phan-Van et al., 2008). Resuspension itself may oxidize sediments, but this is a short-term and local effect, while reductions in invertebrate abundance impacts N removal via denitrification and P sequestration. Moreover, fish predation reduces zooplankton populations, resulting in low grazing on phytoplankton (Jeppesen et al., 1999). Fish excretions are very soluble and rich in N and P which stimulate periphyton growth and negatively affect macrophytes (Tarvainen et al., 2002; Williams et al., 2002). Excreted nutrients are dispersed horizontally and vertically and from littoral to pelagic areas (Schindler et al., 1996; Persson and Svensson, 2006).

## SYNTHESIS

Cyanobacterial blooms in the Curonian Lagoon arise from multiple interacting factors, which include external forcing (riverine discharge and wind conditions) and internal processes (consumer-mediated nutrient cycling and sediment-water nutrient exchange). We summarize the information discussed in this review through a graphical representation of the multiple mechanisms that drive cyanobacterial blooms in the Curonian Lagoon (**Figure 6**).

During spring, the lagoon is diatom-dominated due to a combination of low water temperatures, high river discharge and availability of inorganic N and Si, in excess to P. The system alternates phases with clear and turbid water depending on the intensity of the spring diatom bloom and the occurrence of wind-associated sediment resuspension events. During spring, light penetration may attain 1–2 m and grazers exert appreciable control of algal biomass accrual. The water column is generally well mixed and normoxic; under these circumstances the upper sediment layer is oxidized and acts as a nutrient sink.

The spring-summer transition is marked by a decline in discharge of the Nemunas River, which is accompanied by the depletion of N and Si within the lagoon. Reductions in external loadings, together with processes occurring within the lagoon (high spring denitrification rates and Si sequestration via uptake and accumulation in sediment), result in the onset of N and Si limitation. Cyanobacteria become dominant, resulting in a series of cascade effects that include increased algal-associated turbidity and water stratification. Positive feedbacks arise as large colonies of cyanobacteria limit the capacity of grazers to control biomass accrual and high respiration rates promote oxygen undersaturation. Climatic conditions, which are highly variable, play an important role, as low wind conditions may further push the system toward hyperbloom events, with extensive surface scums (**Figure 7**). Water column respiration, not sediment oxygen demand, promote oxygen depletion in the system, due to the large availability of labile organic matter from decaying algal cells. By this mechanism, hyperblooms promote their persistence as hypoxia results in sediment P release. Other factors that may contribute to these large periodic outbreaks include the presence of waterbirds. Large colonies of cormorants settle in spring along the Curonian Spit and have their most intense period of activity during summer. Cormorants, through the production of guano, make large amounts of P available in surface water. Besides cormorants, colonies of seagulls, swans and duck have large numbers and may contribute to make fish or macrophyte P pools readily available to cyanobacteria. Other biological agents supporting algal blooms include the invasive freshwater mussel (D. polymorpha), which excretes large amounts of reactive P relative to the native unionid mussels and may therefore contribute to the low DIN:DIP ratio in the lagoon.

FIGURE 7 | Cyanobacteria hyperbloom with scum formation in the Curonian Lagoon, August 2013.

# MANAGEMENT POLICIES AND PERSPECTIVES

fmars-05-00434 November 13, 2018 Time: 19:23 # 9

The Curonian Lagoon ecosystem provides a number of provisioning and cultural ecosystem services, most directly linked to the main economic activities in the lagoon area – recreation and fishery (Rashleigh et al., 2012; Razinkovas-Baziukas et al., 2012). The lagoon also provides ecosystem services of relevance to the Baltic Sea region such as denitrification and phosphorus burial. Management efforts to improve water quality in the Curonian Lagoon have targeted reductions of nitrogen loads by 15 % and phosphorus by 8%. Model simulations (Ertürk et al., 2016) revealed that reductions of nitrogen loads by 14 % and phosphorus loads by 6%, will bring about a 10% reduction in the abundance of cyanobacteria (Razinkovas et al., 2008). Further reductions in riverine nutrient loads (40% decrease in both N and P) produced only a 10% decrease in peak chl-a concentrations (Razinkovas et al., 2008). Further efforts to improve water quality may require withinsystem bioengineering solutions (biomanipulations, mussels, reed harvesting).

Climate change projections for the Curonian Lagoon (Jakimavicius et al., 2018 ˇ ) indicate an increase of average water temperature up to 1.7–2◦C by the middle of this century, consistent with trends observed during the last three decades (Dailidienë et al., 2011). According to this modeling study, the increase in water temperature was mostly confined to the summer-early autumn period, which may therefore favor the development of cyanobacteria blooms. Biogeochemical cycles of the Curonian Lagoon will be affected by changes in the water balance of the lagoon. A decline in contributions from Nemunas River coupled with an increase in Baltic water intrusions (due to sea level rise) will alter the water balance during the winter–spring period. A shift in the timing of peak discharge from spring to winter, as observed in recent decades (Dailidienë et al., 2011), may diminish algal blooms if a larger proportion of the nitrogen load from the Nemunas River passes through the Curonian Lagoon during the period of low phytoplankton productivity. However, the predicted decrease of ice cover is expected to reduce winter hypoxia, which would result in reductions in denitrification. Despite the intensive studies of this system, we are not able to predict whether climate change will exacerbate or mitigate cyanobacteria blooms. However, it is apparent that further management actions are needed to reduce nutrient loads and restore ecosystem services. There is a need for additional studies, both at the watershed-scale and the lagoon scale, to facilitate science-based management decision. At the watershed scale, long-term monitoring is needed to better understand the effectiveness of improved agricultural practices and water treatment on N, Si, and P export from the Nemunas basin to the Curonian Lagoon. Watershed practices may differentially affect the three elements further modifying their ecological stoichiometry, with implication for algal blooms

(Yunev et al., 2007; Bresciani et al., 2014; Vybernaite-Lubiene et al., 2017).

# CONCLUSION

This review analyzes the available information on the mechanisms driving cyanobacterial blooms in the Curonian Lagoon. Results from our analysis suggest that blooms are a consequence of multiple, interplaying factors, producing a cascade of processes and positive feedbacks. The hot moment for cyanobacteria blooms is the summer, due to combination of favorable nutrient stoichiometry (N and Si limitation), elevated water temperature, low wind speed, unbalanced internal recycling (P > N) and low grazing pressure. The hot spots of cyanobacteria are stagnant areas where limited water circulation and stratification provide these organisms a competitive advantage. These hot spots may serve as bloom initiation areas from which cyanobacteria are dispersed by prevailing winds. Ecological interactions among aquatic organisms, and how these respond to changes in climate and to species invasions remain understudied. The combination of satellite remote sensing, traditional monitoring of environmental parameters, detailed analysis of processes at the macro and microscale and the application of ecological network models, have proved to be useful tools for understanding the mechanisms underlying the development of cyanobacteria blooms. Our further efforts seek to improve our capacity to predict the occurrence and severity of algal blooms and guide prevention measures.

#### AUTHOR CONTRIBUTIONS

MBa collected and homogenized the material with the help of MZ, who realized the conceptual scheme of **Figure 6** and PB, who organized the different sections and improved the clarity of the paper. All authors contributed to the writing.

# FUNDING

MBr, JP, and DV were supported by Research Council of Lithuania (LMT) grant "Phosphorus as driver of cyanobacterial hyperblooms in the Curonian Lagoon (Patchy)" (Agreement No. S-MIP-17-11), MZ was supported by LMT grant "The role of atmospheric nitrogen fixation in the largest eutroficated European lagoon (NitFix)" (Agreement No. P-MIP-17-126), IV-L and AR-B were supported by the BONUS project "Nutrient Cocktails in Coastal zones of the Baltic Sea (COCOA)" (No. BONUS-2/2014). This work was partly funded by the European Community's Seventh Framework Program (FP7/2007-2013) under Grant 606865 (INFORM) and by the EOMORES project belonging to the European Union's Horizon 2020 research and innovation Program (grant agreement n◦ 730066). Participation in this research effort by PB was aided by a Fulbright award.

#### REFERENCES

fmars-05-00434 November 13, 2018 Time: 19:23 # 10


T. Bregnballe, J. Lynch, R. Parz-Gollner, L. Marion, S. Volponi, J.-Y. Paquet, et al. (Aarhus: Aarhus University), 69–71.


estuaries and coastal marine ecosystems. Front. Ecol. Environ. 9, 18–26. doi: 10.1890/100008


by cyanobacterial blooms. Estuaries Coast. 39, 1386–1402. doi: 10.1007/s12237- 016-0078-0



**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 © 2018 Bartoli, Zilius, Bresciani, Vaiciute, Vybernaite-Lubiene, Petkuviene, Giordani, Daunys, Ruginis, Benelli, Giardino, Bukaveckas, Zemlys, Griniene, Gasiunaite, Lesutiene, Pilkaityte and Baziukas-Razinkovas. 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(s) 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.

# Phytoplankton Community Dynamic: A Driver for Ciliate Trophic Strategies

Lumi Haraguchi <sup>1</sup> \*, Hans H. Jakobsen<sup>1</sup> , Nina Lundholm<sup>2</sup> and Jacob Carstensen<sup>1</sup>

<sup>1</sup> Department of Bioscience, Aarhus University, Roskilde, Denmark, <sup>2</sup> National History Museum of Denmark, University of Copenhagen, Copenhagen, Denmark

Phytoplankton plays a key role as primary producers and mediating biogeochemical

Edited by:

Riina Klais, University of Tartu, Estonia

#### Reviewed by:

Urania Christaki, Université du Littoral Côte d'Opale, France Sai Elangovan S, National Institute of Oceanography (CSIR), India

> \*Correspondence: Lumi Haraguchi luh@bios.au.dk

#### Specialty section:

This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science

Received: 27 April 2018 Accepted: 18 July 2018 Published: 07 August 2018

#### Citation:

Haraguchi L, Jakobsen HH, Lundholm N and Carstensen J (2018) Phytoplankton Community Dynamic: A Driver for Ciliate Trophic Strategies. Front. Mar. Sci. 5:272. doi: 10.3389/fmars.2018.00272 cycles in the water column. The understanding of the temporal dynamic of primary grazers channeling energy and carbon from primary producers is important for evaluating aquatic ecosystems functioning. This study investigates the coupling between phytoplankton and ciliates from live samples collected with approximately daily frequency during an almost 2-year cycle. The study site is a nutrient-rich temperate estuary, Roskilde Fjord (Denmark). Our aim is to evaluate the importance of protist grazers, especially ciliates, as predators on phytoplankton and to evaluate differences among multiple nutritional strategies through different seasons. The phytoplankton community, was mostly dominated by small organisms (<20µm) with few observations of diatoms. In most of observations, heterotrophic dinoflagellates biomass was smaller than biomass of ciliates (<10%), indicating that ciliates are the main component of microzooplankton. Except for the spring 2016, the ciliate community closely followed the phytoplankton community, showing a tight coupling between the primary producers and grazers during all seasons. This somehow contradicts the general assumption that ciliate dominance is restricted to periods of nutrient limitation dominated by the microbial food web and suggests a year-round key role of ciliates as consumers of phytoplankton biomass. Biomasses of ciliates increased during spring and were highest during summer. Relative importance of mixotrophs were high due to occurrence of Mesodinium rubrum blooms as well as other mixotrophic ciliates in late spring/early summer. M. rubrum biomass had the opposite pattern of the cryptophyte prey Teleaulax spp., and the coupling between the two populations was very strong in late spring. Ciliates that grazed on selected phytoplankton, had a smaller potential grazing impact regarding their biomasses, likely due to food limitation; conversely ciliates that feed on diverse prey items were less constrained by food limitation, and their seasonality appear to be driven by other factors. These findings suggest that the ciliate community structure and dynamics is important in structuring the phytoplankton community on short and seasonal scale.

Keywords: phytoplankton community, ciliates, grazing rates, mixotrophy, trophic strategies

# INTRODUCTION

Phytoplankton primary production supports higher trophic levels and fuels microbial remineralization (Azam et al., 1983; Sherr and Sherr, 1988). The dominant pelagic grazers of phytoplankton are typically associated with distinct operating modes of the food web compartments and nutrient cycling. Heterotrophic protist grazers and microzooplankton dominance is usually associated with the microbial loop and regenerated production; while mesozooplankton is associated with a linear food chain and export production (Fenchel, 1988; Buitenhuis et al., 2006). Grazing on particulate primary production in the global ocean surface is ∼10–15% for mesozooplankton and 59– 75% for microzooplankton (Behrenfeld and Falkowski, 1997; Calbet, 2001; Landry and Calbet, 2004; Buitenhuis et al., 2010), with estimates for coastal and estuarine systems usually in the a lower range (Landry and Calbet, 2004).

Ciliates constitute an important component of the microzooplankton community with preference for smallsized preys, in contrast to mesozooplankton, and many ciliate species are also grazed by mesozooplankton (Hansen et al., 1997). Thus, ciliates can be an important link between small cells and higher trophic levels (Nielsen and Kiørboe, 1994). Besides their significant role in carbon transfer, ciliates are also considered high quality food, as a source of proteinaceous compounds with a low C:N ratio in comparison to phytoplankton (Stoecker and Capuzzo, 1990; Gifford, 1991).

Although many ciliates are heterotrophs, a number of pelagic species are mixotrophic, combining both phagotrophic and phototrophic nutrition (Stoecker, 1998). The recognition of mixotrophy in the marine plankton food web has challenged the classical understanding of pelagic food webs, as autotrophy and heterotrophy are not necessarily two distinct functional compartments (Flynn et al., 2013). Classical understanding of ecological interactions among plankton, such as competition for nutrients, indicates that nutrient uptake affinity decreases with organism size (Edwards et al., 2012), favoring smaller sizes under resource limiting conditions. Mixotrophy is advantageous to organisms under nutrient limited conditions, allowing them to reduce direct competition by grazing on smaller prey and increase direct ingestion of nutrients (Mitra et al., 2014). Modeling results suggest that mixotrophy favors larger organisms, and therefore enhances trophic transfer efficiency (Mitra et al., 2014; Ward and Follows, 2016). On top of that, mixotrophy appears to be important over both, space and time, in marine systems (Leles et al., 2017), stressing the need for ecological field studies to further elucidate the role of mixotrophy.

Today, the importance of ciliates in the marine environment, including coastal and estuarine systems, is well recognized (Calbet and Landry, 2004). However, the role of ciliate nutrition mode and its impacts on ecosystem productivity is understudied for a number of reasons. One is that most plankton monitoring programs focus on analyzing phytoplankton and mesozooplankton only and similarly, many field studies do not include analyses of ciliates. Second, fixation can destroy cells and change their characteristics, such as color, size, and shape (Choi and Stoecker, 1989; Stoecker et al., 1994), constraining the distinction between mixotrophs and heterotrophs. In addition, even if ciliates are properly recorded, many studies employ monthly sampling, which hampers the investigation of ecosystem trophodynamics due to the fast growth responses of ciliates and phytoplankton.

The use of in-flow systems (e.g., flow cytometry) have routinely been used to assess plankton communities, including different size fractions (Dashkova et al., 2017). Using these technologies in high frequency monitoring of plankton has demonstrated that short-term events can be easily missed with sampling frequencies typically employed for monitoring (Thyssen et al., 2008; Campbell et al., 2013; Dugenne et al., 2014). Furthermore, in-flow systems allow analysis of live samples, avoiding loss and shrinkage of cells due to fixation (Jakobsen and Carstensen, 2011; Haraguchi et al., 2017). Thus, the use of in-flow systems can improve our knowledge on the coupled dynamics of phytoplankton and ciliates, by allowing a large number of samples to be analyzed in relatively short time.

This study aims to assess the temporal coupling between phytoplankton and its protist grazers in a temperate mesohaline estuary (Roskilde Fjord, Denmark), evaluating differences in potential grazing rates of distinct trophic strategies over different time scales. More specifically, this study seeks to answer the following questions: (1) Are ciliates the dominant pelagic grazers in Roskilde Fjord? (2) Do mixotrophic ciliates comprise a significant proportion of the total ciliate biomass and thereby contribute significantly to the transfer of energy to higher trophic levels?

## MATERIALS AND METHODS

#### Study Area and Sampling

Roskilde Fjord (RF; see **Figure S1** for a map of the study area and sampling pier) is a mesohaline, shallow and well-mixed estuary with a long residence time (up to 2 years) due to low river discharge and low tidal influence (Kamp-Nielsen, 1992). The estuary consists of two larger broads connected by a long narrow channel oriented in a south-north direction. It receives relatively high nutrient inputs due to dominance of agriculture in the RF catchment, which enhance primary production (Staehr et al., 2017).

Surface water (2 L) was sampled with a bucket from the Risø pier (55◦ 41′ 30.19′′N, 12◦ 4 ′ 55.24′′E; **Figure S1**) almost every day from 15 February 2016 until 01 November 2017. Samples were delivered to the laboratory for immediate analysis within 10 to 20 min after sampling. Temperature and salinity were measured with an YSI Professional Plus multiparameter handheld meter (YSI, USA). Phytoplankton was analyzed with a pulse shape recording flow cytometer, whereas ciliates were analyzed by a color FlowCAM IV (see below).

#### Laboratory Analysis Nutrients

Dissolved inorganic nutrient samples were stored frozen in 30 ml acid-washed plastic bottles. The samples were analyzed on a San ++ Continuous Flow Analyser (Skalar Analytical B.V, Breda, NL) as previously described (Grasshof, 1976; Kaas and Markager, 1998). Detection limits were 0.04, 0.1, and 0.3 µmol L−<sup>1</sup> for NO<sup>−</sup> 2 , NO<sup>−</sup> 3 , NH<sup>+</sup> 4 . Dissolved inorganic nitrogen (DIN) concentrations were calculated as the sum of the concentrations of NO<sup>−</sup> 2 , NO<sup>−</sup> 3 , and NH<sup>+</sup> 4 .

#### FlowCAM

Ciliate abundances and body volumes were analyzed from live samples using a color FlowCAM IV (Fluid Imaging Technologies, USA), following Calbet et al. (2014). From 15 March 2016 until 20 February 2017, samples were analyzed using a 4x objective and a flow cell FC300, whereas samples from 21 February until 1 November 2017 were analyzed using a 10x objective and a flow cell FC100x2. User calibration with standard beads (polymer microspheres of 50µm, Thermo scientificTM) were done for both magnification to validate counts and volumes calibration. The instrument was run in auto image-mode for both magnifications, capturing all particles in the range of 15–1,000µm. The analysis time for each sample was ca. 40 min. (4x) or ca. 3 h (10x), corresponding to the analyzed volume of 20 and 10 ml, respectively. During analysis, samples were gently stirred (approx. 3.14 rad s−<sup>1</sup> ) and kept under dim light at room temperature (about 15– 20◦C). We assumed that cell loss during the analysis was insignificant, as no differences were observed between cell numbers recorded at the start and end of the runs. After sample processing, recorded images were manually sorted into ciliate morphotypes and dinoflagellate trophy, based on features such as cell size, color, and general morphology. Equivalent Spherical Diameter (ESD) and body volume were estimated by the software package VISP 3.17 (FluidImagineTM). Cell size was estimated by the area based diameter (ABD) algorithm of VISP 3.17 (Jakobsen and Carstensen, 2011), except for tintinnids. This group of loricated ciliates can have their volumes overestimated from FlowCAM images, thus their individual volume was calculated as a prolate spheroid with diameter equal to the lorica width, and the length as 120% of width. Carbon biomass was obtained by converting volume to biomass using a generic protist volume-to-carbon conversion formula (Menden-Deuer and Lessard, 2000). The higher magnification allowed for identification of more morphotypes, which were grouped when necessary to match the 4x morphotypes. In all samples non-identified blurred images of ciliates were present, however these accounted for <5% of the observations in all samples.

#### Flow Cytometer

We employed a pulse-shape recording flow cytometer (PFCM) (CytoSense, Cytobuoy, NL) to analyse phytoplankton. This technique is suitable for rapid analysis of the phytoplankton size spectra, providing cell counts comparable to those obtained with traditional microscopy and more reliable information for picoplankton (Haraguchi et al., 2017). Additionally, it also provides information on cell size and morphology due to its capacity to store the optical profile for each particle, recorded as they travel through the flow cell. The instrument has a 488 nm laser, fluorescence sensors (yellow/green ∼550 nm, orange ∼ 600–650 nm and red ∼650–700 nm) and two scatter sensors, for light scattered parallel (forward scatter) and orthogonal (sideward scatter) to the incident laser beam. All the optical sensors are duplicated (except for the forward scatter) but set to different sensitivity for precise recording of both larger and smaller particles. Optical particle profiles from live samples (500– 1000 µL, sampled at a flow rate of 8 µL s−<sup>1</sup> ) were collected using the software CytoUSB (cytobuoy.com), with a lower threshold of 30 mV for the high sensitivity red fluorescence sensor. This trigger was set to include only particles containing chlorophyll a (phytoplankton cells). Recorded cells were clustered according to similarities in their optical properties [length and total Forward Scatter (FWS), total red fluorescence (FLR); total orange fluorescence (FLO); total Sideward Scatter (SWS)], using the software CytoClus3 (cytobuoy.com). Particles were assigned to one cluster only and the same clustering algorithm was employed for all samples. Taxonomical information was obtained for some of the clusters based on their optical characteristics and photos taken by the equipment, which were cross-referenced with qualitative information obtained from live samples examined by light microscopy. Carbon biomass was obtained by converting total FWS to volume by applying the empirical formula in Haraguchi et al. (2017) and then converting volume to biomass using a generic protist volume-to-carbon conversion formula (Menden-Deuer and Lessard, 2000). Note that for some characteristic and/or abundant groups (e.g., chains, pico-eukaryotes, Teleaulax spp.) group-specific clusters were identified based on cells characteristics (size, shape, fluorescence). Other clusters, like nano-flagellates and micro-phytoplankton, comprised multiple species with lower relative abundance and no specific clusters could be drawn. For those, cluster definition was based on general functional features such as size and fluorescence levels.

#### Ciliates Potential Grazing Rates Ciliate Trophic Strategy Definitions

During the study, 13 different ciliate morphotypes were identified (**Figure 1**). Heterotrophs were divided into three groups: (i) herbivores that were always observed with the same type of food in vacuole; (ii) herbivores with varying food content; and (iii) carnivores (ciliates that can also feed on other ciliates). Mixotrophs were identified as ciliates with a strong colouration and chloroplasts located throughout the cell periphery, and were assigned into two different groups following Mitra et al. (2016). Based on compiled information from the literature (**Table 1**), the ciliate morphotypes were grouped into five trophic strategies:


Strombidium cf. capitatum; (4) Strombidium cf. conicum; (5) Tinitinnids; (6) Mesodinium cf. velox; (7) cf. Pelagostrobilidium; (8) Balanion comatum; (9) cf. Urotricha; (10) Choreotrichida; (11) Oligotrichida; (12) Askenasia; and (13) Didinium. Assigned trophy strategies are displayed as letters, accompanying the morphotypes numbers: (a) Specialist Non-Constitutive Mixotroph (SNCM); (b) General Non-Constitutive Mixotroph (GNCM); (c) Selective phagotroph (SP); (d) Generic phagotroph (GP); and (e) Carnivore (Cv). Note that some of displayed cells of M. rubrum (1a) are grabbing particles that are not their preferential prey (Teleaulax spp).



#### Prey Definition

FlowCAM IV allowed for identification of different ciliate morphotypes, and for the identification of prey items inside many of the ciliates. Food items were identified from food vacuoles, and based on their characteristics (size, shape, color), they were related to phytoplankton groups when possible. While some morphotypes were always observed with similar food vacuoles, others were observed with food vacuoles of different shapes, colors and sizes, indicating selective or generic prey preferences among ciliates types (**Figure 1**). Potential prey items used in the modeled grazing rates were assigned to each ciliate morphotype based on FlowCAM images (**Figure 1**; **Table 1**).

#### Potential Grazing

The potential grazing of different ciliate groups were estimated by calculating ingestion and clearance rates (see **Table 2** for nomenclature) following Hansen et al. (1997), with the exception of Mesodinium rubrum (see below). Generic maximum ingestions and clearance rates (normalized to predator volume) at standard temperature of 20◦C were estimated from the volume of the ciliate cell using parameters from Hansen et al. (1997).

$$I\_{\text{max}} = 50.1 \cdot V\_{cl}^{-0.225} \tag{1}$$

$$C\_{\text{max}} = 70.6 \cdot 10^{-6} \cdot V\_{\text{cil}}^{-0.225} \tag{2}$$

From these, the half-saturation food density was found

$$K\_m = \frac{I\_{max}}{C\_{max}} = 0.710 \cdot 10^6 \tag{3}$$

equivalent to 0.710 ppm. Cell-specific ingestion rates for the different ciliate groups were subsequently calculated as a function of food density of preferred prey (d) and water temperature (T) using the overall average Q<sup>10</sup> =2.8 from Hansen et al. (1997).

$$\begin{split} I\_{\rm cell}(d, T, V\_{\rm cl}) &= \, V\_{\rm cl} \cdot \frac{I\_{\rm max} \cdot d}{(K\_m + d)} \cdot Q\_{10} \, \frac{^{(T-20)}}{^{10}} \\ &= \, V\_{\rm cl} \cdot \frac{50.1 \cdot V\_{\rm cl} \, ^{-0.225} \cdot d}{(K\_m + d)} \cdot Q\_{10} \, \frac{^{(T-20)}}{^{10}} \end{split} \tag{4}$$

The ingestion rate for the entire ciliate group was found by summation across all cells within the group and accounting for the sampling volume

$$I(d, T, V\_{cl}) = \frac{1}{V\_{sample}} \cdot \sum I\_{cell}(d, T, V\_{cl}) \tag{5}$$

This volume-specific ingestion rate was subsequently converted into carbon units with a scaling factor (r<sup>C</sup> :V), which was calculated from the volume-to-carbon conversion formula (f()) of Menden-Deuer and Lessard (2000) applied to the sample distribution of prey cell volumes (Vprey)

$$r\_{C:V} = \frac{\sum \text{cellcarbon}}{\sum \text{cellvolume}} = \frac{\sum f(V\_{\text{prey}})}{\sum V\_{\text{prey}}} \tag{6}$$

When prey densities (d) were smaller than the half-saturation food density (Km), potential food limitation is indicated. When food-limitation is observed, the prey ingested by a given ciliate is smaller than what it could potentially ingest based on its body volume, thus resulting grazing rates are smaller than expected for that ciliate.

