# THE RESPONSES OF MARINE MICROORGANISMS, COMMUNITIES AND ECOFUNCTIONS TO ENVIRONMENTAL GRADIENTS

EDITED BY : Hongyue Dang, Martin G. Klotz, Charles Lovell and Stefan M. Sievert PUBLISHED IN : Frontiers in Microbiology

#### 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 978-2-88945-807-3 DOI 10.3389/978-2-88945-807-3

#### About Frontiers

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

#### Frontiers Journal Series

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

#### Dedication to Quality

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

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

#### What are Frontiers Research Topics?

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

# THE RESPONSES OF MARINE MICROORGANISMS, COMMUNITIES AND ECOFUNCTIONS TO ENVIRONMENTAL GRADIENTS

Topic Editors:

Hongyue Dang, Xiamen University, China Martin G. Klotz, Washington State University, United States Charles Lovell, University of South Carolina, United States Stefan M. Sievert, Woods Hole Oceanographic Institution, United States

Sunrise over the tropical Western Pacific Ocean. The photo was taken during the 2006 summer Chinese-French joint MD155/MARCO POLO 2/ IMAGES XIV cruise on R/V Marion Dufresne. Image: Hongyue Dang.

Marine environments are fluid. Microorganisms living in the ocean experience diverse environmental changes over wide spatiotemporal scales. For microorganisms and their communities to survive and function in the ocean, they need to have the capacity to sense, respond to, adapt to and/or withstand periodic and sporadic environmental changes. This eBook collates a variety of recent research reports and theoretical discussions on the ecoenergetic strategies, community structure, biogeochemical and ecosystem functions as well as regulatory processes and mechanisms that marine microorganisms employ in response to environmental gradients and variations.

Citation: Dang, H., Klotz, M. G., Lovell, C., Sievert, S. M., eds. (2019). The Responses of Marine Microorganisms, Communities and Ecofunctions to Environmental Gradients. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-807-3

# Table of Contents

*07 Editorial: The Responses of Marine Microorganisms, Communities and Ecofunctions to Environmental Gradients*

Hongyue Dang, Martin G. Klotz, Charles R. Lovell and Stefan M. Sievert

#### CHAPTER I

#### MICROBIAL ECOENERGETIC RESPONSES TO ENERGY SOURCES AND DYNAMICS

*12 Radiative Energy Budgets of Phototrophic Surface-Associated Microbial Communities and Their Photosynthetic Efficiency Under Diffuse and Collimated Light*

Mads Lichtenberg, Kasper E. Brodersen and Michael Kühl

*29 Ecological Energetic Perspectives on Responses of Nitrogen-Transforming Chemolithoautotrophic Microbiota to Changes in the Marine Environment* Hongyue Dang and Chen-Tung A. Chen

#### CHAPTER II

#### MICROBIAL COMMUNITY RESPONSES TO NATURAL AND ANTHROPOGENIC GRADIENTS AND THEIR IMPACTS ON MARINE C, N, S AND FE CYCLING

*52 Seasonal Succession and Spatial Patterns of* Synechococcus *Microdiversity in a Salt Marsh Estuary Revealed Through 16S rRNA Gene Oligotyping*

Katherine R. M. Mackey, Kristen Hunter-Cevera, Gregory L. Britten, Leslie G. Murphy, Mitchell L. Sogin and Julie A. Huber

*63* Synechococcus *Assemblages Across the Salinity Gradient in a Salt Wedge Estuary*

Xiaomin Xia, Wang Guo, Shangjin Tan and Hongbin Liu

*75 Patterns and Processes in Marine Microeukaryotic Community Biogeography From Xiamen Coastal Waters and Intertidal Sediments, Southeast China*

Weidong Chen, Yongbo Pan, Lingyu Yu, Jun Yang and Wenjing Zhang

*89 Distinct Seasonal Patterns of Bacterioplankton Abundance and Dominance of Phyla* a*-*Proteobacteria *and* Cyanobacteria *in Qinhuangdao Coastal Waters Off the Bohai Sea*

Yaodong He, Biswarup Sen, Shuangyan Zhou, Ningdong Xie, Yongfeng Zhang, Jianle Zhang and Guangyi Wang