For M. rubrum, potential grazing was defined based on cell counts, as daily specific prey intake has previously been reported to vary between 0.4 and 5 Teleaulax cells for each Mesodinium (Smith and Hansen, 2007). Three potential ingestion scenarios were estimated based on minimum (0.4 cryptophyte

TABLE 2 | Variables used for estimating ciliate grazing rates in Eq. 1–6.


cells Mesodinium−<sup>1</sup> day−<sup>1</sup> ), maximum (5 cryptophyte cells Mesodinium−<sup>1</sup> day−<sup>1</sup> ), and average (2.7 cryptophyte cells Mesodinium−<sup>1</sup> day−<sup>1</sup> ). Daily intake rates (min, average, and max in cells L−<sup>1</sup> d −1 ) were calculated by scaling the cell-specific rates with the observed M. rubrum cell density. Finally, these daily intake rates were converted into carbon units by multiplying with the average individual cryptophyte C biomass of that particular sampling day.

#### STATISTICAL ANALYSIS

To evaluate the seasonality of the different ciliate morphotypes, a linear mixed model was fitted to quantify the biomass variation across months. Biomass observations were log-transformed to account for scale-dependent variability. The model included a random factor for the seasonal variation between the years and residual variation included an autoregressive process AR(1) to account for potential autocorrelation between the daily samples. The mixed effect models were fitted in R (R Core Team, 2017), using the nlme package (Pinheiro et al., 2017).

Prey-predator biomasses dynamics were quantified as the distance between consecutive observations, as the distance is proportional to differences in biomasses of both prey and predator. The distances were calculated as the Euclidean distance of log-transformed biomass data of phytoplankton and ciliates, which were separated according to the ciliates trophic strategies and their assigned preys (**Table 1**). Therefore, the Euclidean distance was used as a descriptor of the prey-predator dynamics, with longer distances associated with more variability in the biomasses of prey-predator pairs. As the study does not encompasses two full year cycles, differences between years were assessed considering only the productive season (15 March until 1 November).

To summarize the potential C flux through ciliates, simplified estimates of prey C biomass and potential grazing by the different trophic strategies were calculated as daily average for the productive season of each year.

#### RESULTS

#### Physical Environment

Water temperature in the study area varied seasonally from ∼4 ◦C in winter to 20◦C during summer, whereas salinity variation did not exhibit any consistent seasonal pattern, except for a small decrease during winter (**Figure 2A**). Although the 2 years appeared similar, subtle differences were observed: (i) spring warming was faster and summer temperatures remained at maximum level longer in 2016 than in 2017 (**Figure 2A**); (ii) salinity was lower in 2016 (**Figure 2A**). DIN varied seasonally, from >20µM in winter to < 2µM in summer, as well as inter-annually with 2-fold higher winter concentrations observed in 2016 in comparison to 2017 (**Figure 2A**).

#### Phytoplankton and Microzooplankton Overall Dynamics

Interannual variations were observed, with considerably higher phytoplankton biomass in 2016 (**Figure 2B**), and higher ciliate biomass in 2017 (**Figure 2C**). The low phytoplankton biomass in 2017, compared to 2016, was mainly associated with lower abundance of cryptophytes, small nano-flagellates, and picoeukaryotes in 2017 (data not shown). The average ciliate biomass for the entire study period was 22.3 µg C L−<sup>1</sup> , and the average biomass during the productive season (March– November) was 12.3 and 55.8 µg C L−<sup>1</sup> in 2016 and 2017, respectively. Heterotrophic dinoflagellates were also more important in 2017 than in 2016 with average biomass during the productive season of 4.45 and 2.37 µg C L−<sup>1</sup> , respectively. Overall, heterotrophic dinoflagellate biomass was much smaller than ciliate biomass, averaging 2.72 µg C L−<sup>1</sup> for the entire period, although biomass peaks could reach up to 40 µg C L −1 (**Figure 2D**). Higher summer biomasses of heterotrophic dinoflagellates were associated with increased abundance of large-sized heterotrophic dinoflagellates, such as Polykrikos, Protoperidinium, and Warnowiids.

The higher ciliate biomass in 2017 was associated with increasing abundance of representatives of all trophic strategies (**Table 1**). The dominance of heterotrophs (herbivores + carnivores) was higher during spring and summer of 2016, while 2017 was characterized by slightly varying high biomass for most of the year (**Figure 3A**). Biomass of GNCM (mixotrophs excluding M. rubrum) was generally higher in 2017 than in 2016, with values peaking in late spring and summer for both years (**Figure 3B**). The biomass proportion between GNCM and heterotrophic ciliates was variable and higher contributions (up to 90% of the ciliates biomass, excluding M. rubrum, in 2017) of GNCM were recorded in summer (**Figure 3C**). Biomass of M. rubrum was also higher in 2017 and generally above 10 µg C L−<sup>1</sup> from April to October (**Figure 3D**).

#### Seasonality of Ciliate Morphotypes

Some ciliate morphotypes exhibited distinctive and recurring seasonal pattern (**Table 3**). cf. Pelagostrombilidium occurred during winter at temperatures <10◦C, whereas Askenasia and Strombidium cf. capitatum were associated with higher

temperatures, mainly during late spring and summer (**Figures 4A–C**). Other ciliate morphotypes exhibited distinct seasonal patterns of high biomass, although their presence was not restricted to a specific seasonal window. S. cf. conicum and tinitinnids were most abundant during late spring and summer, but also found during autumn (**Figures 4D,E**). Other morphotypes, such as cf. Urotricha (**Figure 4F**), did not display any pronounced recurring seasonal pattern despite of the high

temporal variability. In spite of the significant differences among months (**Table 3**), some ciliates (Didinium, Mesodinium cf. velox, and Choreotrichida) did not have any clear occurrence pattern over the study period.

#### Potential Grazing by Ciliates

Potential grazing varied broadly over time for the five trophic strategies of ciliates and their associated preys (**Table 1**), displaying highly variable patterns of potential food limitation (**Figure 5**).

Temporal variations in biomass of M. rubrum (SNCM) and its cryptophyte prey were inversely related (**Figure 5A**). Teleaulax spp. was most abundant in winter when SNCM biomass was low, and decreased in late spring and summer when blooms of M. rubrum occurred. The seasonal pattern in SNCM biomass also shifted between the 2 years; the biomass was low in summer 2016, whereas it remained high from mid-April until late October in 2017.

GNCM ciliates attained similar biomass levels in 2016 and 2017, but their potential ingestion was lower in 2017 due to stronger food-limitation (**Figure 5B**), which was observed in 98% of the occasions in 2016 and in all observations in 2017. Consequently, biomass and potential ingestion observations were decoupled for this group. On the other hand, it also indicates strong grazing pressure on small phytoplankton cells by these ciliates during spring and summer (**Figure 5B**).

TABLE 3 | Seasonality test for the log-transform of carbon biomass of different ciliate morphotypes.


Significant variation (P < 0.05) among months are highlighted in bold.

SH ciliates were often food limited (82% of observations), even during winter and spring (**Figure 5C**). SH grazing impact varied interannually and was most intense from May to July in 2016, and from June to October in 2017. SH food limitation was observed more frequently in 2017 than in 2016 (60 and 77% of the observations in 2016 and 2017, respectively).

GH ciliates were food limited in 49 and 66% of the observations in 2016 and 2017, respectively (**Figure 5D**). The potential grazing for this group of ciliates seemed to be more intense in autumn 2017 compared to autumn 2016, and potential food limitation appeared to have started earlier in 2017 (around April) than 2016 (around June). A shift in the most abundant morphotype was observed between years, with Choreotrichida being the main GH morphotype in 2016, and cf. Urotricha in 2017. Together with SH, these ciliates appeared to be the main grazers during the winter period.

The carnivore ciliates (Cv) had high grazing rates during summer (**Figure 5E**), associated with high biomass of Askenasia (**Figure 4B**), but carnivores grazing can also be high in other periods, such as the spring bloom with high abundance of Didinium (**Figure 5E**). During 2016, potential grazing had higher peaks, observed in distinct periods of the year, while potential grazing rates had lower maximum values in 2017 but were consistently higher throughout a longer period (until autumn). Food limitation was observed in 32% of the observations (14% in 2016, and 55% in 2017), and as for the other trophic strategies, more frequent in 2017.

#### Interannual Differences in Prey-Predator Dynamics

Dynamics of prey-predator biomasses were analyzed for the productive season (15 March to 01 November) of each year (**Figure 6**). The arrows represent the distance between observations, and longer arrows can be used as a proxy for higher variability in the prey-predator relationships.

For the SNCM-prey coupling, the average distance between observations was lower in 2016 (d = 1.19) than in 2017 (d = 1.36), indicating higher variability in 2017. For this pair, variation between sampled days were higher (41% higher than the rest of productive season) during late spring to summer (day of year 110–200) in 2016 (**Figure 6A**). Distances in 2017 were more evenly distributed, with the distances found in late spring to summer, being only 8% longer than for the rest of the productive season (**Figure 6B**).

For GNCM-prey and SH-prey pairs, average distance between observations were higher in 2016 than in 2017, with d<sup>2016</sup> = 1.48 and d<sup>2017</sup> = 0.74 for GNCM; and d<sup>2016</sup> = 0.94 and d<sup>2017</sup> = 0.85 for SH (**Figures 6C–F**).

GH-prey dynamics did not appear to differ between years, with d = 1.07 and d = 1.10 for 2016 and 2017, respectively (**Figures 6G,H**). Similarly, Cv.-prey observations average distance did not vary between years (d<sup>2016</sup> = 1.22 and d<sup>2017</sup> = 1.26), however differences over specific periods of the productive season were observed for this group (**Figures 6I,J**). From spring to early summer (day of year 75–160) in 2016, distances were about 15% longer than for the rest of the period, while in 2017, distances in the same period were 58% longer than the rest of the productive season.

A simplified scheme (**Figure 7**) summarizes the daily average biomass for each group of phytoplankton preys and the potential daily C intake for the different ciliate trophic strategies during the productive season of each year. In 2016, biomasses of all phytoplankton prey groups were higher than in 2017, except for nano-flagellates (5–15µm). The intake by ciliates was generally lower in 2016, with an overall relative removal of about 9% of phytoplankton daily standing biomass (**Figure 7A**). The potential grazing impact over the phytoplankton varied among the different trophic strategies, being low for SNCM and GH (∼3% of C prey biomass being grazed) and higher for GNCM and SH (about 19 and 14% of C prey biomass being grazed, respectively). In 2017, phytoplankton standing biomasses were modest when compared to the previous year, and potential grazing was higher for most ciliate trophic strategies (**Figure 7B**). The relative impact of grazing by the ciliates was about 31% of phytoplankton standing biomass in 2017, with GH being able to remove 12% of its prey biomass, while GNCM and SNCM being able to graze 37 and 46% of their preys standing biomasses daily.

#### DISCUSSION

Our results underline that microzooplankton, especially ciliates, can be an important component of the pelagic food web in temperate nutrient-rich estuaries at all times. On top of that, we also demonstrate that different components of the heterotrophic protist plankton have distinct ecological strategies, each affecting the phytoplankton community in its own way. Even though our results are based on standing biomasses and do not include any direct rate measurement (neither primary production nor grazing), we believe that the high temporal resolution data on both phytoplankton and microzooplankton support our findings. Furthermore, our results explore the diets of different ciliates in detail, revealing the existence of various prey-predator

couplings that are likely to influence the community functioning in distinct ways. Additionally, we structured potential grazing calculations as proposed by Hansen et al. (1997), following the Michaelis-Menten kinetics, in contrast to the rectilinear model approach suggested by other authors, e.g., Zervoudaki et al. (2009). The Michaelis-Menten kinetics allow grazing estimation at varying prey concentrations, whereas a rectilinear model assumes a linear increase in ingestion rate until food saturation is reached, which can overestimate the potential grazing.

Classically, the relative importance of trophic pathways of phytoplankton primary production in any specific ecosystem is highly dependent on the balance between nutrient inputs and recycling, with increasing dominance of larger phytoplankton species and mesozooplankton in areas with higher input of new nutrients (Azam et al., 1983; Fenchel, 1988; Buitenhuis et al., 2006). Although Roskilde Fjord (RF) is a nutrient rich system (Staehr et al., 2017), its inner portion, which is also the area of interest in this study, has always been dominated by small phytoplankton organisms (<20µm), not diatoms, evidenced by the long term monitoring data. Besides microzooplankton, potential consumers of phytoplankton in RF are pelagic mesozooplankton and benthic filter feeders; however, benthic grazers were not sampled in this study, therefore we cannot compare directly the relative importance of pelagic vs. benthic grazing. We also acknowledge that protozooplankton

further details on Km definition and calculation, see "Potential Grazing" in the section Materials and Methods.

grazing in RF was only calculated for larger protozooplankton (ciliates and heterotrophic dinoflagellates). Other organisms, such heterotrophic nano-flagellates and small heterotrophic dinoflagellates (<15µm), were not properly recorded by any of the methods used in this study, and likely increase the real contribution of protozooplankton grazing.

Heterotrophic dinoflagellates can be an important component of microzooplankton and microbial food webs (Sherr and Sherr, 2007), but their importance in RF seems low compared to ciliates. Heterotrophic dinoflagellates biomass was less than 10% of ciliate biomass in most samples, even though it was higher in a few instances. On top of that, dinoflagellates have lower growth and grazing rates than ciliates (Hansen, 1992; Jakobsen and Hansen, 1997), which indicates a minor role of this group for carbon cycling in RF. Seasonality of heterotrophic dinoflagellates was marked by the increased occurrence of large sized cells during summer. The observed increase in size of heterotrophic dinoflagellates is in agreement with what has previously been described for other open areas, such as Kattegat and Kiel Bight (Smetacek, 1981; Hansen, 1991), and might reflect an adaptation in the dinoflagellate assembly to the larger prey items. However, summer is also characterized by higher biomass of ciliates, and

FIGURE 6 | Prey-predator biomass dynamics for the five trophic strategies in 2016 (left) and 2017 (right): SNCM, Specialist Non Constitutive Mixotroph (A,B), GNCM, Generalist Non Constitutive Mixotroph (C,D), SH, Selective Herbivore (E,F), GH, Generic Herbivore (G,H), and Cv, Carnivore (I,J). Arrows represent the distance between observations (Euclidean distance) and color gradient indicate the day of year. Text depicts the phytoplankton prey assigned for each predator group (green), and the average Euclidean distance (d) between observation in the productive season of each year (black). Right limits of the yellow colored area indicate the half-saturation constant of food density (Km), and the observations to the left, within the colored area, indicate potential food limitation (except for SNCM, see explanation in Figure 5).

the increasing proportion of large heterotrophic dinoflagellates might indicate a change in ecological strategy which aims to avoid competition with ciliates that graze on smaller phytoplankton.

Salinity variations are mainly associated with freshwater discharges during winter, enhancing dissolved inorganic nitrogen (DIN) concentrations that are mainly consumed by

phytoplankton during spring (Staehr et al., 2017). Although inter-annual differences in salinity were modest, DIN inputs were higher in 2016. This could explain the intense spring bloom observed in 2016 compared to the modest one in 2017, but it cannot explain entirely the consistently higher biomass of all ciliate types and heterotrophic dinoflagellates in 2017 compared to 2016. Differences in phytoplankton biomass could be due to the higher microzooplankton biomass in 2017 compared to 2016, indicating a more intense grazing in 2017. Thus, it is likely that phytoplankton is top-down controlled by microzooplankton in RF, and that due to fast turnover rates of both communities, primary production is higher despite relatively lower phytoplankton biomasses. Additionally, it is likely that the high DIN inputs in the winter of 2016 disrupted the coupling between phytoplankton and their ciliate predators that was better depicted in 2017. DIN concentrations were depleted to similar levels after the spring bloom in both years, indicating that higher concentrations in the winter of 2016 probably supported the larger bloom observed in this year. This large perturbation in the prey-predator coupling at the beginning of the productive season probably shaped also the rest of the year, causing interannual differences in prey-predator coupling, and reduced the number of observations with foodlimitation in 2016. The prey-predator coupling varied among trophic strategies, resulting in the following contrasts between the studied years: (1) larger variation in the coupling in 2016 than in 2017 (for GNCM and SH); (2) differences in the timing when variations were observed during the productive season (SNCM and Cv.); and (3) changes in the dominance of the main morphotype (GH).

Previous studies have described various seasonal dynamics of ciliates in coastal and estuarine areas, and their importance in the trophic energy transfer. In Gulf of Maine, distinct ciliate assemblages were dominant at specific seasons, with larger ciliates found in spring, associated with larger preys, and smaller ciliates in other periods, when available prey were smaller (Montagnes et al., 1988). Conversely, ciliates in Coos Bay were reported to control small-sized phytoplankton throughout the year (Cowlishaw, 2004). In our study, ciliate biomass and their estimated potential grazing rates were in the same range as biomass of phytoplankton prey, especially for sizes <15µm and cryptophytes, indicating selective pressure by ciliates grazing on small phytoplankton cells. The phytoplankton groups with lower standing biomass in 2017 compared to 2016 (cryptophytes, pico-eukaryotes, and small nano-flagellates) were subject to intense grazing pressure by different ciliate morphotypes, which probably substituted each other over the year in succession. These ciliate morphotypes also tended to be more food limited (GNCM and SH). Furthermore, as potential ingestion rates were modeled from standing biomass and not directly measured, it is likely that our grazing rates, especially for 2017, are underestimated due to intense grazing that resulted in low standing biomass of some phytoplankton groups. Yet, they provide valuable information on potential impact of ciliates, suggesting that ciliate grazing is an important driver for the phytoplankton succession, and that the microbial loop is a key pathway in the RF food web, despite of the high new nutrient input from land, which was assumed to promote the short-chain food web (Azam et al., 1983). Thus, ciliates might be a key player in the overall trophic transfer in RF, as an intermediate step between small-sized phytoplankton and mesozooplankton (Calbet and Saiz, 2005) and benthic filter feeders (Zeldis et al., 2004).

Organism with acquired phototrophy, including ciliates, are considered to have an increased gross growth efficiency, and therefore enhancing carbon export and nutrient cycling (Stoecker et al., 2009, 2017; Mitra et al., 2016). Experimental studies have shown that among non-constitutive mixotrophs, the acquired capability of photosynthesis can increase the specific growth rates. However, this should not be generalized, as specific growth rate varies among species, and for some mixotrophs, it can still be low compared to autotrophs or heterotrophs (Jakobsen et al., 2000; Jakobsen and Strom, 2004; Schoener and McManus, 2012). Despite the fact that these organisms can often dominate microzooplankton communities and be responsible for a considerable proportion of primary production and/or grazing, in situ ecological data describing processes and importance of those organisms are still sparse (Stoecker et al., 2017). It is hypothesized that autotrophic and heterotrophic organisms dominate during the developmental phase of the ecosystems, while mixotrophs dominate in mature systems, benefiting from a flexible nutrition (Mitra et al., 2014). Our study shows that mixotrophic nutrition is increasingly important at the decline of spring bloom and during summer, when the system shifts from net autotrophy to net heterotrophy (Staehr et al., 2017). Additionally, our results demonstrated that mixotrophs (SNCM and GNCM, and even morphotypes among GNCM) have distinct temporal dynamics, reflecting different ecological strategies.

Mesodinium rubrum acquires its chloroplasts from specific prey (Teleaulax spp.), taking strong control over the photosynthetic apparatus of its prey, and being able to survive at low prey concentrations (Smith and Hansen, 2007; Peltomaa and Johnson, 2017). These organisms have been the focus of many ecophysiological studies. Blooms of M. rubrum are common in coastal areas and have been associated with occurrence of cryptophytes (Johnson et al., 2013; Hamilton et al., 2017; Lips and Lips, 2017). Our data shows that periods with high M. rubrum biomass are associated with low cryptophyte biomass, except for the spring bloom (March–April). Even though high biomass of M. rubrum occur around summer, M. rubrum was present in most (>95%) of samples during all seasons. Using the higher magnification of FlowCAM in 2017 allowed us to observe cell discolouration and size reduction of M. rubrum individuals during late summer/autumn, coinciding with the period of higher temperatures and low DIN in RF. At the same time, cells of M. rubrum were observed capturing particles other than cryptophytes with their oral tentacles (see **Figures 1,1a**lower row). This emphasizes that the ecological flexibility of M. rubrum remains poorly understood, but that the flexibility could also explain the apparent success of M. rubrum in many environments and, in the case of RF, over the entire year.

In contrast to M. rubrum, oligotrichids that retain chloroplasts (GNCM), do not seem to be able to maintain them and need to feed at higher rates than M. rubrum, to replace aging plastids at higher rate (Jakobsen and Strom, 2004; Schoener and McManus, 2012). The plastid turnover rate depends on the availability of the plastid source, and GNCM usually use plastids of different origin, although they have a preferred type, depending on the species (Stoecker et al., 1988; Stoecker and Silver, 1990; Schoener and McManus, 2012). Under controlled conditions, Strombidium rassoulzadegani had newly acquired plastids positioned at the cell periphery within 30 min after being offered new prey, and it replaced all plastids after 2–3 days, yielding a plastid turnover rate slower than other Strombidium species (Stoecker and Silver, 1990; Schoener and McManus, 2012). Thus, GNCM can ingest phytoplankton prey at similar rates as heterotrophic ciliates (Jakobsen and Strom, 2004; Schoener and McManus, 2012), having a potential impact on carbon cycling, especially when larger species are present. GNCM in our study were composed of different species, with higher biomasses being associated with the proliferation of large morphotypes (Strombidium cf. conicum and S. cf. capitatum). These morphotypes were usually found during spring and summer, but higher biomasses were observed mainly in late spring. As those ciliates were food limited in most occasions (<95% of observations), their estimated grazing rates were around 10 times smaller than their potential ingestion based on cell volume. This indicates that despite of having a strong presence in RF, grazing imposed by mixotrophs is lower than by heterotrophic ciliates. On the other hand, mixotrophic nutrition supports their survival until suitable prey is encountered.

High volatility in the dominance of different types of ciliates was observed in this study. The temporal dynamics of the different trophic strategies demonstrated that mixotrophs (SNCM and GNCM) and carnivores exhibited strong seasonality in their occurrence and potential impact, predominating during summer. Conversely, herbivores (SH and GH) appeared to be widely distributed throughout the year, but apparently with different ecological strategies. GH types are present in low concentrations in most of the samples; whereas SH types occur with variable importance over the annual cycle (cf. Pelagostrobilidium in winter and tintinnids in summer), but together covering different seasons. This illustrates how complex and flexible ciliate communities can be, showing a great potential to control and drive changes of phytoplankton. We also observed a gradient in the prey-predator coupling among trophic strategies of ciliates, in which the more selective ciliates tend to be more efficient in removing their prey and are more often food limited, while the opposite is observed for ciliates that have a wider prey range, with exception of SNCM. However, those patterns are subjected to variability introduced by exogenous forcing (i.e., nutrient loads).

# CONCLUSION

The use of in-flow techniques supports the analysis of phytoplankton and their microzooplankton grazers with high frequency. Additionally, the use of live samples further allows exploring trophic characteristics of ciliates, providing insights to the trophic strategies and specific prey classification. Ciliates are likely the main pelagic grazers in RF and probably play an essential role in the food web, linking primary production of small-celled organisms to higher trophic levels. Although ciliates are most abundant during summer, they are still important in other seasons. This is due to the range of different ecological strategies within the diverse ciliate community, combining different trophic strategies with different physiological adaptations. Thus, the ciliate community structure is highly complex and most likely an important driver for structuring the phytoplankton community.

# AUTHOR CONTRIBUTIONS

LH, HJ, JC, and NL contributed conception and design of the study; LH and HJ collected the data and organized the database; LH, HJ, and JC performed the data analysis; LH wrote the first draft of the manuscript. All authors contributed to manuscript writing, revision, read and approved the submitted version.

#### FUNDING

LH was supported by a grant from the Brazilian program Science without Borders/CAPES (Grant No. 13581-13-9). This study is a

#### REFERENCES


contribution from the BONUS COCOA project (grant agreement 2112932-1) supported by Innovation Fund Denmark and the European Union.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars. 2018.00272/full#supplementary-material

Figure S1 | Map showing the sampled station location (red star) at Roskilde Fjord (Denmark).


**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 © 2018 Haraguchi, Jakobsen, Lundholm and Carstensen. 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) and the copyright owner(s) 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.

# The Importance of Mesodinium rubrum at Post-Spring Bloom Nutrient and Phytoplankton Dynamics in the Vertically Stratified Baltic Sea

Inga Lips\* and Urmas Lips

*Department of Marine Systems, Tallinn University of Technology, Tallinn, Estonia*

The inter-annual dynamics of the photosynthetic ciliate *Mesodinium rubrum* in the central Gulf of Finland in spring-summer continuum during 5 years were followed. The analysis was mainly based on high-resolution measurements and sampling in the surface layer along the ferry route Tallinn-Helsinki. The main purpose was to analyze the dynamics of *M. rubrum* biomass, its contribution to the photosynthetic plankton biomass, and the influence of water temperature and variations of inorganic nutrients in the surface and sub-surface layer on its dynamics. The analysis revealed that the outcome of the *M. rubrum* bloom in spring was largely related to the surface layer water temperature—in the years of earlier warming, the higher biomass of this species was formed. The photosynthetic ciliate was an important primary producer in all studied years during the late phase or post-spring bloom period in the Gulf of Finland. The maximum proportion of *M. rubrum* in the photosynthetic plankton community was estimated up to 88% in May and up to 91% in June. We relate the observed post-spring bloom decrease of phosphate concentrations in the surface layer to the dominance and growth of *M. rubrum*. We suggest that this link can be explained by the vertical migration behavior of *M. rubrum* and phosphate utilization in the surface layer coupled with inorganic nitrogen assimilation in the sub-surface layer. Thus, the dynamics of *M. rubrum* could strongly influence the amount of post-spring bloom excess PO3<sup>−</sup> 4 in the euphotic layer and the depth of nitracline in the Gulf of Finland.

Keywords: Mesodinium rubrum, spring bloom, nutrients, stratification, Baltic Sea

#### INTRODUCTION

Phytoplankton production, together with the terrestrial organic carbon load, is the largest primary source of organic carbon to the Baltic Sea (Kulinski and Pempkowiak, 2011). As total annual ecosystem respiration in temperate estuaries and estuarine type seas like Baltic Sea exceeds gross primary production, the temporary shift to autotrophy state only occurs during seasonal and episodic bloom events of photosynthetic plankton when photosynthesis exceeds total system respiration (Cloern et al., 2014). In the Baltic Sea, during the phytoplankton spring bloom, up to 60% of annual carbon fixation takes place, and 40–80% of this fixed carbon sinks out from the surface layer (Heiskanen, 1998; Tamelander and Heiskanen, 2004). The spring bloom

Edited by:

*Riina Klais, University of Tartu, Estonia*

#### Reviewed by:

*Matthew D. Johnson, Woods Hole Oceanographic Institution, United States Harri Kuosa, Finnish Environment Institute, Finland*

> \*Correspondence: *Inga Lips inga.lips@ttu.ee*

#### Specialty section:

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

Received: *21 September 2017* Accepted: *29 November 2017* Published: *12 December 2017*

#### Citation:

*Lips I and Lips U (2017) The Importance of Mesodinium rubrum at Post-Spring Bloom Nutrient and Phytoplankton Dynamics in the Vertically Stratified Baltic Sea. Front. Mar. Sci. 4:407. doi: 10.3389/fmars.2017.00407* leads to depletion of inorganic nutrients in the euphotic layer and through the sedimentation of phytoplankton-derived organic carbon to the acceleration of benthic respiration and nutrient regeneration rates (Conley and Johnstone, 1995).

Spring bloom in the Baltic Sea is co-dominated by diatoms and dinoflagellates (e.g., Kononen and Niemi, 1984; Wasmund and Uhlig, 2003). The late phase of the spring bloom (in May) in the Gulf of Finland is dominated by vertically migrating dinoflagellates together with ciliates (Heiskanen, 1995; Höglander et al., 2004; Lips et al., 2014). Ciliates are an important trophic link between primary producers and metazoa consuming a significant fraction of smallsized phytoplankton and bacterioplankton production and are important in remineralization of macronutrients (Rivkin et al., 1999; Calbet and Landry, 2004). Besides this, ciliates can also be significant contributors to primary production through mixotrophy which is the occurrence of phagotrophy and phototrophy in the same organism. Mixotrophic oligotrichs have been reported both in freshwater and in seawater ecosystems (Esteban et al., 2010).

Mass occurrences of photosynthetic ciliate Mesodinium rubrum Lohmann 1908 (Myrionecta rubra Jankowski 1976) are reported around the world (e.g., Mackenzie and Gillespie, 1986; Crawford, 1989; Wilkerson and Grunseich, 1990; Cloern et al., 1994; Johnson et al., 2013; Kang et al., 2013). In the Baltic Sea, the highest abundances/biomasses and largest size distribution of M. rubrum are observed after the diatom-dinoflagellate dominated spring bloom, usually in May–June (Lindholm, 1985; Passow, 1991; Rychert, 2004; Thamm et al., 2004). The peak of M. rubrum biomass mostly coincides with the period when nitrates are exhausted from the upper mixed layer and the increase of photosynthetic biomass is mostly regarded to be based on regenerated nutrients (according to Dugdale and Goering, 1967).

M. rubrum is extremely mobile, known to be fastest autotroph in the sea with a swimming velocity that is reported to reach 8.5 mm s−<sup>1</sup> (30 m h−<sup>1</sup> ; Smayda, 2010) and showing marked phototaxis and vertical migrations (Lindholm, 1985). Some studies already a long time ago demonstrated the very high rate of primary production of this species (e.g., Mackenzie and Gillespie, 1986; Crawford, 1989; Stoecker et al., 1991; review by Johnson, 2011). Increased temperature and water column stability, decreased salinity and depletion of dissolved inorganic nitrogen from the surface layer are known to have positive influence to the occurrence and abundance of M. rubrum (Lindholm and Mörk, 1990; Cloern et al., 1994; Montagnes et al., 2008; Johnson et al., 2013) in different locations worldwide. In several studies, the ability of directly utilize nitrate, ammonium, dissolved organic nitrogen (Lindholm and Mörk, 1990; Wilkerson and Grunseich, 1990; Tong et al., 2015) and phosphates (review by Lindholm, 1985; Tong et al., 2015) have been reported. M. rubrum mass occurrences tend to develop in a chemical environment where competing photosynthetic species are a resource (nutrient) limited or are not able to migrate vertically to exploit the pools of dissolved inorganic nutrients below the euphotic layer. Ability to migrate vertically complemented with efficient nutrient uptake has been considered to enable M. rubrum to compete with phytoplankters (Stoecker et al., 1991).

The main aim of this paper is to present the interannual dynamics of photosynthetic ciliate M. rubrum in the central Gulf of Finland in spring-summer continuum and to analyze how the increase in mixotrophic ciliate biomass affects the spatial distribution (both horizontal and vertical) and temporal variation of nutrients in the stratified water column. We hypothesize that the magnitude and intensity of M. rubrum bloom has a significant impact on the inorganic nutrient concentrations after the spring bloom and hence may influence the outcome of summer phytoplankton blooms. The analysis is based on high-resolution measurements and sampling in the surface layer along the ferry route Tallinn-Helsinki complemented with vertical profiling and sampling through the water column at one station close to the ferry line. We recognize that M. rubrum belongs to a species complex (Johnson et al., 2016) and that our data may include M. major and/or multiple variants of M. rubrum. However, since we did not measure the diversity of genetic variants, we will refer to all observed Mesodinium ciliates as M. rubrum.

# MATERIALS AND METHODS

#### Study Region

The dataset analyzed was collected during 5 years (2009– 2012, 2014) in the central part of the Gulf of Finland, the easternmost basin of the Baltic Sea (**Figure 1**). The Gulf of Finland is a stratified elongated estuarine basin where the general water movement in the surface layer is anticlockwise (Alenius et al., 1998) but the dynamics of water masses are very much meteorologically driven at the mesoscale. The surface layer salinity in the area is typically between 4 and 6 g kg−<sup>1</sup> , decreasing from west to east due to the major river discharge at the eastern end of the Gulf and slightly from south to north due to the anti-clockwise general circulation. A seasonal variation of inorganic nutrient concentrations is observed in the Gulf of Finland upper layer—minimum values in summer and maximum in winter. Nitrogen is considered the limiting nutrient in the Baltic Sea, and after the development of thermal stratification in spring the nitrogen-rich deeper layers are separated from the nitrogen-depleted surface layer causing the rapid decline in the phytoplankton biomass co-dominated by diatoms and dinoflagellates. At the same time there can be observed the residual amounts of phosphates and silicates in the surface layer after the spring bloom (e.g., Tamelander and Heiskanen, 2004). In summer, the strong stratification and nitrogen limitation give competitive advantages for cyanobacteria (Lips and Lips, 2008) able to fix molecular nitrogen and photosynthetic species able to migrate vertically in the water column (Lips et al., 2011).

#### Measurements and Sample Analysis

Measurements were conducted using autonomous ferrybox system (-4H- Jena Engineering GmbH) installed on board the passenger ferry "Baltic Princess" (AS Tallink Grupp) plying between Tallinn and Helsinki (**Figure 1**) in 2009–2012 and 2014. Seawater was pumped through the measuring system from ∼4

to 5 m depth while the ferry was moving at an average speed of 15–16 knots. The temperature, salinity, and chlorophyll a (Chl a) fluorescence were measured with the time resolution of 20 s corresponding approximately to a spatial resolution of 160 m.

Weekly-biweekly water sampling from up to 17 locations along the 75-km-long ferry route was conducted using an automatic refrigerating (4◦C) sampler (Sigma 900 MAX), being part of the ferrybox system. Sampling dates and number of samples collected for nutrient and phytoplankton analyses on each date along the south-north transect are shown in **Table 1**. Altogether 753 samples were collected from the surface layer in five studied years and analyzed to determine the concentrations of PO3<sup>−</sup> 4 , NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 , Chl a, and phytoplankton species composition, wet weight, and carbon (C) biomass.

Sampling and measurements on board the research vessel SALME were performed at the station AP5 (**Figure 1**) in springsummer 2010–2012 and 2014 (the sampling days and depths can be seen in **Figure 3**). CTD measurements using an Ocean Seven 320plus CTD probe (Idronaut S.r.l.) equipped with a Seapoint Chl a fluorometer were performed, and water samples with a vertical resolution from 5 to 10 m were collected. Collected water samples were analyzed to determine the same parameters as from ferrybox samples. On 20–21 May 2014, the 24 h campaign for measurements and sampling was performed. Vertical profiles of temperature, salinity, Chl a fluorescence and dissolved oxygen content were registered together with phytoplankton sampling with 2 h interval. Samples for nutrient analysis were collected with 6 h interval.

Inorganic nutrients were analyzed with the automatic nutrient analyzers µMac 1000 (Systea S.r.l.) and Lachat QuikChem 8500 Series 2 **(**Lachat Instruments, Hach Company). The nutrient analyses were performed according to the guidelines of the American Public Health Association (APHA, 1992; methods 4500-NO<sup>3</sup> and 4500-P for µMac 1000) and recommendations made by USEPA, ISO, and DIN standards (methods 31-107-04-1- D NO<sup>3</sup> (Egan, 2000) and 31-115-01-1-I PO<sup>4</sup> (Ammerman, 2001) for the Lachat instrument). The lower detection range for PO3<sup>−</sup> 4 and NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> <sup>3</sup> was 0.03 and 0.07µM, respectively.

The Chl a concentration in the water samples was determined using Whatman GF/F glass fiber filters following extraction at room temperature in the dark with 96% ethanol for 24 h. The Chl a content from the extract was measured spectrophotometrically (Thermo Helios γ) in the laboratory (HELCOM, 1988). Chl a fluorescence measured on board the research vessel and by the ferrybox system was calibrated against Chl a measured in the water samples. For each device and season, a linear regression equation between fluorescence and Chl a was found and used to convert fluorescence values into Chl a content values.

Phytoplankton sub-samples (100 ml) were preserved and analyzed according to the HELCOM recommendations and EVS-EN 15972:2011 standard. The wet weight biomasses were calculated according to Olenina et al. (2006), and

TABLE 1 | Ferrybox sampling dates and number of samples (*n*) in different years.


the phytoplankton carbon (C) content was calculated using C:biovolume factors according to Menden-Deuer and Lessard (2000). Carbon biomass of naked ciliates was calculated according to the method described by Putt and Stoecker (1989).

#### Calculations

For the years, when the autonomous profiler data were available, the stratification parameter P - J m−<sup>3</sup> was estimated after Simpson et al. (1990) as

$$P = \frac{1}{h\_2 - h\_1} \int\_{-h\_2}^{-h\_1} (\rho\_A - \rho \text{ (z)}) \, \text{gzdz}, \rho\_A = \frac{1}{h\_2 - h\_1} \int\_{-h\_2}^{-h\_1} \rho \text{ (z)} \, dz$$

where ρ(z) is the density profile in the water layer between the depths h<sup>1</sup> and h2. The obtained estimates of P characterize the strength of stratification between the depth of 40 m (h2) where nutrients were always available and the ferrybox sampling depth of 4 m (h1). The pre-processed CTD-profiles with a depth step of 0.5 m acquired at station AP5 were used.

All correlations between the data series are given as Pearson correlation coefficients. Only significant correlations are referred with a p-value < 0.05.

The consumption rate of phosphates was estimated as suggested by Lips et al. (2014) assuming that the temporal changes in phosphate concentrations, using a large enough number of analyses over the entire transect, were mostly related to the consumption. A linear regression equation between the measured phosphate concentrations and date (day of the year) was found for each post-spring bloom period using the least squares method. The slope of the found regression line was taken as the estimate of the consumption rate of PO3<sup>−</sup> 4 (units µM day−<sup>1</sup> ). The related need for NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> <sup>3</sup> was calculated according to the Redfield ratio (N:P ratio 16:1).

#### RESULTS

#### Inter-Annual Changes in Sea Surface Temperature and Salinity

The Gulf of Finland was partially ice-covered in 2009 and 2014, and ice-covered in 2010, 2011, and 2012; however, in 2012 the ice winter was 2 weeks shorter than on average. The spring warming of the surface layer differed in timing and rate of temperature increase in the studied years (**Figure 2**). The earliest warming was observed in 2014 when the average cross-gulf surface layer temperature exceeded 4◦C by 21 April and 5◦C 1 week later (see **Figure 2**, where relevant dates are marked with vertical red and black lines). The average cross-gulf surface layer temperature exceeded 4 and 5◦C a few days earlier in 2009 compared with 2012 (**Table 2**). The warming of the surface layer in 2010 and 2011 was slower compared with other years, and spring 2011 was characterized by several warming-cooling periods in April-May. Due to several cooling periods in spring 2011, the average cross-gulf temperature stayed around 4◦C until 21 May.

The surface layer salinity (**Figure 2**) in spring-early summer differed between the years, indicating the complex wind-driven circulation patterns and mixing in the Gulf of Finland surface layer in the studied years. On average, the lowest surface layer salinity was registered in 2009 and the highest in 2011. Most probably the high surface layer salinity in 2010 and 2011 was caused by intense vertical mixing in winters 2009–2010 and 2010–2011, respectively. For instance, in winter 2010–2011, there occurred two longer periods with westerly-southwesterly winds resulting in estuarine circulation reversals (eastward flow in the surface layer and westward flow in the deeper layers) that led to intense vertical mixing and a temporal collapse of vertical stratification (Liblik et al., 2013). The latter could also influence the concentrations of nutrients in the surface layer at the onset of the spring bloom.

A typical north-south gradient of the surface layer salinity (on average, salinity is higher near the Estonian coast than near the Finnish coast; e.g., (Kikas and Lips, 2016) was well seen in spring 2009. In spring 2012, a water tongue with slightly lower

TABLE 2 | The warming of the surface layer: dates (Julian day) when the certain average cross-gulf temperature was reached and dates when maximum *M. rubrum* biomass was observed.


\**First warming was registered on 27.04., after that, several colder periods with water temperature around 4*◦*C were registered.*

\*\**Warming over 5*◦*C registered on 08.-12.05.2011 was followed by lower water temperatures until 21.05.2011.*

salinity occurred in the central part of the Gulf. The horizontal distribution of salinity was uniform across the Gulf in March-April-May 2014, while slightly less saline waters appeared in the northern Gulf in June 2014 and the ordinary north-south salinity gradient was established.

Vertical stratification of the water column at low temperatures in April–early May is mostly controlled by the vertical distribution of salinity. The strongest stratification in the upper 40 m layer at station AP5 until mid-May was found in 2010 and 2012 (**Figure 3**) with stratification parameter varying from 32.5 to 45.7 J m−<sup>3</sup> and from 34.5 to 46.9 J m−<sup>3</sup> , respectively. At the same time, the stratification was weak in April–early May both in 2011 and 2014. The stratification parameter varied from 12.0 to 22.0 J m−<sup>3</sup> in 2011 and from 12.4 to 24.6 J m−<sup>3</sup> in 2014, although the surface layer salinity was clearly higher in 2011 than in 2014. In all studied years, vertical stratification strengthened in late May–June due to the formation of the seasonal thermocline. Vertical stratification in June was still stronger in 2010 and 2012 (stratification parameter exceeded 70 J m−<sup>3</sup> ) than in 2011 and 2014, but a change from a weak to strong stratification was also clear at the measurement site in June 2014 with the estimated stratification parameter up to 55.4 J m−<sup>3</sup> in late June.

## Inter-Annual Changes in Inorganic Nutrient Concentrations

Sampling for the analysis of inorganic nutrient concentrations started in the second week of April (during the phytoplankton spring bloom and was usually performed until the end of May-beginning of June when the NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 concentrations were below or close to the detection limit (**Figure 4**). In April, during the development of phytoplankton spring bloom, the NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 concentrations decreased weekly in all studied years. The highest initial NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 concentrations were measured in 2011 among the all five spring periods. The NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 concentrations were below or close to the detection limit by the 24 April in 2014, by the 2 May in 2011 and by the 7 May in 2012. In 2009 and 2010, the concentrations of NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 fell close to detection limit by 10 May (except in the southern part of the study transect in 2010). A late spring increase in the surface layer NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 concentration was detected in 2010 and 2011.

The measurements of NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 concentrations at station AP5 in 2010–2012 and 2014 (**Figure 3**) were conducted during the same period as sampling along the ferry route, and the analysis results allow following the depletion of inorganic nitrogen in the surface and sub-surface layer and deepening of the nitracline with time. By mid-April in 2010, the NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 concentrations in the upper 10 m layer were in the range of 0.2– 0.4µM whereas high levels were measured at the depths of 15 and 20 m (1.6 and 6.8µM respectively). A significant deepening of the nitracline was observed at the beginning of May, and higher concentrations were reintroduced to the upper layer due to the rise of the pycnocline at the end of May (**Figure 3**). In 2011, the upper 10 m layer was depleted of NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 by the start of sampling on 21 April whereas relatively high concentrations were measured at a depth of 15 and 20 m (0.7 and 1.9µM respectively). The sharp deepening of the nitracline down to 25 m was observed at the beginning of May, and a similar rise of the pycnocline, as it was registered in 2010, took place by 18 May. Due to this process, higher NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 concentrations were detected in the upper layer again in both years. The subsequent samplings in 2011 were conducted with an ∼2-week time lag, and probably the NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> <sup>3</sup> were depleted faster than seen from the interpolated field in **Figure 3**. In 2012, the deepening of the nitracline was observed from the beginning of May, and no significant rise in NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 concentrations was detected in the second half of May at station AP5; instead, the continuous deepening of the nitracline down to 25–30 m was observed. In 2014, the NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> <sup>3</sup> were depleted down to the 25 m depth and, like in 2012, no significant rise in nitracline depth was revealed after the spring bloom.

After the depletion of NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 , there was always some PO3<sup>−</sup> 4 left in the surface layer (**Figure 4**) and the concentrations of excess PO3<sup>−</sup> <sup>4</sup> were quite different in the studied years. In 2009 and 2012, the average concentrations were 0.13µM (in the range of 0.12–0.16µM) and 0.18µM (in the range of 0.08– 0.24µM), respectively. For the same period, the concentrations of PO3<sup>−</sup> <sup>4</sup> were on average 0.33µM (in the range of 0.20–0.42µM) and 0.37µM (in the range of 0.23–0.48µM) in 2010 and 2014 respectively, and 0.76µM (in the range of 0.44–0.97µM) in 2011. The observed late spring rise of the pycnocline increased the surface layer PO3<sup>−</sup> 4 concentrations in 2010 and 2011 remarkably (**Figures 3**, **4**).

## Consumption of Inorganic Nutrients in May–June

The PO3<sup>−</sup> <sup>4</sup> was depleted in the surface layer by 31 May in 2009, an increase in concentration in the surface layer was observed in mid-June, and the PO3<sup>−</sup> <sup>4</sup> was depleted again by the end of the month. In 2010, the sampling for PO3<sup>−</sup> 4 analysis was performed until 14 June, and there was still on average 0.16µM PO3<sup>−</sup> 4 left (range of 0.11–0.24µM) in the surface layer. In 2011, the sampling period for nutrient analysis was shorter compared with other years in this study, but by 1 June, there was still on average 0.42µM PO3<sup>−</sup> 4 left (range of 0.30–0.56µM) in the surface layer along the cross-gulf study transect. In 2012, the sampling of nutrients was conducted until 7 June, and by that time, the

phosphate concentrations in the surface layer were close to the detection limit. In 2014, nutrients were sampled until 25 June, and PO3<sup>−</sup> 4 levels were below detection limit by 10 June.

with vertical lines, and sampling depths are shown as white circles.

According to the availability of PO3<sup>−</sup> 4 in the surface layer after the depletion of NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 , the consumption of surplus PO3<sup>−</sup> 4 was estimated (**Table 3**). The consumption rates of PO3<sup>−</sup> <sup>4</sup> were estimated for the following periods: from 10 May until 31 May in 2009 (days 130–151), from 10 May until 14 June in 2010 (days 130–165), from 2 May until 1 June in 2011 (days 122–152), from 7 May until 7 June in 2012 (days 129–160), and from 4 May until 25 June in 2014 (days 125–146). The consumption of PO3<sup>−</sup> <sup>4</sup> within these periods was much lower compared with the consumption during the antedate phytoplankton spring bloom (data not presented). Relatively low consumption of PO3<sup>−</sup> 4 in the post-spring bloom period was found in 2009 and 2012, and the estimates were significantly higher in 2010, 2011, and 2014. The probable consumption/need of NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 , assuming that nutrients were consumed according to the N:P ratio of 16:1, was also estimated. Higher inorganic nitrogen needs to deplete surplus PO3<sup>−</sup> 4 from the surface layer were found in 2010, 2011, and 2014 compared with 2009 and 2012. The statistically significant relationship between the decrease of PO3<sup>−</sup> 4 and an increase of M. rubrum biomass in the surface layer was found (R = 0.70, p < 0.01, n = 27) for all springs except the year 2010.

).

TABLE 3 | Estimated consumption rates of PO3<sup>−</sup> 4 after the depletion of nitrites-nitrates in the surface layer (*n* = number of observations and *R* = correlation coefficient; for all series *<sup>p</sup>* <sup>&</sup>lt; 0.01), standard errors of the estimates seb, and calculated potential need for NO<sup>−</sup> 2 +NO<sup>−</sup> 3 to consume the observed excess PO3<sup>−</sup> 4 according to the N:P ratio of 16:1.


#### Interannual Changes in Mesodinium rubrum Biomass in May–June

The ciliate M. rubrum was an important primary producer in all studied years during the late phase or post-spring bloom (dominated by diatoms and dinoflagellates) period in the Gulf of Finland. In April, during the spring bloom peak, its average contribution to the photosynthetic plankton community was low, on average 8% (n = 213) in all studied years, being the lowest in 2011 when it stayed around 4% (n = 50). The average contribution of M. rubrum to the photosynthetic plankton community in the sea area between Tallinn and Helsinki in May was very variable in studied years. The lowest contribution in 2010 and 2011 (11–13% with n = 51 and n = 66, respectively) and the highest in 2009 and 2014 (53–61% with n = 81 and n = 44, respectively) were registered. The average contribution to the photosynthetic plankton biomass in May 2012 was 28% (n = 48). Even the M. rubrum biomass seemed modest in spring-summer 2010 its proportion in the overall phytoplankton community was similar to the year 2012—on average 32% if the period from the beginning of May until the end of June was taken into account.

The most significant differences between the years were in the timing of the maximum contribution of M. rubrum. The highest proportion in the photosynthetic plankton community measured in May was in 2009 (81%, n = 85), 2012 (72%, n = 48), and 2014 (86%, n = 44). The highest contributions in June were observed in 2010 (97%, n = 99), 2011 (91%, n = 27), and 2014 (86%, n = 33). In addition, the remarkable differences in biomass and distribution of M. rubrum were observed (**Figure 4**). The most intensive blooms, distributed quite evenly across the Gulf, were registered in 2009 and 2014 when very high biomass values were measured at the first half of May – on 10 and 17 May (91–457 and 106.6–510.1 µgC l−<sup>1</sup> , respectively) in 2009 and on 15 May (136– 439 µgC l−<sup>1</sup> ) in 2014. The year 2014 is characterized by a longer period with high biomass values of M. rubrum either along the entire cross-gulf transect or in the different parts of it (28 April until 4 June, **Figure 4**). High values of M. rubrum biomass were also detected in 2011 and 2012 but the cross-gulf distribution was patchy and the intensive bloom period shorter with biomasses over 300 µgC l−<sup>1</sup> only at one sampling date, on 22 May in 2011 (9.5–508 µgC l−<sup>1</sup> ) and on 1 May in 2012 (24.2–479.4 µgC l−<sup>1</sup> ). Remarkably lower M. rubrum biomass values were observed in May–June 2010 when the highest values were measured a month later, compared with the other studied years, on 14 June (76–276 µgC l−<sup>1</sup> ).