*104 Long-Term Survey is Necessary to Reveal Various Shifts of Microbial Composition in Corals*

Shan-Hua Yang, Ching-Hung Tseng, Chang-Rung Huang, Chung-Pin Chen, Kshitij Tandon, Sonny T. M. Lee, Pei-Wen Chiang, Jia-Ho Shiu, Chaolun A. Chen and Sen-Lin Tang

*115 Community Composition and Transcriptional Activity of Ammonia-Oxidizing Prokaryotes of Seagrass* Thalassia hemprichii *in Coral Reef Ecosystems*

Juan Ling, Xiancheng Lin, Yanying Zhang, Weiguo Zhou, Qingsong Yang, Liyun Lin, Siquan Zeng, Ying Zhang, Cong Wang, Manzoor Ahmad, Lijuan Long and Junde Dong

*128 Marine Group II Dominates Planktonic Archaea in Water Column of the Northeastern South China Sea*

Haodong Liu, Chuanlun L. Zhang, Chunyan Yang, Songze Chen, Zhiwei Cao, Zhiwei Zhang and Jiwei Tian

*139 Evaluating Production of Cyclopentyl Tetraethers by Marine Group II*  Euryarchaeota *in the Pearl River Estuary and Coastal South China Sea: Potential Impact on the TEX86 Paleothermometer* Jin-Xiang Wang, Wei Xie, Yi Ge Zhang, Travis B. Meador and

Chuanlun L. Zhang


Kai Tang, Yao Zhang, Dan Lin, Yu Han, Chen-Tung A. Chen, Deli Wang, Yu-Shih Lin, Jia Sun, Qiang Zheng and Nianzhi Jiao


Beverly K. Chiu, Shingo Kato, Sean M. McAllister, Erin K. Field and Clara S. Chan

*255 Analysis of Bacterial Community Composition of Corroded Steel Immersed in Sanya and Xiamen Seawaters in China via Method of Illumina MiSeq Sequencing*

Xiaohong Li, Jizhou Duan, Hui Xiao, Yongqian Li, Haixia Liu, Fang Guan and Xiaofan Zhai

#### CHAPTER III

#### REGULATION OF MICROBIAL RESPONSES TO ENVIRONMENTAL GRADIENTS AND VARIATIONS

#### *271 Genome-Wide Detection of Small Regulatory RNAs in Deep-Sea Bacterium* Shewanella piezotolerans *WP3* Muhammad Z. Nawaz, Huahua Jian, Ying He, Lei Xiong, Xiang Xiao and Fengping Wang

#### *283 Biofilm Formation and Heat Stress Induce Pyomelanin Production in Deep-Sea* Pseudoalteromonas *sp. SM9913*

Zhenshun Zeng, Xingsheng Cai, Pengxia Wang, Yunxue Guo, Xiaoxiao Liu, Baiyuan Li and Xiaoxue Wang

# Editorial: The Responses of Marine Microorganisms, Communities and Ecofunctions to Environmental Gradients

Hongyue Dang<sup>1</sup> \*, Martin G. Klotz 1,2, Charles R. Lovell <sup>3</sup> and Stefan M. Sievert <sup>4</sup>

<sup>1</sup> State Key Laboratory of Marine Environmental Science, Institute of Marine Microbes and Ecospheres, and College of Ocean and Earth Sciences, Xiamen University, Xiamen, China, <sup>2</sup> School of Molecular Biosciences, Washington State University, Richland, WA, United States, <sup>3</sup> Department of Biological Sciences, University of South Carolina, Columbia, SC, United States, <sup>4</sup> Biology Department, Woods Hole Oceanographic Institution, Woods Hole, MA, United States

Keywords: marine microbiology, microbial ecology, biogeochemical cycles, environmental gradients, global change, ocean acidification, greenhouse gases