A significant relationship (R = 0.60, p < 0.01, n = 22) was found between the start (a week with a noticeable increase in biomass compared with the previous sampling) of the M. rubrum bloom and warming of the sea surface layer. The clear increase in M. rubrum biomass was observed after the cross-gulf average surface layer temperature had reached over 4◦C (**Figure 5**). The maximum M. rubrum biomass was higher and established earlier in warmer springs (2009 and 2014; **Table 2**). Also, the average cross-gulf biomass of M. rubrum in May was greater in the years characterized by earlier surface layer warming (2009–155 µgC l−<sup>1</sup> , 2012–119 µgC l−<sup>1</sup> , and 2014–193 µgC l−<sup>1</sup> ). Springs with slower surface layer warming or very dynamic temperature pattern were characterized by lower M. rubrum average biomass in May (2010–51 µgC l−<sup>1</sup> and 2011–62 µgC l−<sup>1</sup> ).

A moderate relationship was found with average cross-gulf surface layer salinity and M. rubrum biomass build up (R = 0.40, p < 0.05, n = 28).

A qualitative relationship between the vertical stratification and bloom outcome could be demonstrated based on the estimated stratification parameter at station AP5 (located at the distance of 22 km north from the southern end of the crossgulf study transect; **Figure 3**) and M. rubrum biomass in the area. The lowest biomass of M. rubrum in spring was observed in 2010 when also the vertical stratification was very strong at station AP5. The highest biomass was found in spring 2014, characterized with the weak vertical stratification, and the decline of M. rubrum biomass in the southern part of the study transect coincided with the strengthening of stratification at station AP5 in late May 2014. Nevertheless, similar outcomes of the bloom with occasional high biomass of M. rubrum were registered in 2011 and 2012, although the vertical stratification differed significantly between these years—the stratification parameter varied from 12.0 to 35.6 J m−<sup>3</sup> in 2011 and from 26.3 to 51.0 J m−<sup>3</sup> (72.1 J m−<sup>3</sup> in early July) in 2012.

Moderate but statistically not significant agreement of M. rubrum biomass in May with preceding period cryptophyte biomass was found (**Figure 6**). Still, the years with higher cryptophyte biomass in April showed the earlier establishment of the ciliate bloom, except in 2011 when the surface layer temperature was very variable in April. In addition, the decline of the M. rubrum bloom corresponds with the development of

Dinophysis acuminata (Claparède and Lachmann 1859) biomass increase in June (**Figure 6**; R = 0.51, p < 0.05, n = 23).

## Diel Vertical Dynamics in Mesodinium rubrum Abundance in Spring 2014

On 21–22 May 2014, the 24 h measurement and sampling campaign was performed at station AP5. By this time, the water temperature in the upper 5 m layer was above 7◦C (**Figure 7**). The surface layer salinity was about 5.45 g kg−<sup>1</sup> , and a clear vertical salinity gradient was observed in the sub-surface layer below 20 m depth. The NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> <sup>3</sup> was depleted down to 20 m depth, but there was still some PO3<sup>−</sup> 4 left in the surface layer (**Figure 7**).

Phytoplankton sampling at 2 h intervals and 5 m vertical resolution allowed following the vertical displacement of M. rubrum cells (**Figure 7**). Although the highest abundances were usually obtained from the 1 m depth, the clear increase in abundance in the sub-surface layers can be observed at night. Cells started to descend after 7 p.m. (local time) and maximum abundances at 25–30 m depth were registered in the early morning between 1 and 7 a.m. Most of the descending cells did not migrate deeper than 20 m depth, where the start of nitracline was located. By 11 a.m. next day, the cells were again mainly concentrated in the upper 15 m layer.

# DISCUSSION

The biomass of the photosynthetic ciliate M. rubrum had an evident influence on the primary production of plankton community and nutrient cycling. The inorganic nitrogen, accumulated in the upper layer in the Gulf of Finland during winter, is consumed on average by the beginning of May (**Figure 4**) by rapidly growing spring bloom diatoms and dinoflagellates (Lips et al., 2014). Due to the low N:P ratio

d−2012, e−2014). The number of the week is shown on the x-axis.

in the winter pool of nutrients and nutrients in the subsurface layer after the formation of stratification, vertical mixing, and/or advection introduces PO3<sup>−</sup> 4 into surface layer always in excess compared with nitrogen (Laanemets et al., 2011). The depletion of excess PO3<sup>−</sup> 4 in the surface layer before the increase of biomass of N-fixing cyanobacteria (at the end of June-beginning of July) was recognized in the present study.

The depletion of surplus PO3<sup>−</sup> <sup>4</sup> would need significant amounts of inorganic nitrogen which is depleted from the surface layer after the spring bloom. It is possible to roughly estimate the potential need for inorganic nitrogen (**Table 3**) and to predict biomass increase of photosynthetic plankton according to the Redfield ratio (C:N:P of 106:16:1) based on the available PO3<sup>−</sup> 4 in the surface layer and neglecting the remineralization process and consumption by bacteria. For example, the predicted increase in photosynthetic biomass in the surface layer for the period from 3 May to 17 May 2009 (after the depletion of NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 in the surface layer; days 122–136) could be ∼100 µgC l−<sup>1</sup> taking into account the average concentration of available PO3<sup>−</sup> 4 (0.08µM). The real measured biomass increase in photosynthetic plankton (including ciliate M. rubrum) was on average 42% higher. This discrepancy can be explained either by the fact that net community production may be underestimated if it is based on nutrient concentrations and Redfield ratios only because the contributions of recycled nutrients cannot be taken into account (Thomas et al., 1999) or there are other potential nutrient sources unnoticed when sampling only from the surface layer. In the present study, during the selected period, the shift to species able to migrate vertically in the stratified water column, took place. In fact, the main biomass increase was formed by phototrophic ciliate M. rubrum—the biomass increased more than two times from an average 95–220 µgC l−<sup>1</sup> in the study area. During the first week (3–10 May), the total biomass of phytoplankton increased only by 50 µgC l−<sup>1</sup> while the biomass of M. rubrum increased at the same time by 120 µgC l−<sup>1</sup> (the difference is due to the disappearance of spring bloom species from the community after the inorganic nitrogen depletion from the surface layer). Within the next week (10–17 May), the average increase in total biomass of phototrophic plankton was 130 µgC l−<sup>1</sup> , whereas, on average 100 µgC l−<sup>1</sup> was due to the increase of biomass of M. rubrum. The estimated consumption

rate of PO3<sup>−</sup> 4 in the surface layer would have supported the total photosynthetic biomass increase approximately by 65 and 40 µgC l−<sup>1</sup> only, respectively for these 2 weeks. Hence, the observed biomass increase in the whole period of 3–17 May was suggested to be mainly established due to the change in the photosynthetic plankton community composition and the ability of particular species to migrate to the sub-surface layers and assimilate nutrients (both nitrates and phosphates) necessary for their growth.

Still, there are other features and mechanisms to be considered. One should note the increase in NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 concentration in the surface layer by 17 May 2009 (day 136, **Figure 4**). When the measured decrease of PO3<sup>−</sup> 4 in the surface layer for the period of 10–17 May was rather low (0.03µM), the increase in NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 concentration at the same time was remarkable (0.34µM) and coincided with the highest M. rubrum biomass values measured in spring 2009 in the study area. If to assume that the increase of inorganic nitrogen concentration in

white circles.

the surface layer was induced by the rise of pycnocline, as it was observed in May 2010 and 2011 (**Figure 3**; no vertical profiles available for spring 2009), there should have also been observed the increase in PO3<sup>−</sup> 4 concentration in the surface layer. Instead, the slight decrease in PO3<sup>−</sup> 4 concentration coincided with the increase in NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 concentration and M. rubrum biomass. The first major increase in M. rubrum biomass (from 3 to 10 May) was proposed to be mainly related to the species ability to migrate to the lower layers and exploit nutrient reserves there to be able to multiply later in the surface layer. The N-depleted surface layer together with significant biomass increase support this assumption. The second biomass increase (from 10 to 17 May) of M. rubrum is most probably a combination of previous and following growth, and advection of water masses with lower salinity into the study area (**Figure 2**), which probably had higher inorganic nitrogen concentration. Hence, high-resolution measurements and sampling enable to see a more comprehensive picture.

Similar calculations were made for May-June 2014, a period when a decrease in PO3<sup>−</sup> 4 concentration in the surface layer during M. rubrum bloom was observed. The predicted average biomass increase of primary producers for a period from 22 April to 15 May (∼100 µgC l−<sup>1</sup> ) was also lower than the average real outcome (155 µgC l−<sup>1</sup> ). It is important to note, that the increase of M. rubrum biomass was as high as 4.7 times during the considered period, increasing its contribution to the overall photosynthetic plankton biomass from 15 to 56%. These noticed simultaneous dynamics of nutrients and phototrophic plankton biomass suggest the assimilation of inorganic nutrients (both – nitrates and phosphates) in the sub-surface layers and their transport to the surface layer by vertically migrating M. rubrum.

The year 2010 was characterized with the lowest maximum biomass of M. rubrum. Calculations showed that in the period from 31 May to 14 June, the decrease of PO3<sup>−</sup> 4 by 0.11µM could have supported the biomass increase according to the Redfield ratio approximately by 140 µgC l−<sup>1</sup> , which was slightly lower but still quite close to the measured average biomass increase for this period (159 µgC l−<sup>1</sup> ). For the same period, the average increase in M. rubrum biomass was 100 µgC l−<sup>1</sup> , and the contribution of this species to the total phototrophic plankton biomass was increasing within 2 weeks from 50 to 85%. The biomass increase seemed to be mostly based on the PO3<sup>−</sup> 4 left in the surface layer and assimilation of sub-surface NO<sup>−</sup> <sup>2</sup> <sup>+</sup>NO<sup>−</sup> 3 after the rise of nitracline at the second half of May (**Figure 3**).

These kind of calculations, without taking into account all possible sources and sinks of inorganic nutrients, are very rough. In addition, dynamic mesoscale features on the background of meteorologically forced transport and mixing (as intensive horizontal flows of water masses with different salinity and/or nutrient concentration to or through the study area) make these simplified calculations/assessments complicated. At the same time, the decrease in the surface layer PO3<sup>−</sup> 4 concentration in spring is very often significantly associated with the M. rubrum biomass increase and dominance in the community. Hence, the contribution of this species to the dynamics of inorganic nutrients cannot be neglected. The biomass of photosynthetic plankton in the second half of May was dominated by M. rubrum (70%) in all studied years. As the main increase in total photosynthetic plankton biomass was due to the growth of this species, the significant amounts of available inorganic nutrients (both in horizontal and vertical scale) were consumed most probably by it.

The spring bloom is predominantly regarded as a new production according to the definition of Dugdale and Goering (1967). After the nitrate depletion and decline of the spring bloom, the primary production in the Baltic Sea is mostly assumed to be based on the availability of regenerated nutrients (e.g., Kivi et al., 1993). Earlier studies (Jimenez and Intriago, 1987; Lindholm and Mörk, 1990and references therein, Crawford and Lindholm, 1997; Lips and Lips, 2014) are supporting the results of the present study about the importance of vertical migration to the nutrient dynamics and autotrophic growth in periods characterized by inorganic nutrient limitation in the surface layer. The remarkable growth of M. rubrum and formation of red tides in different seasons are based on new nutrients introduced to the surface layer either by physical processes (rise of the thermocline/pycnocline, advection of surface layer water masses, upwelling) or biological capabilities (vertical migration through pycnoclines). The dominance of photosynthetic ciliate M. rubrum after the spring bloom not only increases the retention time of newly produced material in the nutrientlimited euphotic layer (Lips et al., 2014) but its contribution to the overall photosynthetic community and primary production can be outstanding. Leppänen and Bruun (1986) estimated that Mesodinium contributed about 10% of primary production in spring (April-May) in the open Northern Baltic Proper having at the same time on average only 2% of the total biomass of primary producers. Three decades later the overall contribution of this photosynthetic ciliate to the spring primary producers biomass seems to be increased significantly allowing the estimation on average 8–38% of total primary producers for the same period. Hence, 4–19 times higher contribution to the spring primary production can be expected, and the even greater contribution due to the overall climate change and an increase in sea surface temperature can be foreseen.

The very dynamic nature of M. rubrum blooms in spring would make difficult predictions of impact to the other trophic levels without knowledge of this species ecological preferences. Considering the regulating factors in dynamics of plankton community, limiting factors affecting population growth directly and controlling factors influencing the outcome of growth processes exist (Thingstad and Sakshaug, 1990). Johnson et al. (2013) have shown a significant positive correlation of M. rubrum field population with temperature in spring. Also, Montagnes et al. (2008) demonstrated a significant effect of temperature on M. rubrum abundance; the one-degree increase would increase the mean abundance by 1.42, and explained it with a decreased growth rate at low temperature. The limitation of cellular metabolic capacity by the thermal stress (Moeller et al., 2011) and a decrease in swimming velocity with decreasing temperatures (Riisgård and Larsen, 2009) have also been suggested. The motion of M. rubrum is characterized by jumping after shorter or longer periods of motionless, but it is also capable of sustained swimming (Fenchel and Hansen, 2006). Besides escape from predators, the jumping is necessary to create fluid motion surrounding the ciliate to increase the contribution of advective transport in nutrient uptake (Jiang, 2011). Slow warming (**Figure 2**, **Table 2**) in the study area at the end of April in 2010 probably created a physical environment not supporting the intensive growth and active motions, including the vertical migrations to assimilate nitrates from deeper layers, of M. rubrum compared with other studied years. Hence, when migratory capabilities were reduced, the competitive advantages over other migrating photosynthetic plankton were downgraded. The other important factor in combination with the low surface water temperature to influence the growth of M. rubrum is the low position of nitracline (below 20 m already by the end of April) in 2010 (**Figure 3**). Although the ability of M. rubrum to migrate through the density gradients is well-documented (e.g., Figueroa et al., 1998), the present results suggest that vertical stratification could notably influence the growth and bloom outcome of this species. The latter is also supported by the data from spring 2014. The observed clear decrease of M. rubrum biomass in late May 2014 in the southern part of the transect (**Figure 4**) coincided with the strengthening of vertical stratification (**Figure 3**).

M. rubrum is an obligate phototroph obtaining most of its carbon from photosynthesis, and only one cryptophyte prey per cell is required to maintain its maximum growth (Hansen and Fenchel, 2006). The suitable prey and predator relationships are hypothesized to support the M. rubrum surface bloom formation (Stoecker et al., 2009). Cryptophytes-M. rubrum relationships can also be followed in the data set of studied years with significant bloom maximum in 2009, 2011–2012, and 2014 (**Figure 6**), but the same does not hold for the year 2010 when the initial biomass of cryptophytes was several times lower in April compared with other years. Most probably, the combination of regulating factors (both limiting and controlling) influenced the M. rubrum biomass outcome in 2010.

Complex migratory patterns observed in the present study and described by others (e.g., review by (Crawford, 1989) might be related to a combination of requirements for light, cryptophyte prey, and nutrients, especially nitrates. The migrations in response to nutrient and light conditions in the stratified sea can lead to a vertical distribution in which the majority of the population may be concentrated close to the surface during the day and in deeper layers at night. Most commonly, only part of the populations are performing such daily migrations (Pérez et al., 1999; Rychert, 2004), but getting a more realistic picture of the extent of such migrations temporally very high-resolution measurements are needed. The suggested vertical migrations of M. rubrum between the surface layer and the nitracline should create incidents when during the high biomass period of this ciliate in the plankton community the Chl a concentrations could occasionally be elevated in the sub-surface layer. Subsurface maxima have been reported earlier in the Gulf of Finland in summer in the cases when the dinoflagellate Heterocapsa triquetra (Ehrenberg) Stein 1883 was present in the community in high abundances (Lips et al., 2011; Lips and Lips, 2014). The high-resolution measurements with a profiling mooring and a towed undulating vehicle (data not presented here) registered the sub-surface Chl a maxima in spring 2012. During the period of the high biomass of M. rubrum in the first half of May, thin layers of relatively high Chl a values were observed on several occasions. In the frame of the present study, the vertical sampling conducted at 2 h intervals and resolution of 5 m might have been too low as M. rubrum could have theoretically made several migrations within 2 h or was missed in the depths not sampled. Still, the integrated biomass values increased slightly from 21 p.m. until 1 a.m. (240–380 µC l−<sup>1</sup> ) and decreased again afterwards, indicating the high probability of success in sampling during the active migration. The high swimming speed allows M. rubrum to descend to the nitracline in the evening, stay there to assimilate nutrients and migrate back to the well-lit surface layer by midday. From the presented 24-h study, the observed migration pattern allows suggesting the diurnal vertical migrations of this species. Still, extensive studies should be made in the future to see the longer pattern and regularity of such migrations and assess more precisely the influence not only to the horizontal but also to the vertical distribution of inorganic nutrients. The ability to migrate vertically and exploit the nutrient pools from the lower layers may significantly influence the nutricline depth after the development of stratification in spring. Interestingly, even the nitracline was located at 20 m depth in May 2014, part of the M. rubrum cells migrated deeper—down to 25 and 30 m depth.

The photosynthetic ciliate M. rubrum might be a key player in the trophic transfer of energy after the decline of spring bloom and establishment of late summer phytoplankton blooms. This phototrophic ciliate not only prolongs the autotrophic production in the nutrient-depleted surface layer but also acts as an important food supply to other organisms (e.g., Park et al., 2006; Fileman et al., 2007; Lee et al., 2014; **Figure 6**). Also, the excretion of nutrients through mineralization and cell explosion can be a significant source of nitrogen (Lindholm, 1985; Miller et al., 1995) to phytoplankton species present in the surface layer community. High nitrate and phosphate assimilation rates reported in previous studies (Dugdale et al., 1987; Jiang, 2011; Tong et al., 2015), support the assumption that inorganic nitrogen available in spring-summer continuum, either brought close to the surface through pycnocline rise or from adjacent areas, will be mostly assimilated by dominating M. rubrum if the other environmental conditions support its growth. Also, the dominance of M. rubrum in May-June, its migration behavior, and phosphate utilization in the surface layer is strongly influencing the amount of excess PO3<sup>−</sup> 4 that is usually regarded to support the summer cyanobacterial bloom development (e.g., Janssen et al., 2004; Laanemets et al., 2006; Raateoja et al., 2011). The years, when PO3<sup>−</sup> <sup>4</sup> was depleted or close to the depletion by the end of our study period, are characterized by lower cyanobacterial biomass development and vice versa (Kahru and Elmgren, 2014). Hence, the dynamics and intensity of M. rubrum blooms in May-June have, besides nutrient distribution, the significant impact also on the late summer phytoplankton communities. The interactions between M. rubrum biomass development and other phytoplankton groups should be studied further as its contribution to the total photosynthetic biomass has increased in all seasons in the Baltic Sea (Jaanus et al., 2011). Thamm et al. (2004) have demonstrated the shift of the peak occurrence of this species from spring (in 1987) over spring/summer (in 1990) to summer (in 1997).

# CONCLUSIONS

The clear relationship between the start and outcome of the M. rubrum bloom and average cross-gulf surface layer temperature emphasize the potentially high impact of this species to the spring–early summer plankton community in the background of overall climate change and continuous increase in sea surface temperature. The very high proportion of M. rubrum in the phototrophic plankton community has created the shift from, previously acknowledged, regenerated production toward new production at the period between spring bloom and summer cyanobacterial bloom in the Baltic Sea. The dominance of M. rubrum after the spring bloom in vertically stratified Gulf of Finland strongly influences the amount of excess phosphates in the surface layer and vertical inorganic nutrient dynamics. Within present study, comprising sampling with high temporal and spatial resolution, the understanding of dynamics and possible impacts of spring M. rubrum blooms to the Gulf of Finland ecosystem was increased.

## REFERENCES


# AUTHOR CONTRIBUTIONS

UL organized and partly conducted the measurements and sampling in the study area. IL conducted all phytoplankton and partly the nutrient analysis. Both authors contributed to the data analysis and manuscript writing.

#### FUNDING

The present work was supported by institutional research funding IUT (IUT19-6) of the Estonian Ministry of Education and Research, by the Estonian Science Foundation grant no. 8930 and by EU Regional Development Foundation, Environmental Conservation and Environmental Technology R&D Programme project VeeOBS (3.2.0802.11-0043).

## ACKNOWLEDGMENTS

We appreciate the help of crew and colleagues during sampling and analysis.


**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 Lips and Lips. 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.

# Ecosystem Effects of Morphological and Life History Traits in Two Divergent Zooplankton Populations

Konrad Karlsson\* and Monika Winder

Department of Ecology Environment and Plant Sciences, Stockholm University, Stockholm, Sweden

Little is known about the ecosystem effects of locally adapted populations. The filter feeding copepod Eurytemora affinis is an abundant and important zooplankton in coastal waters that consist of a cryptic species complex with locally adapted populations. We used a mesocosm setup to investigate population and ecosystem interactions of two populations from the Baltic Sea with different morphology and life history traits. One population is laterally wider, larger-sized, more fecund, and have higher growth rate than the other. The experimental ecosystems varied in algae community (pelagic algae, and pelagic algae + benthic diatoms) with two resource supply scenarios. Results showed that the large-sized population is a more effective grazer. In low resource supply the small-sized population starved, whereas the large-sized population was unaffected, resulting in a larger population increase of both nauplii and copepodites than for the small-sized population. Addition of benthic diatoms to the pelagic algae community had much more negative effects on the reproduction of the large-sized population. This suggests that the large-sized population feeds near benthic to a greater extent than the small-sized population, and that filamentous benthic diatoms interfere with the grazing process. Despite the negative effects of benthic diatoms, the large-sized population could maintain similar or higher reproduction than the small-sized population. In addition, the high grazing efficiency of the large-sized population resulted in a different community composition of algae. Specifically, flagellated species and small sized benthic diatoms were more grazed upon by the large-sized population. Our results show that morphologically divergent, yet phylogenetically closely related zooplankton populations can have different ecosystem functions, and in turn have different population increase in response to resource supply and algae community.

Keywords: local adaptation, common gardening experiment, intraspecific variation, ecological-evolutionary dynamics, resource specialization, morphological divergence, niche partitioning, size efficiency

# 1. INTRODUCTION

The number of species in an ecosystem and their traits affect diverse ecosystem processes (Hooper et al., 2005). A key question is whether the species or their traits are the best predictor of ecosystem functions because variation within species, such as differences in resource specialization and life history traits (e.g., growth and fecundity) can affect ecosystem processes (Harmon et al., 2009; Bassar et al., 2010; Walsh et al., 2012). Consequently, variation in ecosystem processes can be larger

# *Edited by:*

Riina Klais, University of Tartu, Estonia

#### *Reviewed by:*

Sami Souissi, Lille University of Science and Technology, France Ana Marta Gonçalves, University of Coimbra, Portugal

> *\*Correspondence:* Konrad Karlsson konrad.karlsson@su.se

#### *Specialty section:*

This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science

*Received:* 29 April 2018 *Accepted:* 15 October 2018 *Published:* 13 November 2018

#### *Citation:*

Karlsson K and Winder M (2018) Ecosystem Effects of Morphological and Life History Traits in Two Divergent Zooplankton Populations. Front. Mar. Sci. 5:408. doi: 10.3389/fmars.2018.00408

**153**

within species than between species (Gianuca et al., 2016) and not all conspecifics can be regarded as ecologically equal (Bolnick et al., 2002). Furthermore, interactions of species and ecosystems are bilateral so that species may diverge depending on the type of habitat (Marklund et al., 2018).

Many species exhibits adaptations dependent on attributes in their local habitat. Differences between locally adapted populations can be quantified by rearing populations in a common garden where the environments are identical for all individuals (Falconer and Mackay, 1996), thereby the environmental source of phenotypic variance between populations can be eliminated. Hence the phenotypic variance of a quantified trait is caused by genetic variance when the environment is identical. In addition, locally adapted populations may have different ecological effects (Schoener, 2011), which can be estimated by common gardening experiments. The difference between a common garden and a common gardening experiment is the switch of focal point. In the former the focus is on the environmental effects on phenotypes. In the latter the focus is on the effect of phenotypes on the environment. Therefore, there are two focal points for delineating evolutionary diversification: the effect of ecology on phenotypes, and the effect of phenotypes on ecology. By constructing common gardening experiments, where one put specific phenotypes (e.g., populations) in replicated ecosystems, it is possible to quantify both of these effects (Matthews et al., 2011b, 2014).

Overlooking divergence within species results in loss of information about how organisms interact with and shape their ecosystems. A common procedure when conceptualizing ecosystems is to use a trait-based approach (Litchman et al., 2013; Colina et al., 2016). However, to describe trait diversity correctly one has to acknowledge that traits diverge also within species. For instance, in a fish species when averaging consumption of benthic and pelagic resources it seems as if fish are connectors of the pelagic and benthic food webs (Schindler and Scheuerell, 2002; Vander Zanden and Vadeboncoeur, 2002). However, when accounting for that fish can specialize for either benthic or pelagic resources it seems as if they disconnect these two food webs (Quevedo et al., 2009). In fish, there are many examples of how adaptive radiations and plastic specializations in resource use have caused intra-specific morphological variation, which results in different ecosystem effects by the different morphotypes (Harmon et al., 2009; Palkovacs and Post, 2009; Post and Palkovacs, 2009; Lundsgaard-Hansen et al., 2014). In addition to morphological traits, life history traits such as divergence in populations' growth rates and fecundity can affect ecosystem processes differently (Bassar et al., 2010; Walsh et al., 2012). Furthermore, traits that diverge under artificial selection can have different effects in experimental ecosystems (Becks et al., 2010; Pantel et al., 2015), giving a direct link from adaptation to ecosystem effects.

The focal species in the present study is the calanoid copepod Eurytemora affinis (Poppe, 1880), a dominant zooplankton species in estuaries in the northern hemisphere and an important grazer and prey for fish (Hernroth and Ackefors, 1979; Diekmann et al., 2012; Rajasilta et al., 2014). It consists of a cryptic species complex (Lee and Frost, 2002) and inter-population crosses have shown that certain populations cannot reproduce further than two generations (Lee, 2000). Specifically, in the Baltic Sea the great morphological variation in E. affinis have led to past taxonomic confusion and classification into invalid species (Lee and Frost, 2002; Sukhikh et al., 2016) because the females of some populations can be larger, laterally wider, and carry more eggs than others do (Gurney, 1931).

Revisions of the E. affinis species complex have been ongoing since it was first described by Poppe (1880) and Gurney (1931). Recently, a new species within the complex, Eurytemora carolleeae, was described (Alekseev and Souissi, 2011). This species originate from the North American east coast and is listed as an invasive species in the Baltic Sea, with occurrences in the Gulf of Finland and the Gulf of Riga (Alekseev and Souissi, 2011; Sukhikh et al., 2013). The morphological traits used to discriminate between E. affinis and E. carolleeae are nonadaptive (Alekseev and Souissi, 2011; Sukhikh et al., 2013; Lajus et al., 2015), and thus unlikely to have any ecosystem effects (Matthews et al., 2011a). Furthermore, taxonomic classification based on morphology alone is not distinct because some key traits overlap between the two species (Sukhikh et al., 2013; Vasquez et al., 2016) and differentiate between E. affinis populations (Sukhikh et al., 2016). Therefore, we use the E. affinis species name throughout and refer to the E. affinis species complex in the present study.

A previous common garden experiment revealed differences in development time as a response to temperature between Baltic Sea E. affinis populations, where a population from the Gulf of Riga (Pärnu Bay) had similar (12 and 17◦C) or shorter (22.5◦C) development time than a population from the Swedish coast (Stockholm Archipelago) (Karlsson and Winder, unpublished data). Females from the Gulf of Riga population also appeared to have a larger body size than the Swedish coast population, which was however not quantified. For copepods in general, development time increase with species size and fecundity (Allan, 1976; Gillooly, 2000), thus this large-sized population breaks this norm. The trait differences between the two populations' may affect ecosystem processes, such as differences in prey size spectra, prey selection, and grazing efficiency (Bassar et al., 2010; Walsh et al., 2012; Gianuca et al., 2016).

The aim of this study is to (i) describe and quantify morphology and fecundity in a common garden experiment and (ii) quantify ecosystem responses and reciprocal effects in a common gardening experiment (mesocosms) of these two E. affinis populations from the Baltic Sea. We constructed mesocosm environments with two types of algae (prey) community and two types of resource supplies. To investigate if the described trait differences affect feeding efficiency and algae community, and the reciprocal effects of algae community and resource supply on population growth, female prosome length, and clutch size. Algae community included two treatments one with pelagic species and one with pelagic species plus benthic diatoms, resource supply included two levels of nutrients. The mesocosm experiment lasted for 14 days, which corresponds to approximately one generation time for both populations at the experimental temperature of 17◦C (Karlsson and Winder, unpublished data). We hypothesized that populations would differentiate over the algae treatment if they diverged in their feeding behavior and resource specialization, and differentiate over the resource supply treatment if they diverge in their demand for food and feeding efficiency.

# 2. MATERIALS AND METHODS

#### 2.1. Sampling and Rearing of Cultures

No specific ethical permits were needed for research on invertebrate crustaceans. This study does not include vulnerable populations and endangered animal species. Populations were sampled in spring 2014 in the Stockholm archipelago at Askö monitoring station B1 (hereafter referred to as STHLM) 58◦ 48.19′ ,17◦ 37.52′ (latitude, longitude) and in summer 2014 in the Pärnu Bay a shallow inner part of Gulf of Riga (hereafter referred to as GOR) 58◦ 21.67′ , 24◦ 30.83' by vertical tow nets. More than 300 adult copepods were taken from the samples and used to start stock cultures. Before measurements of morphology and clutch size (number of eggs), stock cultures were kept for a minimum of three generations to reduce imprints of maternal effects (Sanford and Kelly, 2010) in 3 L (Exo TerraTM) aquariums at 17◦C and 7 PSU and fed with the cryptophyte Rhodomonas salina. Aquariums were aeriated by gently stirring the water once a day. Before the mesocosm experiment, cultures were kept for at least three generations in 45 L plastic tubs at 15◦C and 7 PSU and fed with the cryptophyte Rhinomonas nottbecki. The two experiments were separated in time by 2 years, from 2014 to 2016 and during this time we predict that approximately 42 generations have past, given a generation time of 17 days at 15◦C (Karlsson and Winder, unpublished data).

The zooplankton were at all times kept in tap water that had been circulated for about a week in an aquarium with gravel from a freshwater stream to condition the water for aquatic organisms (reduce chlorine and excessive gas); to adjust the right salinity we used Instant OceanTM sea salt.

### 2.2. Common Garden Experiment: Measurements of Morphology and Clutch Size

In autumn 2014 we started up 5 cultures per population by taking 10 egg carrying females from our stock cultures into 3 L aquariums with GF/F WhatmanTM filtered aquarium water of 2 PSU. The copepods were fed ad libitum of the cryptophyte Rhodomonas salina (100,000–200,000 cells ml−<sup>1</sup> ). More than 27,000 cells ml−<sup>1</sup> did not increase egg production (a proxy for growth in adult female copepods) in Acartia tonsa (Kiørboe et al., 1985) a copepod of similar size as E. affinis. As soon as the daughters of the inoculated females had developed egg sacks they were picked out to be photographed dorsal or ventral side, measured (prosome length), and the number of eggs in each female clutch were counted. We sampled 50 females per population, 10 from each culture.

We measured shape and size of the females' prosomen with 14 landmark data points from the digitized images. To visualize shape of the two populations we ran a principal component analysis with the landmark data (14 x and 14 y coordinates, in pixel units), made by R package geomorph (Adams and Otarola-Castillo, 2013), and depicted the individuals of each population that had minimum and maximum values on principal component 1 (**Figures 1**, **2**). We computed landmarks with the software packages tpsUtil and tpsDig2 (Rohlf, 2011b, 2013). We computed partial warps, uniform variables, and centroids with tpsRegr (Rohlf, 2011a). To test if shape differed between populations we used a MANCOVA with partial warps (11 x and 11 y) and uniform variables (1 x and 1 y) as response, size (centroid of each specimen) was included as covariate, and population as explanatory variable. The centroid size is calculated in pixel units as the square root of the sum of squared landmark distances to the centroid, which is a two-dimensional measure of size (1 pixel correspond to ca 0.89 µm). The interaction of population and size was not significant, thereby we concluded that the size and shape slopes of both populations do not differ, and therefore removed the interaction and use size as covariate (Engqvist, 2005). We used a two-way ANOVA to test the interaction of prosome length and population on clutch size. Furthermore, we followed up this analysis with a oneway ANOVA to test for differences in prosome length between populations.

### 2.3. Common Gardening Experiment: Set Up and Treatments

For the common gardening experiment, we used three types of factors each with two levels of treatment: two zooplankton

Karlsson and Winder Pelagic and Benthic Zooplankton Morphotypes

populations (STHLM and GOR), two nutrient treatments (low and high), and two types of algae community (pelagic and pelagic + benthic diatoms). We replicated each factor combination 10 times, summing up to 80 mesocosms (2 × 2 × 2 × 10). For control treatments we used factor combinations of nutrient levels and algae community without zooplankton, replicated 5 times, summing up to 20 control mesocosms (2 × 2 × 5).

The mesocosms were 3 L plastic aquariums (Exo TerraTM) filled with GF/F WhatmanTM filtered aquarium water, placed in a rooftop greenhouse at the Department of Ecology, Environment and Plant Sciences at Stockholm University. The mesocosms received natural light as well as light from metal halide lamps in 12:12 h night:day cycle, with an average 55.8 (range 16.7, 202.1) µmol m−<sup>2</sup> s <sup>−</sup><sup>1</sup> on the bottom of the mesocoms. Average temperature during the experiment was 17.0◦C, but oscillated during the night day cycle and was highest just before lamps turned off (average 18.8◦C, range 16.8–20.6◦C) and lowest just before lights turned on (average 14.9◦C, range 13.6–15.9◦C). We added nutrients and Instant OceanTM sea salt while the water was separated into two containers one for low and one for high resource supply treatment, water was later portioned out to the mesocosm after thorough mixing. We added 57.6 µmol L−<sup>1</sup> SiO4, 28.4 µmol L−<sup>1</sup> of NO3, and 2.4 µmoles L−<sup>1</sup> PO<sup>4</sup> to the high resource supply treatment, for the low nutrient treatment the respective concentrations were 30.6, 12.9, and 1.1 µmol L−<sup>1</sup> . Ratios of Si:N:P were similar for the high and low resource supply treatment with 24:12:1 and 28:12:1, respectively. We further added micronutrients, B vitamins, and peat extract (**Table S1**). Concentrations of NO<sup>2</sup> + NO<sup>3</sup> and SiO<sup>4</sup> (**Figure S1**) were higher than deliberately, suggesting that additional nutrients came into the mesocosm from either the algae or zooplankton cultures. Either way, the contrasts of resource supply treatments were clearly visible and quantifiable. We added salt to reach a salinity of 10 PSU, which was a salinity where both zooplankton and algae could coexist.

The two communities of algae consisted of either five pelagic species or the same pelagic species plus an addition of three benthic species (**Table S2**). The pelagic species were: the pooled sample of R. nottbecki and R. salina referred to as cryptophytes, Heterocapsa triquetra, Pseudoscourfeldia marina, Skeletonema marinoi, and Isochrysis sp. The benthic species where the pooled sample of Nitzschia aurariae and Navicula perminuta, Melosira sp., and Fragilaria sp. We selected algae species that belonged to some of the commonly found taxonomic classes in the Baltic Sea, with the compromise that they should be readily cultured in lab conditions on the same growth media (**Table S1**) and at the same salinity (10 PSU). We added the pelagic species in equal amounts to all mesocosms and the benthic species in equal amounts to half of the mesocosms. Therefore, at the beginning of the experiment the benthic treatment contained more food than the pelagic, but over time the amount of food was constrained to the concentration of nutrients. We inoculated the mesocosms by adding 1 ml from the same homogenized culture to each mesocosm for each species of algae. The inoculation concentrations were measured on the start day (October 4th) from Lugol-preserved samples diluted 10 times and placed in a tubular plankton chamber. We then

on principal component 1 from the respective population, these individuals are marked out in Figure 1. Pictures include locations of the 14 landmarks used in the principal component analysis. The prosome lengths of the individuals depicted are 833 and 776µm for STHLM min and max, and 833 and 923µm for GOR min and max, respectively. Note that the pictures are meant to show variation in shape and not in size and length. The scale of the pictures are 2,300µm wide and 1,725µm high.

took pictures by inverted light microscopy at specific locations in the chamber and calculated the number of cells per picture by EBImage R package (Pau et al., 2010), these numbers were calculated back to cells ml−<sup>1</sup> algae ml−<sup>1</sup> algae culture (**Table S2**).

One day after mesocosm setup, we added zooplankton from the stock cultures that consisted of similar ratios of nauplii, copepodites, and adults for both populations. Zooplankton were added by taking aliquots of the thoroughly mixed stock cultures to each mesocosm. We added slightly less individuals from the GOR population since these individuals are larger to get similar zooplankton bio-volumes for both populations. Aliquots (n =10) of the starting concentrations for each population were put in 4% formalin to be counted and measured. For these samples we counted the number of nauplii in stage 1–3 and 4–6 and copepodites in stage 1–3 and 4–6 (where stage 6 are the adults), and made measurements of length, width, and depth of the prosome body for all individual life stages. We then calculated the average bio-volume for an individual within one of these four groups by using the volume of an ellipse and multiplied the bio-volume with the number of individuals in the corresponding life-stage interval. Subsequently, we used the bio-volume sum in each sample to test if initial volumes were different for the two populations by a t-test.

## 2.4. Common Gardening Experiment: Sampling and Sample Analysis

We took two types of algae samples, one to measure relative fluorescence units (RFU) and one to count and identify the algae. We took samples directly after the mesocosms were gently stirred. RFU samples were measured in a TrilogyTM fluorometer (Turner design), samples for counts were preserved in acidic Lugol, and were counted with an inverted light microscope.

We counted cryptophytes, H. triquetra, Fragilaria sp., and Melosira sp. in continuous transects (field of view) over the full diameter of the chamber at 100 times magnification. We counted N. aurariae and N. perminuta in a continuous transect in 400 times magnification, and the smallest species P. marina, Isochrysis sp., and S. marinoi in 20 fields of view in a transect at 1,000 times magnification. We took samples of both RFU and Lugol on day 2, 6, 10, and 14. For RFU we measured all 100 mesocoms at each sampling occasion. Samples for counts of the pelagic species were 5, 2, 5, and 2 samples per treatment combination (including controls) and respective day. Samples for counts of the benthic species were 5, 2, 5, and 2 per treatment combination of the controls; and 5, 4, 7, and 2 per treatment combination with zooplankton.

At the same days as algae sampling we took two samples per treatment combination of free phosphate (PO4), nitrite + nitrate (NO<sup>2</sup> + NO3), ammonia (NH4), and silicate (SiO4) by filtering 10 ml water through a 0.45 µm filter and then analyzing the filtrate in a segmented flow analysis (**Figure S1**).

At the end of the experiment on day 15 (14 days after copepod innoculation), which corresponds to the development time for one generation (given ad libitum food conditions and 17◦C, Karlsson and Winder, unpublished data), the mesocosms were poured through a 45 µm net and all inhabiting zooplankton were filtered out and put in 4% formalin. From these samples we counted the number of nauplii and the pooled number of copepodites and adults. Two egg-carrying females from each sample (mesocosm) were picked at random and their prosome lengths were measured and clutch size counted. We expected the nauplii to have been born in the mesocosm during the experiment, and that the copepodites and adults mainly to be part of the individuals inoculated at the beginning.

#### 2.5. Statistical Analyses of the Common Gardening Experiment

We used R for all analyzes (R Core Team, 2017). The response variables: counts of nauplii (N1-N6) and copepodites + adults (C1-C6), prosome length of females, and female clutch size, were fitted as Gaussian response models by functions lme or gls from the nlme (Pinheiro et al., 2017) package. When there was more than one observation per mesocosm we used mesocosm ID as random effect in the linear mixed effect models (lme), and when variances in the different treatments were unequal we fitted extended linear models with power covariates in the linear models (gls). Models were evaluated by starting with a full model containing all treatment interactions (population, algae, and nutrients) and subsequently removing non-significant interactions, based on AIC. We only included the highest order of significant interactions or main effects in the running text, full model outputs can be found in the **Supplementary Material**.

RFU and species counts were analyzed by extended linear mixed models (function lme), with power covariates to control for the increasing residual variation due to algal growth. For both RFU and species counts we omitted the control to test for the difference between populations. The pelagic species occurred both with and without benthic diatoms, whereas the benthic diatoms only occurred with pelagic species, and hence the main effect of the addition of benthic diatoms was only estimated for the pelagic species. The RFU analysis was split up between pelagic and pelagic + benthic algae treatments as we expected the emitted fluorescence to differ between pelagic flagellates and benthic diatoms, and therefore a direct comparison could be spurious. The effects of resource supply were estimated for both RFU and species counts. We included a third degree polynomial of time as an interaction with the treatments (population, algae, resource supply) in both the RFU and the species counts models. This tests if the polynomial relationship with time differs between treatment contrasts. However, we did not include the estimates of the treatment main effects (i.e., time = 0), and the main effect of time in the results as it is not the focus of this study.

# 3. RESULTS

#### 3.1. Common Garden Experiment: Morphology and Clutch Size

The shape of the females prosomen differed significantly between the populations [MANCOVA, F(24, 74) = 5.19, p < 0.001] the GOR population was wider and rounder than the more laterally compressed STHLM population (**Figures 1**, **2**) and the size of the prosomen had no significant effect on shape [MANCOVA, F(24, 74) = 1.51, p = 0.090].

We found a significant interaction of population and prosome length on clutch size [two-way ANOVA, F(1, 96) = 6.68,

p = 0.011]. Hence, the relationship between prosome length and clutch size differs between the two populations (**Figure 3**), and a 1µm increase in length increased clutch size with 0.25 and 0.09 eggs for the GOR and STHLM populations respectively. Furthermore, the main effects of population [two-way ANOVA, F(1, 96) = 48.06, p < 0.001] and prosome length [two-way ANOVA, F(1, 96) = 25.64, p < 0.001] were significant, the average clutch sizes were 34.6 (28.1, 41.0; 95% CI) for the STHLM population and 54.4 (49.8, 58.9) eggs female−<sup>1</sup> for the GOR population. Furthermore, the females' prosome length differed between populations [one-way ANOVA, F(1, 98) = 124, p < 0.001] and was on average 762.1 (743.1, 781.1; 95% CI) for the STHLM population and 869.8 (856.4, 883.3)µm for the GOR population.

#### 3.2. Common Gardening Experiment

The averages of zooplankton bio-volumes at the start of the experiment were not significantly different between the populations [t-test, t(18) = -0.541, p = 0.595] and was 3.08 (± 0.24, SE) for the STHLM population and 2.91 (± 0.22) mm-3 for the GOR population. Number of individuals per stage at the start of the experiment (averages rounded to whole numbers) were for the STHLM population, N1-N3: 105 (86–116, range), N4-N6: 10 (3–22), C1-C3: 9 (1–17), C4-C6: 30 (18–49). Respective numbers for the GOR population were, N1-N3: 69 (51–92), N4-N6: 28 (15–38), C1-C3: 14 (5–20) and C4-C6: 17 (10–27).

Algae concentrations measured as RFU were significantly lower with the GOR population than with the STHLM population. The lower values for the GOR population were

FIGURE 4 | Relative fluorescence units as proxy for algal biomass for the control (black), STHLM (orange) and GOR (blue) E. affinis populations. Solid and dotted lines are respective high and low nutrient treatments, triangles and circles are respective pelagic and benthic treatments. Estimates and 95% CI are fitted by restricted maximum likelihood.

consistent over both resource supply and algae treatments, suggesting that this population is a more efficient grazer than the STHLM population (**Figure 4**, **Table 1**).

Concentrations of all the pelagic species (cryptophytes, H. triquetra, Isochrysis sp., and P. marina) were significantly lower with the GOR population than the STHLM population, whereas for S. marinoi there was no difference between the populations (**Figure 5**, **Table 2**). Concentrations of the large filamentous benthic diatoms Fragilaria sp. and Melosira sp. did not differ between the populations, while the smaller single celled benthic diatoms N. aurariae + N. perminuta were significantly lower with the GOR population than the STHLM population (**Figure 5**, **Table 3**).

The addition of benthic diatoms had a negative effects on the concentrations of all the pelagic species except H. triquetra (**Figure 5**, **Table 2**). Addition of nutrients had positive effects on the concentrations of all species except H. triquetra and Fragilaria sp. (**Figure 5**, **Tables 2**, **3**).

The numbers of nauplii at the end of the experiment depended on the interaction of algae treatment and population [t(5, 72) = 3.15, p = 0.002] and on the interaction of population and resource supply [t(5, 72) = 2.66, p = 0.001]. The GOR population was more negatively affected by benthic algae than the STHLM population, and the STHLM population was more negatively affected by low resource supplies (**Figure 6A**, **Table S3**).

For copepodites + adults we found a significant interaction of population origin and resource supply [t(4, 73) = 5.24, p < 0.001], where low resources led to a striking decrease in numbers for the STHLM population, but not for the GOR population. Furthermore, we found a significant main effect of algae treatment [t(4, 73) = 3.88, p < 0.001] indicating that the addition of benthic diatoms had weak negative effects for both populations (**Figure 6B**, **Table S4**).

The females clutch sizes depended on the three-way interaction of population x resource supply x algae [t(69) = 3.02, p = 0.004]. For the STHLM population, clutch sizes were unaffected by algae treatment but smaller in low resource supply. For the GOR population, clutch sizes differed between algae treatments in low resource supplies (**Figure 7A**, **Table S5**). For female prosome length (**Figure 7B**) we found no



The interactions tests for differences in the trend over time for the treatments: population and resource supply. Output of the treatments main effects, and the main effect of time are omitted.

GOR and STHLM population in the pelagic algae treatment, gray is the pelagic + benthic diatom treatment (same for both populations), solid and dashed lines are respective high and low nutrient treatment, filled and open circles are respective high and low nutrient treatment. Lines are fitted by local regression by the R package ggplot2 (Wickham, 2009), points are observed values.

interactive effects, but both the main effects of population [t(74) = 14.45, p < 0.001] and resource supply were significant [t(74) = 3.87, p < 0.001], whereas the effect of algae was not [t(74) = 0.19 p = 0.848]. Females were generally smaller in the STHLM population compared to the GOR population, and in low resource supply compared to high resource supply.

#### 4. DISCUSSION

In the common garden experiment we could show that the two E. affinis populations from the Baltic Sea differentiate genetically in female morphology and clutch size. These trait differences further affect how the populations interact with


The interactions tests for differences in the trend over time for the treatments: population, algae, and nutrients. Output of the treatments main effects, and the main effect of time are omitted.

TABLE 3 | ANOVA output from the mixed models on the benthic algae.


The interactions tests for differences in the trend over time for the treatments: population and nutrients. Output of the treatments main effects, and the main effect of time are omitted.

various experimental ecosystems. In general, the small-sized STHLM population was more sensitive to resource supply, whereas the large-sized GOR population was more sensitive to the type of algal community.

## 4.1. Common Garden Experiment: Morphology and Clutch Size

The two populations clearly differentiated in females' average shape, length, and clutch size. The differences in average length persisted from the first common garden experiment and throughout the mesocosm experiment made 2 years and ca 42 generations later. Thus, there should be little doubt that these differences are due to genetic differentiation. A previous study by Winkler et al. (2011) found neutral (mitochondrial DNA) genetic differentiation between populations from the Gulf of Riga and the Swedish Baltic coast. Our study confirms this by showing genetic differentiation in quantitative traits. There were however large individual variation within populations and individuals from the two populations overlap in shape, length and clutch size. This large morphological variation implies that whenever these two morphotypes coexists complete discrimination of the two will be difficult. The morphological variation in E. affinis has led to previous confusion in species classification. In Northern European estuaries two species of Eurytemora were described in the nineteenth century, E. affinis (Poppe, 1880) and E. hirundoides (Nordqvist, 1888). According to the taxonomic key provided by Gurney (1931) the small-sized population from the Stockholm Archipelago (STHLM) corresponds to a morphotype that was described as E. hirundoides. Our study shows that classification of Eurytemora by prosome length and fecundity can be spurious. Because these traits overlap between individuals even in genetically distinct populations in a controlled environment. Furthermore, prosome length and egg production in zooplankton are highly plastic traits that can be affected by both food conditions and temperature (Ban, 1994), which can further impede discrimination of the two populations unless they are reared under common conditions (Falconer and Mackay, 1996; Sanford and Kelly, 2010).

#### 4.2. Common Gardening Experiment

Based on the populations different effect on RFU we concluded that the GOR population were more efficient grazers than the STHLM population, which was expected because of their larger size. We found that the pelagic flagellates and the benthic diatoms N. aurariae + N. perminuta were in lower concentrations with the GOR population than with the STHLM population. Whereas there was no difference in the concentration of the pelagic diatom S. marinoi between populations. Perhaps this is due to that motile flagellates can escape the copepods feeding current (Jakobsen, 2002) whereas non-motile diatoms probably cannot. In general, larger copepods have a stronger feeding current (Peters and Downing, 1984) which could be harder for flagellates to evade and explain why the GOR population is more efficient in feeding on flagellates. Given that N. aurariae + N. perminuta settle on the bottom, their lower concentrations with the GOR population suggest that this population occupy the bottom of the mesocosms to a greater extent than the STHLM population. Furthermore, the concentrations of the two largest filamentous diatoms Melosira sp. and Fragilaria sp. did not differ between the populations. Our results suggest that there is no difference in the size spectrum of resource use between the populations, despite the populations' difference in size.

We found contrasting effects on the amount of nauplii and copepodites + adults for the two populations at the end of the experiment, which depended on resource supply and algae treatment. The large-sized GOR population was a more efficient

FIGURE 6 | Nauplii (A) and copepodites + adults (B) counts for the STHLM (orange) and GOR (blue) E. affinis populations at the end of the experiment. Triangles and circles are respective pelagic and benthic treatments, the resource supply treatment is shown on the x-axis. Estimates and 95% CI are from generalized least square models fitted with a power covariate to control for unequal variances. Observations are included as points. In (B) average number of individuals at start are shown with 95% CI for the STHLM and GOR populations, denoted by the letters S and G.

grazer and could therefore acquire the amount of resources needed even in low resource concentrations, which allowed them to maintain a similar population size in both high and low resource supply (**Figure 4**). In contrast, for the small-sized STHLM population, less resources resulted in a notable decrease in nauplii and copepodites + adults (**Figures 6A,B**). Larger animals are in general more efficient in resource acquisition and can feed on a wider size range of resources than smaller ones (Wilson, 1975; Gianuca et al., 2016), therefore the GOR population have a competitive advantage over the STHLM population. The STHLM population starved in the low nutrient treatment, probably because metabolic demand relative to body size is higher for smaller-sized zooplankton (Hansen et al., 1997), while both filtration and feeding rates of zooplankton increase with increasing size (Peters and Downing, 1984; Gianuca et al., 2016). Thus, the STHLM population acquires fewer resources while needing relatively more, and hence performs poorly at low resource concentrations. Our results suggest that a larger body size enables copepods to graze more efficiently and mitigate starvation at low resource concentrations.

The addition of benthic diatoms had a greater effect on the number of nauplii than on the number of copepodites + adults, with a stronger negative effect in the presence of benthic diatoms forthe GOR population than for the STHLM population. For calanoid copepods feeding on large aggregates is difficult compared to smaller particled food and can result in starvation (Koski et al., 2017). Thus, large filamentous diatoms may have disturbed the feeding process and reduced the copepods feeding efficiency. Because the GOR population was more negatively affected by benthic diatoms than the STHLM population , it suggests that females of the GOR population feed more from the bottom of the mesocosm where the filaments settled.

Differences in habitat use of E. affinis populations have been reported in the past and according to Gurney (1931) the GOR morphotype corresponds to "Buchtenplankton" (bay plankton) while the STHLM morphotype corresponds to "Kustenplankton" (coastal plankton). This is in agreement with the habitats where the populations were sampled. Similarly, two clades of E. affinis are reported from the St. Lawrence estuary. One clade mainly inhabits the inner reaches of the estuary and has invaded the Great Lakes (Winkler et al., 2008; Favier and Winkler, 2014), and has been reported as epi-benthic in the lakes (Torke, 2001). The clade in the Great Lakes have been classified as E. carolleeae by Vasquez et al. (2016). The other clade in St. Lawrence is mainly found in the middle parts of the estuary, suggesting that it occupies the pelagic habitat (Winkler et al., 2008; Favier and Winkler, 2014). Furthermore, epi-benthic zooplankton are often larger than pelagic zooplankton because a large size in a pelagic habitat increases the risk of predation by planktivorous fish (Brooks and Dodson, 1965). For the GOR population the larger sized females are likely prevented to enter the pelagic because of their conspicuousness, while nauplii and copepodites in the low stages likely stay closer to the surface, which is the general pattern for copepods (Irigoien et al., 2004; Casper and Thorp, 2007). This suggests that apart from morphological differences, the populations' behavior also differ.

Comparisons showed considerably smaller female clutch sizes of a European E. affinis population from the Seine than of two North American east coast populations assumed to be E. carolleeae from the St. Lawrence and the Chesapeake Bay (Beyrend-Dur et al., 2009; Devreker et al., 2012). Hence, the larger clutch size and epi-benthic behavior of the GOR population, similar to what has been reported in North American clades, suggest that this population could be the invasive E. carolleeae. In contrast, similar trait differences as observed in the present study have been reported between European E. affinis populations since the beginning of the 20th century (Gurney, 1931 and references therein, Sukhikh et al., 2016). The species name E. carolleeae has not been fully applied by researchers in the field. Regardless, if one categorize the E. affinis complex into clades (e.g., Favier and Winkler, 2014; Lee, 2016) or species (e.g., Sukhikh et al., 2013), it is important to focus on the contrasting ecological effects (e.g., freshwater invasions, habitat preferences) of adaptive variation within the E. affinis species complex.

Competition for nutrients between benthic diatoms and flagellates led to lower concentrations of flagellated species. However, we do not believe the lower concentrations of flagellates in the benthic algae treatment is the cause for the decline in nauplii for the GOR population. Because the resource supply treatment had no effect even in the treatments without benthic diatoms (**Figure 6A**). In contrast, the STHLM population was sensitive to resource supply, and for this population a reduction in flagellates caused by competition with benthic diatoms could explain why the nauplii count is lower when benthic diatoms were added.

# 4.3. Conclusion

Our study indicates that the large-sized GOR population of E. affinis is a superior competitor in low resource supply mainly because they can feed more efficiently on motile flagellates, than the small-sized population. However, when the resource supply is high and benthic diatoms are present, their competitive advantage disappears. We suggest that the benthic diatoms used in our study interfere in the feeding process and reduce their filtration efficiency. These results suggest that the GOR population exhibits a close relation to benthic habitats and that this morphotype is mainly found in shallow bay areas of the Baltic Sea. Apart from the morphological differences, the populations diverge in habitat choice rather than in resource specialization. We suggest that adaptive radiation in resource use is not the driver of morphological variation in E. affinis, but rather the trade-off between size-efficiency and vulnerability to predators. Perhaps the GOR population can afford to have a larger size because the adults stay close to the bottom. An important question for further research is to address the effects of planktivorous fish on zooplankton of different size and behavior, in pelagic vs. benthic habitats. This would highlight if the shallow littoral zone acts as a refugium for large epi-benthic morphotypes of E. affinis, whereas smaller morphotypes inhabit the coastal pelagic waters.

The past and current eutrophic state of the Baltic Sea (Andersen et al., 2017) may have been more advantageous for the small-sized STHLM population than the large-sized GOR population because the small-sized population has a positive response to increased resource supply. In addition eutrophication has reduced the extent of the littoral zone (Ojaveer et al., 2010) and thus the habitat for the larger-sized population. Given the area of suitable habitats, we assume that the small-sized population is by far the most common in the Baltic Sea. For example, we have only found the small-sized morphotype (i.e., none of the large extremes indicates a large-size morphotype) in samples from the central parts of Bothnian Bay, Bothnian Sea, Gulf of Finland, Gulf of Riga and the Baltic Proper (Karlsson, personal observation). We speculate that the largesized morphotype is not unique to the Pärnu Bay in the Gulf of Riga, but can probably be found in similar shallow bays with outlets of freshwater along the eastern and southern Baltic Proper.

Our results show that morphologically divergent, yet phylogenetically closely related zooplankton populations can have different effects on ecosystem functions, and in turn have different population increase in response to resource supply and algae community. These results underline that the two E. affinis morphotypes cannot be regarded as ecologically equal in their interaction with algae communities. Furthermore, their different habitat choice implies that the pelagic morphotype are more important for pelagic feeding fish such as herring and sprat, whereas the benthic morphotype is more important for fish species that feed in more complex littoral habitats such as perch and sticklebacks. This highlights the large morphological and life history variation in the E. affinis species complex and that populations affect ecosystem properties in different ways.

#### DATA AVAILABILITY STATEMENT

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

#### AUTHOR CONTRIBUTIONS

KK performed the experiments, analyzed the data, and wrote a first draft of the manuscript. KK and MW wrote the manuscript.

#### REFERENCES


#### FUNDING

This is a contribution to the BONUS BIO-C3 project and was supported by BONUS (Art 185), funded jointly by the EU and the Swedish Research Council FORMAS.

#### ACKNOWLEDGMENTS

We thank Marilyn Kalaus and Henn Ojaveer for field collection of zooplankton, and Erika Hedlund for taking mesocosm samples. We thank the two reviewers for their comments, which improved an earlier version of this manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars. 2018.00408/full#supplementary-material

calanoid copepod eurytemora affinis in chesapeake bay, usa. J. Exp. Mar. Biol. Ecol. 416, 72–83. doi: 10.1016/j.jembe.2012.02.010


**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 © 2018 Karlsson and Winder. 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(s) 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.

# Measures and Approaches in Trait-Based Phytoplankton Community Ecology – From Freshwater to Marine Ecosystems

Guntram Weithoff<sup>1</sup> \* and Beatrix E. Beisner<sup>2</sup>

<sup>1</sup> Department Ecology and Ecosystem Modelling, University of Potsdam, Potsdam, Germany, <sup>2</sup> Department of Biological Sciences, University of Québec à Montreal and Groupe de Recherche Interuniversitaire en Limnologie et en Environnement Aquatique, Montreal, QC, Canada

Trait-based approaches to investigate (short- and long-term) phytoplankton dynamics

and community assembly have become increasingly popular in freshwater and marine science. Although the nature of the pelagic habitat and the main phytoplankton taxa and ecology are relatively similar in both marine and freshwater systems, the lines of research have evolved, at least in part, separately. We compare and contrast the approaches adopted in marine and freshwater ecosystems with respect to phytoplankton functional traits. We note differences in study goals relating to functional trait use that assess community assembly and those that relate to ecosystem processes and biogeochemical cycling that affect the type of characteristics assigned as traits to phytoplankton taxa. Specific phytoplankton traits relevant for ecological function are examined in relation to herbivory, amplitude of environmental change and spatial and temporal scales of study. Major differences are identified, including the shorter time scale for regular environmental change in freshwater ecosystems compared to that in the open oceans as well as the type of sampling done by researchers based on site-accessibility. Overall, we encourage researchers to better motivate why they apply trait-based analyses to their studies and to make use of process-driven approaches, which are more common in marine studies. We further propose fully comparative trait studies conducted along the habitat gradient spanning freshwater to brackish to marine systems, or along geographic gradients. Such studies will benefit from the combined strength of both fields.

Keywords: algae, functional traits, ocean, lake, biogeochemistry, community assembly

# INTRODUCTION AND SOME HISTORY OF TRAIT-BASED APPROACHES IN PHYTOPLANKTON ECOLOGY

Phytoplankton has been studied for a very long time, starting with the description and identification of the diverse plethora of species. The linkage between phenotypically-based taxonomy and evolutionary history led to an initial scientifically sound categorization. Since a number of traits are linked to phylogeny, this phylogenetic categorization inherently incorporated a (weak) trait-based component (Bruggeman, 2011; Narwani et al., 2015). Moreover, long before trait-based approaches became popular, what are currently considered phytoplankton traits (e.g., edible or motile)

Edited by:

Kalle Olli, University of Tartu, Estonia

#### Reviewed by:

Alberto Basset, University of Salento, Italy Franziska Wemheuer, University of New South Wales, Australia

> \*Correspondence: Guntram Weithoff weithoff@uni-potsdam.de

#### Specialty section:

This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science

Received: 30 April 2018 Accepted: 28 January 2019 Published: 12 February 2019

#### Citation:

Weithoff G and Beisner BE (2019) Measures and Approaches in Trait-Based Phytoplankton Community Ecology – From Freshwater to Marine Ecosystems. Front. Mar. Sci. 6:40. doi: 10.3389/fmars.2019.00040

were commonly referred to, without explicitly using the term "trait." A good example is the conceptual narrative PEG model for seasonal lake plankton community succession that emerged as early as the 1980s (Sommer et al., 1986). This oft-referred to lake model is essentially trait-based, without using the term trait. The roots of trait-based thinking about phytoplankton are in Ramon Margalef's original work describing the "lifeforms" and "functional morphologies" favorable to remaining in the water column under different nutrient conditions, irrespective of whether one considers freshwater or marine environments (Margalef, 1978). Margalef's Mandala, as it has become known, underlies much thinking on phytoplankton community assembly based on traits. Margalef's thinking formed a strong undercurrent in the pillar of the development of trait-based approaches to phytoplankton community assembly in freshwaters. Subsequently, in the 1980s, Colin Reynolds established a functional classification of lake phytoplankton assemblages based on observations from field data (summarized in Reynolds, 1997 and elaborated further in Reynolds et al., 2002). Using this classification scheme, characteristic phytoplankton communities comprised of key species were identified and related to specific environmental conditions, including season, lake trophy, lake morphometry, light availability. Within each assemblage, functionally different groups were usually represented, demonstrating that for a particular environment, complementarity through different ecological strategies or trait combinations enable thriving communities. It is from these sources that the application of trait-based thinking and theory with respect to phytoplankton has its roots, and from which it is currently rapidly expanding.

#### PHYTOPLANKTON TRAIT TYPES

Traits for phytoplankton have been fairly well-explored and described for both marine and freshwater ecosystem types. In freshwaters, the focus has been on lake phytoplankton, and the conceptual link back to Reynold's earlier work on functional groups (see previous section) can be directly traced in many cases. There are now essentially two schools of thought, although they are highly related. One "explicitly" defines traits as categorical, nominal or continuous and the other groups organisms by morphometric features into functional groups sometimes called "morphospecies," continuing more directly Reynold's classification. The first is common to both marine and freshwater studies while the morphospecies classification has been more commonly used only in freshwater studies, likely owing to the tradition of studying seasonal succession in these environments and less focus on biogeochemical cycling that preclude the need for high quality physiological processing rates.

#### Explicitly-Defined Traits

Using explicitly defined traits lends itself well to the estimation of functional diversity indicators (such as functional dispersion, richness, diversity and evenness; Weithoff, 2003; Villéger et al., 2008; Laliberté and Legendre, 2010) as well as community weighted means to characterize functional composition of communities (e.g., Beisner and Longhi, 2013; Moser et al., 2017). It also enables the examination of trade-offs between different continuously measured traits such as those associated with nutrient uptake and storage (Litchman et al., 2007; Edwards et al., 2012). All of these indicators are relevant to questions of competitive interaction between phytoplankton, interaction with predators, and thus community assembly. Explicitly-defined traits will also characterize ecosystem functioning, including processes such as primary production, trophic transfer (biomass) and nutrient cycling and this has been addressed mainly in studies relating biodiversity to ecosystem function (BEF studies) (e.g., Zwart et al., 2015; Abonyi et al., 2018).

Generally, trait matrices are created for communities of interest whereby each species is characterized by a series of explicit traits. The traits used reflect both functions related to resource acquisition (bottom–up processes) and predation (top–down processes). In the explicit framework, commonly used traits are classified by type (Litchman and Klausmeier, 2008) as follows:

First: Morphological: size, biovolume, mucilage presence/absence, biological form (unicellular/colonial/ coenobium/sincoenobium/filament/chains).

Second: Physiological: (a) presence/absence of silica demand, heterocysts (N2-fixation), mixotrophy, toxin production, resting stages; (b) dominant pigment type; (c) nutrient and/or light uptake parameters; (d) temperature optima.

Third: Behavioral: presence/absence of flagella, aerotypes.

Generally many of these traits are based on literature or expert knowledge to classify traits, although many gaps still exist. There are new and evolving techniques to fill the data gaps on traits, mainly through the use of phylogenetic relationships (e.g., Bruggeman, 2011).

An important potential drawback with this approach, that assigns static traits to species, is that the actual expression of traits in phytoplankton may depend heavily on local (spatial and temporal) environmental conditions (but see section "Study of Communities vs. Selected Taxonomic Groups" on phenotypic plasticity). However (Violle et al., 2007), based on trait-based approaches in plants, argue that traits should be measurable on individuals independently of the environmental conditions. It is clear that such arguments do not necessarily apply in the case of the short-generation time and fast response times characteristic of phytoplankton. Notably, phytoplankton traits related to physiology may not always be expressed to the same degree, nor even active at all– for example classifying a species as a nitrogen-fixer does not characterize temporal variability in this trait that may occur in a changing environment. The next step in functional trait consideration of phytoplankton communities will be to better characterize in situ trait expression using genetic, biochemical or physiological probes. The challenge is doing so at the community level for the wide range of phytoplankton taxa and cells potentially present.

### Morphospecies Classification

fmars-06-00040 February 11, 2019 Time: 17:0 # 3

This approach is also commonly used, mainly in freshwater ecosystems in which it has roots (Reynolds et al., 2002), specifically in Reynolds (1997) trait-based "response group" classification of lake phytoplankton. Response group classification was based largely on how communities changed reliably year-to-year during seasonal succession. The morphospecies approach in particular counts, among other things, on the fact that shape (including size) affects buoyancy, uptake parameters and predator handling time and thus simplifies the original classification proposed by Reynolds (1988).

Morphospecies classification assumes that the empirical observation of the morphology of phytoplankton community constituents reflects the environmental and biotic interaction constraints on the community as well as the physiology of the organisms. It relies heavily on the idea that the morphometry of taxa reflect their autecology (physiology and functioning) (Kruk et al., 2010). Thus, like the response group approach of Reynolds (1988, 1997), it differs from the explicit trait approach in that it considers trait-mediated effects that arise from species interactions (by empirically assessing group composition) (Abonyi et al., 2018). Summarily, clusters of taxa based entirely on morphological features can be defined statistically and should represent meaningful functional phytoplankton groups.

In many ways, the original morphospecies classification resembles a reduced version of the explicitly-defined traits (Kruk et al., 2010). It was based originally on nine morphological traits including many that overlap with the previous classification: volume, maximum linear dimension, surface area, and the presence of mucilage, flagella, gas vesicles (or aero-topes), heterocysts or siliceous exoskeletal structures. Essentially, what is not found explicitly in the morphospecies approach is any explicit reference to physiological parameter rates – these are assumed (and were shown by Kruk et al., 2010) to be correlated with the morphological traits and group classification. The other difference from the explicitly-defined trait approach, is that instead of creating a species x trait matrix, then used to estimate functional composition or diversity, the focus here is on creating morphologically based functional groups (MBFGs) using clustering techniques. The groups thus created can then be related to environmental gradients or ecosystem functioning. In this approach, a potential short-coming however is the need to translate these morpho-groups into fitness (Naselli-Flores and Barone, 2011).

#### "Master" Traits

There has been a large focus on cell or body size as a "master" trait – so-called because of the large number of physiological and morphometric features that are regulated by body size. Also feedbacks with the environment are related to body size, as evident in the metabolic theory of ecology (e.g., Brown et al., 2004), which demonstrates the theoretical link between body size and ecosystem processes. Furthermore, it is recognized that body size and body size distribution (e.g., biomass size spectra) in phytoplankton communities can reflect many environmental factors such as vertical water dynamics and depth, trophic state, predation, and factors affecting growth (reviewed in Mouillot et al., 2006; Litchman and Klausmeier, 2008). It should be noted that the use of biomass size spectra has a long history in both marine and freshwater plankton studies: with older studies usually focused on estimating food web transfer efficiencies (e.g., Kerr, 1974; Gaedke, 1993; Sprules and Goyke, 1994) and not to prediction of other functional traits important for community assembly within specific groups like the phytoplankton themselves. More recent work has directly considered the relations between size spectra and effect traits such as nutrient recycling rates and in response to environmental conditions in aquatic ecosystems (e.g., Rodríguez et al., 2001) as is currently done in more explicit trait-based approaches. In phytoplankton body size is usually estimated as cell size along the maximum linear dimension (abbreviated as MLD or GALD: greatest axial linear dimension). The use of cell size is common to both the explicit trait and morphospecies frameworks and lends itself well to ecosystem modeling (e.g., Acevedo-Trejos et al., 2016).

Elemental stoichiometry of phytoplankton cells could be considered a type of "master trait" because, like size, these relationships are subject to biophysical rules that link growth rates with environments, food web interactions and thus biogeochemical cycles (Finkel et al., 2010). This consideration stems originally and largely from marine studies of phytoplankton (Finkel et al., 2010; Litchman et al., 2015), in large part due to the longer focus on biogeochemical cycles and feedbacks with phytoplankton (e.g., Follows and Dutkiewicz, 2011). Phytoplankton taxa have distinct stoichiometric ratios in both the macronutrients (e.g., C:P, N:P), but also in relation to micronutrients such as Cu, Fe, Zn, Co, Cd, and Mn; relationships that have evolved over time (reviewed in Litchman et al., 2015). Because of the strong potential effect of changes in stoichometry that would accompany changes in phytoplankton communities driven by environmental factors (other than biogeochemical ones), it is argued that the stoichiometric properties or traits of altered communities could inform on how biogeochemical processes are likely to be altered.

More recently, and in a different context, an interest in using phytoplankton to produce biofuel has lead to a fuller investigation of the stoichiometric traits in freshwater phytoplankton as well (Shurin et al., 2013, 2014). The goal has been to optimize long-term culture conditions for commercial-grade algal biofuels through the manipulation of C:N:P ratios that best optimize desired phytoplankton biomass production based on organismal requirements (stoichiometric traits).

#### Other Trait Types

There are a series of other trait types being developed, some focusing on more aggregate properties of communities (i.e., defined at the community and not individual level) and others focussing directly on the many individual phytoplankton present in a community. We will describe these briefly working from the aggregate community to the individual level.

#### Remotely-Sensed Traits

fmars-06-00040 February 11, 2019 Time: 17:0 # 4

In marine ecosystems, researchers have begun to use optical properties measured from satellites to estimate overall phytoplankton community structure. Essentially, from MERIS wavelengths, bio-optical traits (BOTs) have been derived to characterize phytoplankton functional types (PFTs) (Aiken et al., 2007, 2008). Such PFTs can then be linked to the global carbon cycle using bio-mechanistic models (Aiken et al., 2008). As with other trait-based methods, estimating the functional type (PFT) from remotely sensed oceanographic data requires that a specific BOT exist for each taxon. Bio-optical traits are derived more specifically from a suite of variables: Chl-a concentration; accessory pigments (Chl-b, Chl-c, carotenoids, phycobillins); pigment ratios (TChl-a/AP, TChl-a/TP, PPC/TC); phytoplankton absorption at 443 nm (aph443), and the spectral slope of aph.

#### Pigments as Traits

Using high performance liquid chromatography (HPLC), it is possible to characterize phytoplankton communities by their pigments (e.g., Zhao and Quigg, 2014). Post-processing of HPLC results enables observation of changes in communities in terms of often highly phylogenetically conserved pigment groups. Because taxa with different pigments respond differently to the light environment, their pigment profiles can be considered as functional traits.

#### Cell Chemotyping for Macromolecular Traits

This method relies on the fact that there are conserved strategies in energy and C partitioning amongst phytoplankton species that can then be used to characterize their growth rate traits. Cell chemotyping can be done using Fourier transform infrared (FTIR)-spectroscopy to obtain biochemical signatures in phytoplankton cells (Fanesi et al., 2017). In this study, the authors demonstrated that the macromolecular (FTIR absorption peaks of proteins, lipids, carbohydrates, and phosphorylated compounds) composition of phytoplankton cells was able to better characterize details of cell physiology related to growth rates. In this way, a common set of physiological traits can be used to define species and it has the potential to apply to in situ estimates of functioning of cells, even in dynamic environments.

#### Scanning Flow Cytometry Traits

Another recent method applies to characterizing individual cells in a community, rather than assigning traits to all individuals in the same taxon. Thus, the focus is on intra-specific variation. The idea is to scan communities using flow cytometry to obtain a series of parameter estimates (traits) for all individual cells present in a community (Fontana et al., 2014). These traits characterize morphology (length, area) as well as fluorescence characters (chlorophyll a, accessory and degraded pigments) on the individual level, thereby allowing for intra-specific variation in these traits to be estimated.

#### Trait Concept Standardization

To better integrate the various ways in which traits have been defined and used for phytoplankton, recent developments have been underway to better harmonize data and allow for its interoperability (e.g., exchange between computer platforms or software). In particular the development of thesauri is occurring generally in ecology, to standardize the semantic properties for trait labels, including their definitions. For phytoplankton, the PhytoTraits thesaurus has recently been created<sup>1</sup> (Rosati et al., 2017). This thesaurus creates a standard definition of phytoplankton traits and their measurements that will permit better data integration across studies. A thesaurus ensures the integration amongst different datasets that use different terms for the same concept (for example biomass, weight, or dry weight; Rosati et al., 2017). Thesaurus construction uses eco-informatics approaches to acquire, integrate and analyze trait terms, developing a standard terminology or "controlled vocabulary" to which the scientific community can refer. Each term or concept obtains its own Uniform Resource Identifier (URI that is a unique, persistent internet label). New terms can then be added to thesauri such as PhytoTrait via a portal and traits labeled differently but referring to the same concept can be identified and properly classified.

#### Traits Used in Marine and Freshwater Research

Due to the similar nature of marine and freshwater habitats, relevant phytoplankton traits are also very similar between marine and freshwater species. The master trait size is commonly used independently of the habitat under consideration, being a unifying trait (Litchman and Klausmeier, 2008). Other, less frequently used traits, are resource use traits, e.g., nitrate acquisition (N-fixation), mixotrophy or traits related to light harvesting (Barton et al., 2013; Edwards et al., 2013). In freshwater studies, size is often expanded to a more complex measure of shape, taking morpho-functional differences between species into account (e.g., Kruk et al., 2010), often in combination with phylogenetic/taxonomic relatedness (Huszar and Caraco, 1998; Cellamare et al., 2013; Segura et al., 2013). In studies investigating the vertical structure of lakes, the pigment composition (Beisner and Longhi, 2013) and the motility of cells (by gas vacuoles or flagella) are further relevant traits (e.g., Pomati et al., 2012), likely less important in the well-mixed surface layer of the open ocean.

#### LITERATURE SURVEY

To better compare how trait-based approaches have been used in phytoplankton ecological research in marine and freshwater studies, we performed a literature search using the criteria "phytoplankton" AND "trait<sup>∗</sup> " in the Web of Science (Clarivate Analytics) on March 7, 2018. A total of 631 entries matched our criteria; the oldest one is from 1976 (Vladimirskaya et al., 1976), taking 15 years until the next papers appeared. From that point onwards, the number of contributions increased rapidly with a total of 243 papers published in 2015–2017 on phytoplankton and trait<sup>∗</sup> . We discarded > 300 papers from our subsequent analysis because the term trait was not at all related to phytoplankton or because the study did not fit into

<sup>1</sup>http://thesauri.lifewatchitaly.eu/PhytoTraits/index.php

the context of phytoplankton trait-based ecology. Another ca. 150 were also removed because of a lack of a community context, dealing with only a single trait or a single species, or a particular measurement technique. A total of 150 papers remained and these were grouped into marine or freshwater (lakes and rivers) studies, then classified according to study type (field observational surveys, experimental, mathematical modeling and conceptual/review) and according their research question (community assembly, response to environmental change or relations with biogeochemical cycling). A substantial number of the papers could not assigned clearly to each category, because research questions were overarching or the themes were only loosely related to our categories. Thus, the exact numbers we present are debatable, but they do indicate some overall tendencies. From the subset of 150 articles, we found an almost equal share of marine (67) and freshwater (76) studies (**Figure 1** and **Table 1**). Furthermore, there were several modeling or conceptual/review studies that were either not specific to one habitat type or explicitly dealt with both. Grouping these articles into field surveys, experimental, mathematical modeling and conceptual/review studies, it became obvious that experimental studies were underrepresented, while field surveys and mathematical modeling studies dominate the literature. Mathematical modeling studies were more commonly applied to marine ecosystems, while field studies were more often conducted in freshwater studies. This might reflect the accessibility and higher sampling frequencies of freshwater sites, but also the goal of the respective studies (see section "Comparing Freshwater and Marine Trait Studies: Assembly vs. Biogeochemistry"). This literature survey better enabled us to determine the different ways in which traits have been incorporated into the study of phytoplankton ecology in both freshwater and marine environments, which we now elaborate with some examples.

## STUDY OF COMMUNITIES vs. SELECTED TAXONOMIC GROUPS

Trait-based phytoplankton studies generally consider either the whole community or focus on responses in certain phylogenetic groups such as the diatoms or other major phyla. Whole community studies from freshwater sites typically analyze seasonal dynamics or a component of temporal variation, such as the spring bloom, using classical phylogenetic (taxonomic) approaches. A trait component, such as cell size as a master trait, or some measures of functional diversity based on functional traits is sometimes added to broaden the view. However, such studies do not focus on a trait-based approach per se making it difficult to group them unambiguously into our categories.

Another aspect in community studies is the analysis of spatial (mostly vertical) differences in communities from stratified environments (e.g., Longhi and Beisner, 2010; Santana et al., 2017). Often the response to environmental changes or a detailed analysis of the traits within a particular taxonomic group (representing a particular set of traits) has been the focus. For example, diatoms represent a phylogenetic group common to

assigned to four different types according to their predominant approach and according to ecosystem type. The conceptual/review articles were often not specific to a habitat type, so no further distinctions were made.

both marine and freshwater studies. They are abundant in lakes, rivers and oceans alike, and play an important role in food webs and the biogeochemical cycles. As a result, their trait distribution and the consequences for diatom dynamics have been commonly studied in both ecosystem types (e.g., Terseleer et al., 2014; B-Béres et al., 2017; Taherzadeh et al., 2017). Diatoms come with the added advantage that they preserve in sediments and can thus be the focus of longer-term paleo-limnological or paleooceanographic analyses (Smol et al., 2005; McKay et al., 2012). In marine studies dinoflagellates and coccolithophores represent other target groups. For dinoflagellates, important traits relate either to toxin production or to mixotrophy, for coccolithophores relevant traits relate to the tolerance to ocean acidification (see section "Genotypic Trait Plasticity").

# TRAIT VARIABILITY: GENOTYPIC vs. PHENOTYPIC

An important, though often neglected, aspect in the quantification of traits is the phenotypic and genotypic plasticity of traits. Plasticity or trait variation is difficult to measure, but it is important for the parametrization of mathematical models and for the understanding of community processes on timescales relevant for phytoplankton ecology.

# Phenotypic Trait Plasticity

Many traits can vary substantially according to environmental conditions. While some traits can be measured on individuals (e.g., body size), some are more amenable to measurement at the community level, representing the average of a population of



Note in some studies, the focus could not clearly assigned to one of the categories.

individuals, such as nutrient ratios or physiological traits. Size, as a master trait (see section "Phytoplankton Trait Types"), is related to many ecological processes, but it can also be highly variable. To account for cell size variation within a phytoplankton population or community and over extended periods, modern techniques now exist, although not regularly utilized to assess size variation, including flow cytometry or related systems that also provide images (e.g., Flow-CAM, CytoBOT).

Body size variation, in addition to being easy to measure, is ecological very relevant in phytoplankton and thus often considered in ecological studies. For example, because zooplankton herbivory is often size-selective, a selective pressure on distinct prey size classes is introduced, thereby changing the size distribution of a population or community (Sommer et al., 2001; Lewandowska et al., 2014). Furthermore, in freshwater ecosystems, it has been observed that the mere presence of the herbivore Daphnia, can induce changes in the colony size of their algal prey driven by kairomones, even without any herbivory (Hessen and VanDonk, 1993). In another example, the physiological status of phytoplankton cells can also drive their body sizes and thus the variation observed in a population or community: during the regular cell growth cycle the size of the cells might vary by a factor of two or even more (Massie et al., 2010). With increasing nutrient limitation, the cell cycle arrests at a certain limiting nutrient concentration and it continues only under nutrient replete conditions, which may synchronize a population's cell size to a smaller value (Massie et al., 2010). Body size in diatoms is a particularly interesting case of size variation. Diatoms reduce their average cell size through progressive population growth, because after cell division the newly-built theca is formed from the inner (smaller) part of the mother theca leading to a continuous cell size reduction of one of the daughter cells (Lee, 1999). Associated to a smaller body size is an increase in the surface to volume ratio, facilitating the nutrient uptake per unit of volume (Reynolds, 1997).

Physiological traits also vary in response to environmental changes. A common strategy is the optimization of resource use efficiency. This includes an increase in chlorophyll-a or other photosynthetic pigments by phytoplankton cells to better collect photons at low light availabilities or spectral discontinuities (Richardson et al., 1981; Falkowski and LaRoche, 1991). Another example is the production of the enzyme nitrogenase under nitrogen limiting conditions, which enables the fixation of elementary nitrogen. An increasing nitrogenase activity is typically found with increasing duration of nitrogen-depleted conditions (e.g., Paerl, 1988).

#### Genotypic Trait Plasticity

Traits and trait values not only vary in response to environmental factors but also among different genotypes within a population. Information on genotypic trait variation is surprisingly rare and often stems from research in other fields than traitbased ecology. We can draw on several examples involving phytoplankton. First, with respect to toxicity, different strains of potentially toxic algae have been analyzed to understand the mechanisms and physiological costs behind toxin production. In marine habitats, dinoflagellates (e.g., Alexandrium) are the most prominent toxic group, while cyanobacteria (e.g., Microcystis and others) are the main toxin producers in freshwaters. Within one species, the ability and magnitude of toxin production among different strains varies considerably, making it difficult to assign this trait to a species (Ichimi et al., 2002; Kardinaal et al., 2007; Touzet et al., 2008). Another source for information on trait plasticity comes from invasion biology: invasive species usually regarded as species with a high genotypic plasticity facilitating their establishment in newly invaded habitats. For the invasive cyanobacterium Cylindrospermopsis, high genotypic trait variability is notable (Willis et al., 2016; Bolius et al., 2017). Genotypic variation in phytoplankton also comes from global change research as it is related to ocean acidification. The response of coccolithophores to declining pH-levels has been shown to be strain-specific (Müller et al., 2015; Rickaby et al., 2016) and varies at different times scales (Meyer and Riebesell, 2015; Schlüter et al., 2016). Differentiating between such shortterm strain-specific acclimation and long-term adaptation makes it difficult to predict community responses to ocean acidification.

#### Consequences for Trait-Based Modeling

For several applications of trait-based approaches, trait variability might not be a serious disadvantage, as for example with respect to aspects of community assembly in relatively stable environments. However, for modeling ecosystem processes or species interactions more directly and on shorter timescales, the variability in traits may be critical. However, not only does the variability itself complicate trait-based modeling (Coutinho et al., 2016), but the shape of the trait distribution also plays an important role (Gaedke and Klauschies, 2018). The shape of the trait distribution determines the potential for adaptive responses to environmental change. For example, when environmental forces act specifically on a certain trait and a distinct range of the trait values, a bimodal trait distribution might permit a more sustainable response option relative to a normal trait distribution, especially if the pressure is imposed around the mean. A great

deal more work will be necessary to establish the distributions of highly variable individual phytoplankton traits to inform these modeling approaches.

# COMPARING FRESHWATER AND MARINE TRAIT STUDIES: ASSEMBLY vs. BIOGEOCHEMISTRY

Narrowing our literature survey subset of 150 articles further, by including only those studies that had either a broad spatial (many samples over a broader area) or temporal scale (time series data), as well as those that considered the entire (or a substantial part of the) phytoplankton community under a trait-based view, 73 articles remained. The majority were freshwater studies, with 29 dealing mostly with environmental factors driving community assembly or temporal dynamics when assessed using traits (perspective 1, **Table 1** and **Supplementary Table S1**). Other studies at the community level examined trait distributions more directly by focusing on interactions between them and how this could influence community structure or dynamics (without explicit consideration of environmental variation). This group (perspective 2, **Table 1**) was more evenly split between marine and freshwater environments (6–7 per environment). On the other hand, studies that focused more directly on ecosystem consequences of traits (perspective 3, **Table 1**), including ecosystem biogeography and biogeochemical processes including stoichiometry, were much less numerous (total of 13), with the majority (8) from marine environments and another two combining freshwater and marine. Irrespective of research perspective, a total of seven studies used traitbased approaches in marine and freshwaters together, and three considered estuaries.

Generally speaking, for phytoplankton, the nature of the pelagic habitat as well as the main representative phyla and their ecology are relatively similar across both marine and freshwater ecosystems. Despite this similarity, research questions and approaches have evolved, at least in part, separately, in the two environments. One crucial difference between these ecosystem types is the accessibility and the possibility to take frequent samples. For many lakes, routine sampling campaigns are regularly conducted, either by local authorities or scientific institutes and universities. Thus, many data sets with some temporal resolution and good physico-chemical background data exist. In comparison, less data exists for coastal regions, and regular, high temporal resolution biological sampling in open ocean sites is very rare (but see Buitenhuis et al., 2013; Brun et al., 2015).

Traditionally, much ecological research was performed at field stations investigating the on-site environment. This is particularly the case for freshwater biological stations, many of which were established to combine monitoring and research, often in relation to water utilities and providers related to drinking water, fisheries or recreational uses. For these purposes, water quality and phytoplankton community composition has been extensively studied, with frequent and regular sampling of the pelagic food web, to ensure the management of drinking water quality and quantity or of fish harvest. At marine biological stations, local sites have also been sampled, although they have served primarily as entry points to understanding the open ocean, for which cruises are expensive and do not allow for regular sampling.

These different ecosystem accessibility and scientific histories are mirrored in the different research questions addressed in phytoplankton trait-based studies in freshwater and marine systems. Trait-based studies from the open ocean preferentially investigate biogeographic trait patterns (Acevedo-Trejos et al., 2013; Barton et al., 2013) or they have a strong biogeochemical focus aiming for the quantification of globally important processes related to primary production such as carbon-dioxide acquisition, nitrogen fixation (Breton et al., 2017), or oceanic stoichiometry and biochemistry (Litchman et al., 2007, 2015; Strom, 2008; Finkel et al., 2010; Bonachela et al., 2016), but also on phytoplankton dynamics (Alexander et al., 2015) (**Figure 2**). In oceanography, many studies model trait-based processes, aiming for broad-scale extrapolation of these globally (Smith et al., 2016; Vallina et al., 2017). From a conceptual point of view, marine research considers how traits and trait distribution drive processes determining the environmental conditions (e.g., biogeochemistry).

In the more accessible coastal or freshwater sites such as lakes, the conceptual view is often in the opposite direction (**Figure 2**), considering instead how phytoplankton communities respond to the environment and which environmental factors drive the traitbased community assembly. Answers to these research questions rely on a sampling regime with a high temporal resolution. A prominent example for a coastal long-term sampling site is the Western Channel Observatory site in the English Channel off the coast of Plymouth, United Kingdom (Edwards et al., 2013; Mutshinda et al., 2017). Coastal sites that are influenced by the tide or estuaries are subjected to rapid environmental changes so that the response to such changes became into the focus of the investigations (Aubry et al., 2017; Klais et al., 2017; Moser et al., 2017). On a similar time-scale, lakes (and rivers) are also characterized by rapid environmental changes and studies on these systems thus often deal with trait-driven responses to these by the community. Longer-term studies that consider several

months or years are highly suitable to these types of questions, especially when similar environmental changes are repeated during the investigation, representing a form of temporal replication (Pomati et al., 2012; Edwards et al., 2013; Tsai et al., 2014; Weithoff et al., 2015; Bortolini et al., 2016; Weithoff and Gaedke, 2017). Alternatively, instead of temporal replication, trait-based approaches also enable a space-for-time substitution in which many sites receiving a similar environmental signal (e.g., sites along a coastal or river stretch or lakes within the same region) are sampled only once or a very few times (Longhi and Beisner, 2010; Machado et al., 2016; Santana et al., 2017; Vadrucci et al., 2017). Trait-based approaches become particularly helpful in such cases, as they are ataxonomic, facilitating the detection of common response patterns within these potentially disparate communities. Another aspect of freshwater trait-based studies is the analysis of the vertical distribution of phytoplankton (Pomati et al., 2012; Beisner and Longhi, 2013; Santana et al., 2017), when steep physical-chemical gradients exist. One drawback we noted in several studies is that they lack a clear rationale as to why they use trait-based approaches in relation to the ecological goals (as also argued by Hébert et al., 2017 for zooplankton). Clearly outlined goals related to improved understanding of community assembly vs. contributions of phytoplankton to biogeochemical cycles should be a part of every study that uses trait-based approaches.

#### RECOMMENDATIONS AND CONCLUSION

Similarities and differences in phytoplankton trait-based approaches between marine and freshwater studies can be identified by the ways in which traits are assigned and the focus of the primary research questions to which traits are applied. With respect to trait assignment, two related schools of thought exist, either "explicitly" defining several species' traits simultaneously so that there is little overlap between taxa, or by restricting classification to "morphospecies" types, continuing more directly Reynold's earlier classification. Explicitly-defined traits are commonly used in both marine and freshwater studies, while morphospecies classification has been more commonly used in freshwater habitats. We attribute this difference to the tradition of studying seasonal succession and community assembly under varying environmental conditions in highly accessible freshwater environments that preempts the need for higher resolution physiological processing rates. Marine studies often focus on globally relevant biogeochemical cycles, for which key processrelated traits are studied on the most abundant taxa, especially

#### REFERENCES


coccolithophores, diatoms, dinoflagellates and prochlorophytes. In general, researchers are encouraged to better motivate, why they apply trait-based analyses to their studies. They will benefit from applying the strong process-driven approaches used in most marine studies to investigate the relationship between traits and biogeochemical processes. Marine researchers might benefit from the mechanistic approaches used by freshwater researchers to better understand community assembly using traits, if more resolved time-series data can be obtained through autonomous buoys or remote sensing. As another avenue of future research, we propose fully comparative trait studies conducted along the habitat gradient spanning freshwater to brackish to marine systems. Such studies can combine the strengths of both fields and would be of general ecological interest. Other comparative studies could analyze latitudinal, altitudinal, or trophic gradients using phytoplankton traits as a common currency to assess community structure. In some cases, data likely already exist to conduct these studies. Given the low number of experimental studies we noted in our literature survey, it would be very useful to test the field survey patterns more explicitly in manipulated experiments, such as in mesocosms, to ascertain mechanisms influencing trait distribution and expression. Last, but not least, a common challenge for future research in both systems remains to account for within-species trait variation, a critical component to trait-based studies in all aquatic habitats, especially in an era of rapid environmental change.

#### AUTHOR CONTRIBUTIONS

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

#### FUNDING

BEB was supported by the Alexander von Humboldt-Stiftung.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars. 2019.00040/full#supplementary-material

TABLE S1 | List of articles that had either a broad spatial (many samples over a broader area) or temporal scale (time series data), as well as those that considered the entire (or a substantial part of the) phytoplankton community under a trait-based view; sorted by year.

in the oceans. Glob. Ecol. Biogeogr. 22, 1060–1070. doi: 10.1038/srep 08918


for phytoplankton functional types and photosynthetic parameters. Internat. J. Rem. Sens. 28, 497–516. doi: 10.1080/01431160600821036



phytoplankton species to ocean acidification. Sci. Adv. 2:e1501660. doi: 10.1126/ sciadv.1501660



**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 © 2019 Weithoff and Beisner. 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(s) 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.

# Manipulation of Non-random Species Loss in Natural Phytoplankton: Qualitative and Quantitative Evaluation of Different Approaches

Friederike G. Engel <sup>1</sup> \*, Aleksandra M. Lewandowska<sup>2</sup> , Sarah L. Eggers <sup>3</sup> and Birte Matthiessen<sup>1</sup>

<sup>1</sup> Experimental Ecology-Food Webs, Marine Ecology, GEOMAR Helmholtz Centre for Ocean Research, Kiel, Germany, 2 Institute for Chemistry and Biology of the Marine Environment, Carl von Ossietzky University of Oldenburg, Oldenburg, Germany, <sup>3</sup> Hustedt Diatom Study Centre, Polar Biological Oceanography, Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research, Bremerhaven, Germany

#### Edited by:

Kristian Spilling, Finnish Environment Institute (SYKE), Finland

#### Reviewed by:

Maria Stockenreiter, Fakultät für Biologie, Ludwig-Maximilians-Universität München, Germany Punyasloke Bhadury, Indian Institute of Science Education and Research Kolkata, India

> \*Correspondence: Friederike G. Engel fengel@geomar.de

#### Specialty section:

This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science

Received: 20 July 2017 Accepted: 19 September 2017 Published: 27 September 2017

#### Citation:

Engel FG, Lewandowska AM, Eggers SL and Matthiessen B (2017) Manipulation of Non-random Species Loss in Natural Phytoplankton: Qualitative and Quantitative Evaluation of Different Approaches. Front. Mar. Sci. 4:317. doi: 10.3389/fmars.2017.00317 Ecological research in recent decades revealed that species loss has a predominantly negative effect on ecosystem functioning and stability. Most of these studies were based on random species loss scenarios, but extinctions in nature are not random. Recent experimental studies using macroscopic communities largely advanced knowledge about the effects of non-random species loss. However, in microscopic communities like the phytoplankton, implementing realistic species loss scenarios is challenging and experimental data are scarce. Creating more realistic experiments to study the role of phytoplankton diversity for ecosystem functioning is particularly important, as they provide up to 50% of global primary productivity, form the basis of all pelagic food webs, and are important for biogeochemical cycling. In this study, we experimentally tested and evaluated three methods for non-random species loss in a natural marine phytoplankton community. Dilution, filtration, and heat stress removed the targeted rare, large, and sensitive species, respectively. All these species groups are extremely vulnerable to extinction in future climate scenarios and play important roles in the communities. Dilution and filtration with a fine mesh additionally decreased initial biomass, which increased the variability of species left in the respective replicates. The methods tested in this study can be used to non-randomly manipulate phytoplankton species diversity in communities used for experiments. However, in studies where species identities are more important than species richness, the dilution and filtration methods should be modified to eliminate the effect of decreasing initial biomass.

Keywords: phytoplankton, non-random species loss, realistic species loss, species loss manipulation, extinction

# INTRODUCTION

Over the last centuries, humans have increasingly influenced and modified all ecosystems on planet Earth. Habitat destruction, the emission of greenhouse gases, and the introduction of non-native species led to species loss rates comparable to historic mass extinction events (Barnosky et al., 2011). A large number of experimental studies tested the effects of species loss on different ecosystem processes. They uncovered a generally negative effect of species loss on ecosystem functioning and stability (Hooper et al., 2005; Cardinale et al., 2012). Depending on the magnitude of species loss, its effect size on ecosystem functioning is comparable to those arising from direct effects of environmental factors, such as acidification or nutrient pollution (Hooper et al., 2012). To date, in most diversity manipulation experiments, researchers altered species richness by randomly adding species to or excluding them from communities (Bracken and Low, 2012; Mensens et al., 2015; Radchuk et al., 2015). However, loss of species in natural communities is not random, but depends on a multitude of factors including population size (i.e., rarity), body size, and sensitivity to environmental stress (Gross and Cardinale, 2005). Previous research showed that the effects of biodiversity loss on ecosystem functioning can largely differ in randomly assembled communities compared to communities experiencing non-random species loss (Solan et al., 2004; Bracken et al., 2008; Mensens et al., 2015).

Manually removing or adding target species is feasible in communities with larger and substrate-bounded organisms, such as in grasslands (Zavaleta and Hulvey, 2004; Selmants et al., 2012, 2014) or the marine benthos (Bracken et al., 2008; Stachowicz et al., 2008; Bracken and Low, 2012). However, in microbial communities, such as phytoplankton, the manipulation of non-random species loss from natural communities is particularly challenging. A recent laboratory experiment with freshwater phytoplankton showed that it is possible to create communities with distinct species richness gradients using different levels of dilution and disturbance (Hammerstein et al., 2017). Nevertheless, technically it is nearly impossible to remove certain target species or groups from a natural microbial community without significantly altering the overall organism density. Hence, such diversity manipulations are prone to be confounded with a hidden density treatment, which can have two major consequences. First, it can directly affect the total biomass which is often used as a measure for ecosystem functioning. Second, it can enhance the variability of which species are remaining in a treated community and thus indirectly affect ecosystem functioning. To date the latter problem remains largely unquantified, so that the effects of realistic changes in phytoplankton diversity on ecosystem functioning in experiments remain essentially unknown (Gamfeldt et al., 2015).

Even though extinctions in the oceans are not as common as on land, and direct extinctions caused by humans were less often recorded for marine organisms, the influence humans have on aquatic ecosystems is immense and will likely continue to increase (McCauley et al., 2015). In particular for marine phytoplankton, that contribute almost 50% to global primary production (Field et al., 1998) and play an important role in biogeochemical cycling (Falkowski et al., 1998), future changes in biodiversity remain speculative.

In the present study, we tested different methods to nonrandomly remove rare, large and sensitive species from natural phytoplankton communities. These methods can be employed to manipulate community composition before the onset of an experiment or they can be used as factors in the experiment to create different levels of species composition. In the following section we elaborate on the rationale behind focusing on these particular groups.

## Loss of Rare Species

Rare species are often characterized by small population sizes, narrow geographical ranges and little genetic variation within populations (Rabinowitz, 1981; Frankham, 2005; Harnik et al., 2012). These characteristics in combination with stochastic processes make rare species more likely to go extinct than common species (Lande, 1993; Frankham, 2005; O'Grady et al., 2006; Leitão et al., 2016). Due to numerical disadvantages, rarity is often accompanied by competitive inferiority which can lead to extinction by competitive exclusion. An example of competitive inferiority can be seen in the context of priority effects, where the first colonizers exhibit a numerical advantage such that they can exclude later arriving species by monopolizing shared resources (Urban and De Meester, 2009; de Meester et al., 2016). Priority effects can also happen at the onset of a phytoplankton bloom. Eggers and Matthiessen (2013) showed experimentally that the initial structure of a phytoplankton community influenced the identity of the respective dominant species at bloom peak.

Despite their increased risk of extinction, rare species are important in communities because they can maintain ecosystem functions under changing environmental conditions when they substitute for dominant species that are lost or decline in numbers (Walker et al., 1999; Norberg et al., 2001; Elmqvist et al., 2003; Lyons et al., 2005; Mouillot et al., 2013; Jain et al., 2014; Leitão et al., 2016; Jousset et al., 2017). Hence, manipulating the loss of rare species in bloom building phytoplankton can provide important information about possible future scenarios, in which these species might be lost at disproportionally high rates.

#### Loss of Large Species

Experiments and observations have shown that rising seawater temperatures lead to the reduction of average cell size in phytoplankton communities (Morán et al., 2010; Peter and Sommer, 2012; Sommer et al., 2012). Though the underlying reasons are not completely understood yet, current research suggests that it is partly driven by decreased nutrient availability in the euphotic layer of the oceans due to stronger stratification under warmer conditions (Hofmann et al., 2011; Winder and Sommer, 2012; Acevedo-Trejos et al., 2014; Lewandowska et al., 2014). This situation benefits smaller-sized cells. Small cells have more efficient nutrient uptake rates because they have more favorable surface area to volume ratios (Aksnes and Egge, 1991; Raven, 1998; Mara-ón, 2014). Additionally, experiments have attributed the increase in smaller phytoplankton cell sizes with warming to trophic interactions. Depending on the preferred size spectrum of the prey, more intense grazing by zooplankton under higher temperatures can lead to the reduction in average phytoplankton cell size in a community (Sommer and Lewandowska, 2011). Since smaller cell size has been proposed to be one of the universal responses to global warming for aquatic organisms (Daufresne et al., 2009), testing the effect of losing larger species from phytoplankton communities can be very important for future phytoplankton diversity experiments.

## Loss of Sensitive Species and Potential Interaction with Rarity

Species are considered sensitive when they have a narrow environmental optimum and therefore steep reaction norms. They are generally very susceptible to changes in the environment including a variety of biotic and abiotic factors, such as salinity, pH, and temperature. In this study, we tested the effect of losing heat sensitive species, because one of the major changes in the future ocean is predicted to be an increase in the average surface ocean temperature (IPCC, 2014). Many species that are sensitive to environmental change are at the same time low in their abundance, because they have very specialized habitat requirements (Davies et al., 2004). Therefore, we additionally combined the loss of sensitive and rare species as one of the treatments, a valuable addition to the experimental design as it can improve our understanding of the mechanisms underlying sensitive species loss.

The aim of this study was to test the immediate effects of the above described non-random species loss scenarios on a natural phytoplankton community and to evaluate their impact on diversity change at bloom peak. Additionally, we aimed to qualitatively analyze how far the manipulations led to increased variability in species identities present in the communities (i.e., whether the methods manipulated the loss of the same species among replicates).

# MATERIALS AND METHODS

#### Experimental Set-Up

The experimental units consisted of white polypropylene buckets filled with 25 L of sterile filtered (0.2µm) seawater that we collected from the Kiel Fjord (Kiel, Germany) in March 2013. At that time it was possible to obtain nutrient-rich winter water with dissolved inorganic nutrient concentrations of 30.73 µmol L <sup>−</sup><sup>1</sup> nitrate and nitrite (NO<sup>3</sup> + NO2), 0.77 µmol L−<sup>1</sup> phosphate (PO4), and 34.77 µmol L−<sup>1</sup> silicate (SiO4).

We distributed the experimental units randomly in two temperature controlled rooms (10◦C) that were equipped with computer-programmed ceiling light units (Econlux, Hibay LED 100 W, full sun-light spectrum) providing light from above and creating an underwater light intensity of 130 µmol m−<sup>2</sup> s −1 and a light-dark cycle of 16:8 h. This corresponded to the local irradiance levels in Kiel Fjord at the time of the year the experiment was conducted (Brock, 1981).

Depending on the treatment, we inoculated the experimental units with different volumes of a natural phytoplankton stock community that was collected shortly before the experimental start from surface waters of the Kiel Fjord, Germany in April 2013 at the onset of spring bloom. To exclude meso-grazers (copepods) from the experiment, we pre-filtered the stock community with a 200µm sieve. For a detailed description of treatment application see section Manipulation of Different Species Loss Scenarios and **Table 1**.

We stirred the water body of each experimental unit carefully once per day to ensure a homogenous distribution of the phytoplankton in the experimental units. To prevent airborne particle transport into the experimental units, but still allowing for oxygen exchange and light penetration, we loosely covered the buckets with transparent polyethylene foil.

### Manipulation of Different Species Loss Scenarios

We applied three methods (dilution, filtration, and heat stress), resulting in five treatments (**Table 1**): **control** (Co–no manipulation), **dilution** to remove rare species (two levels: D1– weak dilution, D2–strong dilution), **filtration** to remove large species (two levels: F1–coarse filtration, F2–fine filtration), **heat stressed** (S) to remove heat sensitive species, and **dilution of heat stressed** to simultaneously remove sensitive and rare species (two levels: SxD1–weak dilution of heat stressed, SxD2–strong dilution of heat stressed). Each treatment level and combination was 4-fold replicated which resulted in 32 experimental units in total.

In order to lose temperature sensitive species prior to the experimental onset, we separated the collected and pre-filtered stock community into two temperature treatments (control (10◦C) = non-heated stock community; heat stress (22◦C) = heated stock community). We chose 22◦C, because it represents a critical temperature for many Baltic Sea auto- and heterotrophs (Reusch et al., 2005; Eggers et al., 2012; Werner et al., 2016). The stock communities were stored in closed glass bottles (Duran, 2,500 mL) in climate cabinets for 24 h. After 24 h, we inoculated

TABLE 1 | Overview of the experimental design with treatment names and levels, abbreviations, and how the treatments were realized.


The last column shows which species groups were targeted for removal.

the experimental units according to the different treatments (**Table 1**). For the control (Co), we added 10 mL of the nonheated stock community. We applied the two levels of the dilution treatment by adding inocula of 1 and 0.1 mL of the non-heated stock community, resulting in 10% (D1) and 1% (D2) of the control concentrations, respectively. For the two levels of the filtration treatment, we filtered the non-heated stock community with 100µm (F1) and 20µm (F2) sieves, respectively. 10 mL of the respective filtrates were added to the experimental units. For the heat stressed treatment (S), we added 10 mL of the heated stock community to the experimental units. To prepare the combined treatment of heat stress and dilution, we added inocula of 1 and 0.1 mL of the heated stock community to achieve 10% (SxD1) and 1% (SxD2) concentrations of the heat stressed treatment, respectively. For an overview of all treatments, treatment levels and combinations as well as their respective inocula volumes see **Table 1**.

To assess if the targeted species were actually lost due to the specific diversity manipulations, we classified each species as at least one of the following four categories: common, rare, large, and sensitive (**Table 2**, **Figure 1**). We based these categorizations on species abundances (common or rare species) and cell sizes (large species) obtained microscopically from a 100 mL sample of the non-heated stock community. Additionally, we microscopically determined species abundances in a 100 mL sample of the heated stock community to define sensitive species. For a detailed overview of the categorizations see **Table 2**.

Because phytoplankton species often belong to more than one of the above defined four categories, overlapping species were placed as shown in **Figure 1**. Since the sample that we initially analyzed for the control treatment already represented a dilution of the total species pool present in the stock community, it did not contain the full set of phytoplankton species at the onset of the experiment. More precisely, we found single individuals of some species in initial samples of specific treatments but not in the control. We defined these species as rare (**Figure 1**). Resulting from this, we based the qualitative assessment of the initial loss of target species caused by the treatments on the total species pool of 30 species. In addition, we related the quantitative assessment of how many species were effectively lost in the treatments to the control.

TABLE 2 | Categorization of species into separate groups based on 100 mL samples analyzed of the control (non-heated stock community) and heat stressed treatment (heated stock community).


#### Sampling and Analysis

We sampled phytoplankton twice over the course of the experiment: at the onset of the experiment (initial), and after the community had reached stationary phase (bloom peak). To determine when each community had reached the stationary phase, we took daily fluorescence measurements with a fluorometer (Turner Designs 10AUTM Field Fluorometer). We applied a sigmoidal growth model (Equation 1) to confirm the exact developmental stage of the bloom:

$$f\_l = \frac{a}{1 + \left[\frac{a-b}{b}\right]e^{-\mu t}}\tag{1}$$

where f<sup>t</sup> is the relative fluorescence after t days, a is the maximum relative fluorescence (carrying capacity), b is the starting relative fluorescence, and µ is the growth rate. The treatments reached their termination point when the growth curve of at least one replicate of that treatment significantly fit the model for three consecutive days. At this time, the other replicates of the treatment had reached stationary phase and we sampled all replicates of the same treatment. We took the last samples 18 days after starting the experiment, corresponding to approximately 9–18 microalgae generations. We preserved the phytoplankton samples in Lugol's iodine solution and stored them in the dark until further processing. To determine phytoplankton cell numbers and biovolume, we counted the samples using an inverted microscope after Utermöhl (1958). For the initial samples (i.e., at the onset of the experiment), we examined a 100 mL sample of each treatment. This corresponded to cell abundances ranging from approximately 200 (D2, SxD2) to over 14,000 (Co, S) cells counted (see Supplementary Table 1). Identifications were made to the species level when

possible, otherwise genera were determined. Based on procedures described in Hillebrand et al. (1999), we calculated speciesspecific cell biovolume by approximating cell shapes to simple geometric bodies. By summing up species specific biovolumes, we calculated the total biovolume of a sample, which is used as a proxy for total biomass hereafter.

To compare phytoplankton communities, we calculated species richness, Pielou's evenness (Pielou, 1966) and Bray-Curtis dissimilarity (Bray and Curtis, 1957; Clarke et al., 2006) using the vegan package in R version 3.2.3 (Oksanen et al., 2017; R Core Team, 2017). The latter two were calculated based on species-specific biomass data. For the bloom peak samples, we computed Bray-Curtis dissimilarity both as average within (B-Cintra) and between (B-Cinter) treatments. For the initial samples, we could only calculate B-Cinter, because we did not have any replication of the inocula. Therefore, the initial samples could only be compared quantitatively. Taking multiple sub-samples at this point would have led to pseudoreplication. To compare treatments at bloom peak, we performed a multi-factorial ANOVA in R 3.2.3 (R Core Team, 2017). We tested the main effects of the categorical factors dilution, filtration and heat stress as well as the interaction of dilution and heat stress on species richness, Pielou's evenness (J'), and within-treatment variation (B-Cintra). Evenness and dissimilarity data were log-transformed to increase normality and account for non-homogenous variance distribution, respectively. We performed post-hoc tests (Tukey HSD) to specify which treatments significantly differed from one another. Finally, we created multi-dimensional scaling (MDS) plots based on the Bray-Curtis dissimilarity matrix in Primer 6 (Clarke and Gorley, 2006).

# RESULTS

# Initial Treatment Effects

All diversity manipulations initially decreased species richness by at least two species compared to the control (**Figure 2A**). We found the strongest declines in initial richness in the dilution

variation (as Bray-Curtis dissimilarity (B-Cintra); <sup>E</sup>), and Pielou's evenness (J′ , F). Filled gray dots represent treatment means (±95% CI), while open dots show the single replicates of the bloom peak samples. Asterisks denote statistically significant differences from the control detected with Tukey HSD tests (\*p < 0.05, \*\*p < 0.01, \*\*\*p < 0.001).

treatment (D1 and D2) and in the strong dilution of stressed treatment (SxD2). Compared to the control, the decreases were six and nine species, respectively (**Figure 2A**). Both filtration treatment levels, as well as the heat stressed treatment (S), reduced initial richness by two species compared to the control (**Figure 2A**).

The total species pool in the initial samples consisted of 30 species, of which 22 were present in the control treatment (Co) (**Figures 1**, **2A**, **3**). Generally, the vast majority (i.e., 22) of the species in the total pool was rare, including 12 species that were additionally large and/or sensitive (**Figures 1**, **3**).

Dilution successfully implemented the loss of rare species. In both levels of the dilution treatment (D1 and D2), 14 species were absent from the total species pool, of which 13 were initially defined as rare (**Figure 3**). Likewise, in the two levels of the dilution of stressed treatment (SxD1 and SxD2) almost only rare species were absent from the total species pool. That is, in the weak dilution of stressed treatment (SxD1), 11 species were missing from the total species pool, of which 10 were rare. In the strong dilution of stressed treatment (SxD2), 15 of the 17 missing species were rare (**Figure 3**).

Filtration only partially removed large species. In both levels (F1 and F2), five of the ten species that were missing from the total species pool were large (**Figure 3**). However, six species that we initially categorized as large species were still found in both levels of the filtration treatment (**Figure 3**). All of them appeared in low abundances in the fine filtration (F2), but some of them were frequent in the coarse filtration (F1). As a side effect, both filtration treatment levels also removed five small rare species (**Figure 3**). This is comparable to the absence of rare species in the control treatment and can be attributed to detection limits (i.e., not all species were found in the sub-sample of the initial communities).

Heat stress successfully removed sensitive species. Ten species from the total species pool were missing in the heat stressed treatment (S), of which four were sensitive (Se; **Figure 3**). Considering the fact that we initially defined five species as sensitive (**Figure 1**), a high proportion of sensitive species were effectively lost with heat stress. Similarly, in the two levels of

FIGURE 3 | Initial number of species belonging to the different categories. Species were defined as common (C), rare (R), large (L), rare large (RL), rare sensitive (RSe), large sensitive (LSe) or rare large sensitive (RLSe) as seen in Figure 1 and abbreviated as in Table 2. Opaque bars framed by black lines depict the number of species present in each category while transparent bars show categories of species missing from the total species pool of 30 species.

the dilution of heat stressed treatment, three out of five sensitive species were missing (**Figure 3**). All but one of the remaining missing species in the heat stressed treatment (S) were rare (**Figure 3**).

The treatments differentially affected initial total biomass. Reduction of total biomass in the strong dilutions (D2 and SxD2) was highest (99% compared to the control; **Figure 2B**). The heat stressed treatment (S) and the coarse filtration (F1) reduced total initial biomass by 15% compared to the control (**Figure 2B**). Reduction in the fine filtration (F2) and weak dilutions (D1 and SxD1) was intermediate ranging from 68 to 88% (**Figure 2B**).

Those treatments that substantially reduced the initial total biomass (i.e., F2, D1 SxD1, D2, SxD2; **Figure 2B**) had the biggest influence on community composition which is reflected in the alignment of the community similarities along the initial biomass dilution gradient (**Figure 4A**). Whereas, the community composition of the heat stressed treatment (S) and the coarse filtration (F1) remained very close to the control (60% similarity in community structure, **Figure 4A**), the strong dilutions (D2 and SxD2) were least similar in their community structures to all other treatments (<20% similarity, **Figure 4A**).

Initially, Thalassiosira sp. (THA) was the dominant species in most treatments according to biomass data, followed by Skeletonema costatum (SKE) and Detonula confervaceae (DET). These three species together contributed over 70% to total biomass in nearly all treatments. Only the strongly diluted treatments (D2 and SxD2) were dominated by other species. In the strong dilution (D2), Ceratium fusus (CER) was the dominant species (41% contribution to total biomass) and in the strong dilution of stressed treatment (SxD2), the primarily benthic diatom Licmophora sp. (LIC) was the most dominant species (45% biomass; **Figure 4A**, Supplementary Table 2). This was caused by the extremely large cell-size of individual CER and LIC in relation to the here concurrently diluted contribution to biomass of smaller-sized THA and DET. Though community compositions largely differed between the treatments, initial Pielou's evenness remained similar in all species loss scenarios (between 0.5 and 0.6; **Figure 2C**).

# Treatment Effects at Bloom Peak

Compared to the initial values, species richness, evenness, community composition, and hence similarity between the communities changed in all treatments, including the control. In the control, the average species richness at bloom peak was 13 (**Figure 2D**), a decrease by nine species compared to experimental onset (**Figure 2A**). Likewise, evenness at bloom peak decreased by 0.11 in the control compared to the initial value (**Figure 2C**), resulting in an average value of 0.45 (**Figure 2F**). Since a decrease in species richness and Pielou's evenness was apparent in all treatments, the control remained the most species rich and evenly distributed treatment at bloom peak (**Figures 2D,F**).

The strongly diluted treatments (D2 and SxD2), that initially led to the strongest species loss, still showed a significantly lower species richness at bloom peak compared to the control (**Figure 2D**). In the strong dilution (D2), richness declined by

represent average species composition based on biomass data. In (A), the pie charts for S, F1, and Co correspond to highly similar data points overlapping in between these three pie charts. F1 and Co of the initial samples had a similarity of more than 80%, but this could not be represented graphically. Legend with color-coded species abbreviations is valid for both, a and b. 2D stress for (A,B): 0.01.

four species and in the combination of heat stress and strong dilution (SxD2) it declined by six species (**Figure 2D**; Tukey HSD: p = 0.002 and < 0.001, respectively; for ANOVA results see **Table 3**). The latter effect seemed largely driven by the strong dilution because heat stress alone did not result in a significant species loss at bloom peak compared to the control (**Figure 2D**; Tukey HSD: p = 0.99; ANOVA results in **Table 3**). Filtration did not have a lasting significant effect on species richness at bloom peak [**Figure 2D**, Tukey HSD: p = 0.95 (F1) and p = 0.55 (F2); ANOVA results in **Table 3**].


TABLE 3 | ANOVA results showing the effects of the factors dilution, filtration, heat stress and the interaction between dilution and heat stress on species richness, Pielou's Evenness (J'), and within-treatment variation (B-Cintra) at bloom peak.

Statistically significant results (p < 0.05) are displayed in bold letters.

At bloom peak, the treatments were more equal in their species composition compared to the initial samples. Whereas, similarity of the community structure of some treatments were initially <20% (**Figure 4A**), similarity at bloom peak was more than 60% between all treatments (**Figure 4B**). This was mainly caused by the dominance of SKE that contributed at least 52% to total biomass in all treatments at bloom peak (**Figure 4B**, Supplementary Table 3).

Generally, community composition of the treatments at bloom peak was no longer mainly influenced by the dilution gradient; instead the level of heat stress had the most pronounced effect on community composition and similarity between treatments (**Figure 4B**). This effect was driven by even stronger dominance (90%) of SKE in the heat stressed treatment compared to the control (65%), and by the strong joined dominance of SKE and CHA in the dilution of stressed communities (**Figure 4B**, Supplementary Table 3). Even though heat stress did not affect Pielou's evenness initially (**Figure 2C**), the strong dominance of SKE in the heat stressed community significantly decreased evenness at bloom peak (J' = 0.16) compared to the control (J' = 0.45; **Figure 2F**; Tukey HSD: p = 0.038; ANOVA results in **Table 3**). SKE is a species that has been observed to have higher growth rates under warmer conditions and performs best in temperatures between 20 and 24◦C (Sanchez et al., 1995). The initial effects of filtration (F1 and F2) and weak dilution (D1) on community composition (**Figure 4A**) were not mirrored at bloom peak. Thus, community compositions in these treatments were similar to the control (**Figure 4B**).

#### Within-Treatment Variation

As an important side-effect, we observed high within-treatment variation (dissimilarity between replicates of a treatment, B-Cintra) in most dilutions (D1, D2, and SxD2) and in the fine filtration (F2) treatment (**Figure 2E**). Random effects, such as differing species-specific relative contribution to total biomass, or the absence and presence of certain species in replicates of the same treatment (Supplementary Figure 1), could manifest themselves in the treatments that initially decreased total biomass (**Figure 2B**). More precisely, the weak dilution (D1) had two pairs of replicates that were alike but differed from the other two replicates (Supplementary Figure 1). The strong dilution (D2) had one replicate that contained a large biomass of Thalassiosira rotula (THAR), contributing 40% to total biomass, while the other three replicates did not contain any THAR at all. In the strong dilution of stressed treatment (SxD2), two replicates had very high proportions (47 and 61%) of Chaetoceros spp. (CHA), while the other two replicates had minimal amounts of CHA (0.8 and 3.2%). In the strong filtration treatment (F2), one replicate differed from the rest in that it contained a larger proportion of THA and DET compared to the other three replicates (Supplementary Figure 1). Consequently, as stated above, B-Cintra clearly coincided with the amount of initially removed biomass in the different treatments. The only exception to this pattern was the weak dilution of stressed treatment (SxD1), in which the replicates were very homogenous at bloom peak (Supplementary Figure 1), resulting in a significantly lower B-Cintra value than in the control (**Figure 2E**).

## DISCUSSION

Using non-random species loss scenarios for experiments in community ecology is becoming increasingly important, because it allows for a more realistic approach to predict phytoplankton communities' functional reaction to changing environmental conditions. The here tested methods advance the marine sciences in two ways: first, they give the opportunity for more realistic experiments in phytoplankton community ecology and second, they allow a quantification of the hidden treatment effect due to initial variation of phytoplankton density. Certainly, these two factors are linked—in that the dilutive manipulations always alter both species richness and species identities simultaneously.

Essentially, all three treatments (dilution, heat stress, and filtration) reduced richness of the targeted species groups. That is, dilution effectively removed rare (decrease by 93% compared to the total species pool) and heat stress sensitive species (decrease by 80% compared to the total species pool). Filtration removed some of the large species, but the success rate was only 50% removal compared to the total species pool.

### Effects of Dilution

Successful decrease of species richness due to dilution at the onset of the bloom (**Figure 2A**) translated into lower species richness at bloom peak (**Figure 2D**), but did not significantly influence Pielou's evenness at bloom peak (**Figure 2F**). Since rare species are statistically more likely to be excluded from a smaller inoculum, we expected an initial decrease in species richness. Problematic is that even though dilution lead to the effective loss of rare species, it was not possible to control which species were lost in replicates of the same treatment. This lead to a high variability of species present at bloom peak, causing more heterogeneous responses in species composition and evenness between replicates of the same treatment (**Figure 2F**, Supplementary Figure 1). In experiments studying the effect of species loss on ecosystem processes, the high variability of species causes problems, because the effects observed in these experiments cannot clearly be attributed to the decrease in richness. It could just as likely be an effect of species identity carrying certain functional traits. Furthermore, in many experiments—especially those focusing on species traits—species identity is of great importance. For example, in grazing experiments it is essential to distinguish whether edible or inedible species are lost. In certain climate change studies, it is of great interest if calcifying coccolithophores or silicifying diatoms are lost. In these cases, our methods of non-random species loss should be modified (e.g., by using other stress factors) or might not be suited at all.

To eliminate the problem of not only decreasing species numbers, but also losing random species traits in different replicates of the same treatment, the inoculum size could be adjusted so that starting densities between treatments would be more comparable. In culture experiments with bacteria, this is already practiced (Franklin et al., 2001; Hol et al., 2015). For this, the diluted community is allowed to grow for several days such that an increase in biomass can be achieved before the start of the experiment. However, to the best of our knowledge, this method has not been used for phytoplankton and therefore should be tested experimentally. It has to be confirmed that rare phytoplankton species are actually removed from the culture in the long run and cannot grow back to original numbers if the inoculum is re-grown.

Nevertheless, for experiments that only focus on species richness, dilution is a good method to create non-randomly assembled communities. With dilution, distinct gradients of species richness can be created (Roger et al., 2016; Hammerstein et al., 2017), which is a large improvement over traditional culture experiments with randomly assembled communities. Furthermore, some experimental ecologists have used this side effect of dilution to their advantage. Trommer et al. (2012) for example, aimed at reaching high variation within treatments to increase ecological noise and make the community response broader. In addition to being an experimental manipulation method, the dilution treatment can be useful for other applications, such as long-term phytoplankton evolution experiments. In these experiments, researchers use semicontinuous culturing approaches where each new culture is established as a (diluted) inoculum of the former (Lohbeck et al., 2012; Schlüter et al., 2014). For these types of experiments, the tested levels of dilution allow to estimate the potential loss of genotypes at each step.

In general, dilution can be used as a standard method to lose rare species or genotypes in natural phytoplankton communities, for experiments that only manipulate richness. For experiments with an additional focus on species identities, some adjustments should be made to the dilution method and within-treatment variation should be considered.

#### Effects of Heat Stress

Heat stress successfully decreased species richness of sensitive species (**Figure 2A**). This did not have an immediate effect on community composition (**Figure 4A**), but was reflected at bloom peak by the extreme dominance of a warm-adapted species (**Figure 4B**). For this manipulation it is helpful to have a good knowledge of the community present at the onset of the experiment, because that allows for fine-tuning of the heat stress treatment (e.g., setting a specific target temperature or exposing species to a certain temperature for a specific time frame). Other methods to target different types of sensitive species are also feasible. Hammerstein et al. (2017) used a mechanical disturbance by shaking the cultures to create communities with distinct gradients of species richness with the goal to lose sensitive species. Summarizing, heat stress can be one successful method to lose sensitive species in natural phytoplankton communities.

#### Effects of Filtration

In this study, fine filtration (20µm) successfully decreased initial richness (**Figure 2A**), and similarly to the dilution treatment, reduced initial biomass (**Figure 2B**). The observed effects on within-treatment variability at bloom peak were also comparable to dilution (**Figure 2E**). This problem can easily be mitigated by adjusting the inoculum volume of the filtration treatment so that it contains comparable phytoplankton densities to the control.

A disadvantage of the filtration method was that it did not completely remove all large species. Many large species are long and thin, which means that individuals could still pass through the sieve if they reach it in the right angle. This can create a bias in size class categorization (Graham and Jones, 2007). Sieving the samples multiple times could minimize this problem. In contrast to this, cells of some species that were not defined as large were filtered out in this treatment. This included genera like Skeletonema, Chaetoceros, and Detonula, which are chain-forming and therefore accumulate to a greater size (Round et al., 1990). The individual cell size, however, can be smaller than the colony size. In this case, it has to be clarified whether an individual cell or the entire colony is categorized in a specific size class. Related to this, the special morphology of some genera, including Chaetoceros, can also lead to a size class bias. Chaetoceros cells themselves may not have an average dimension exceeding 20µm, but many species of this genus are conspicuous due to long spines protruding from the valves (Round et al., 1990). We also observed this in the present study: the overall largest cell dimension including spines often exceeded 20µm and therefore fewer cells of this genus were found in the filtration treatment. The morphology and life history of species has to be considered when targeting specific species for removal.

Another challenge to the accuracy of the filtration treatment in our experiment is based on the definition for large species that we employed (i.e., the largest cell dimension exceeds 20µm). We chose this definition due to the scarcity of species exceeding 100µm in size. We still found several species initially categorized as large in the coarse filtration treatment (F1), for which we had used the 100µm sieve (**Figure 2**). In other studies, the definition of large species might be chosen differently as there may be predominantly species in larger size classes, depending on the sampling location and time of the year.

If all these complications are considered, filtration can be a successful method to test the ecosystem consequences of manipulating the "master trait" cell size in experimental phytoplankton communities.

#### CONCLUSION

In general, the tested manipulations of non-random species loss allow to assess the effects of realistic phytoplankton diversity change. These methods can be used to manipulate species composition for a wide variety of biodiversity experiments independent of their scale (i.e., micro- and mesocosm) or location (i.e., outdoor and indoor). However, researchers should take special care to avoid the pitfalls of the dilution and filtration methods, especially when species identities are important for the experimental outcome. All in all, these methods are a valuable improvement for hitherto artificially assembled phytoplankton communities in biodiversity—ecosystem functioning experiments, simply because non-random species loss is more realistic (Gamfeldt et al., 2015).

# REFERENCES


#### AUTHOR CONTRIBUTIONS

BM, AL, and SE conceived the study. FE conducted the experiment and analyzed the data. FE, BM, AL, and SE wrote the manuscript. All authors agree on the final version of this manuscript and agree to be accountable for all aspects of this work.

#### FUNDING

The research was funded by internal grants from the GEOMAR Helmholtz Centre for Ocean Research Kiel.

#### ACKNOWLEDGMENTS

We thank Thomas Hansen and Bente Gardeler for technical assistance and two reviewers for their helpful comments on the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2017.00317/full#supplementary-material

effects. Trends Ecol. Evol. 31, 136–146. doi: 10.1016/j.tree.2015. 12.009


Zavaleta, E. S., and Hulvey, K. B. (2004). Realistic species losses disproportionately reduce grassland resistance to biological invaders. Science 306, 1175–1177. doi: 10.1126/science.1102643

**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 Engel, Lewandowska, Eggers and Matthiessen. 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.