**Editorial on the Research Topic**

#### **The Responses of Marine Microorganisms, Communities, and Ecofunctions to Environmental Gradients**

#### Edited by:

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

#### Reviewed by:

Klaus Jürgens, Leibniz Institute for Baltic Sea Research (LG), Germany

> \*Correspondence: Hongyue Dang danghy@xmu.edu.cn

#### Specialty section:

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

Received: 03 November 2018 Accepted: 18 January 2019 Published: 08 February 2019

#### Citation:

Dang H, Klotz MG, Lovell CR and Sievert SM (2019) Editorial: The Responses of Marine Microorganisms, Communities and Ecofunctions to Environmental Gradients. Front. Microbiol. 10:115. doi: 10.3389/fmicb.2019.00115 From estuaries to marginal seas and open oceans, from tropical warm pools to subtropical gyres and polar cryospheres, from sunlit surface water to twilight zone and pitch-black abyssopelagic water, from water columns to sediments and deep subseafloor biospheres, marine ecosystems experience diverse environmental gradients (Karl, 2007; Dang and Jiao, 2014). In addition to these large-scale gradients, small-scale, and micro-scale gradients of various physicochemical factors are common in the ocean; in particular, in marginal seas and coastal environments (Kappler et al., 2005; Stocker, 2012). The diverse gradients of physicochemical parameters, nutrients, and chemicals serving as electron donors and acceptors contribute to the creation of habitat heterogeneity and novel locales along a gradient may create unique niches for any given microorganism. Whether at the surface of a marine snow particle or alga, at the edges of an oxygen minimum zone (OMZ), in marginal sea methane-seep sediments, or on a chimney wall of a deep-sea hydrothermal vent, these interfaces provide hotspot habitats with sharp physicochemical gradients that may host diverse yet unknown microorganisms that facilitate yet unknown biogeochemical processes (Hügler and Sievert, 2011; Wright et al., 2012; Dang and Lovell, 2016). With the progress of marine molecular microbial ecology and "omics" techniques, certain environmental keystone microorganisms have been discovered at some of these interfaces: such as the anaerobic methane-oxidizing (ANME) archaea in methane-rich sediments (Valentine and Reeburgh, 2000), cable bacteria that facilitate electrogenic sedimentary sulfide oxidation (Nielsen and Risgaard-Petersen, 2015), neutrophilic zetaproteobacterial iron-oxidizing bacteria (FeOB) in deep-sea hydrothermal microbial mats and at abyssal basaltic glass-seawater and coastal metal-seawater interfaces (Emerson et al., 2010; Dang et al., 2011; Henri et al., 2016), anaerobic ammonium-oxidizing (anammox) bacteria and SUP05 sulfur-oxidizing bacteria (SOXB) in coastal and oceanic OMZs (Dick et al., 2013; Oshiki et al., 2016), and sulfur-oxidizing and/or hydrogen-oxidizing Campylobacteria in the proposed new phylum Campylobacterota (formerly known as Epsilonproteobacteria; Waite et al., 2018) at seawater, hydrothermal vent, and subseafloor redox interfaces (Campbell et al., 2006; Grote et al., 2012; Dick et al., 2013; Han and Perner, 2015; McNichol et al., 2018). Even the ubiquitous marine ammonia-oxidizing Thaumarchaeota, discovered only a decade ago (Könneke et al., 2005), can now be divided into two distinct ecological groups according to the vertical physicochemical profile of marine water, the "shallow clade" and the "deep clade" (Hatzenpichler, 2012). The ongoing discovery of unique ecophysiological functions of marine Bacteria and Archaea will contribute to a conceptual rewriting of biogeochemical pathways in the marine C, N, S, and Fe cycles.

The characterization of how the abundance and spatial distribution of marine microorganisms, the structure of microbial communities and their provided ecosystem functions respond to the diverse environmental gradients is of fundamental importance to our understanding of the microbial ecology and biogeochemistry of the oceans. This rationale defines the aim and scope of this Research Topic. The contributions of environmental gradients to the diversity of marine microorganisms and their metabolic potentials may play important roles in maintaining the stability and functions of the estuarine, coastal and marginal sea ecosystems, which have been experiencing a multitude of anthropogenic perturbations (Dang and Jiao, 2014; Damashek and Francis, 2018). The responses of the affected microbial communities to human-induced environmental impacts are currently still difficult to predict and the understanding of microbial processes and mechanisms at the community level is the key for predictive modeling, which also requires the collection of large empirical data sets (Haruta et al., 2013; Hanemaaijer et al., 2015; Burd et al., 2016). Greater understanding of microbial responses to natural and anthropogenic environmental gradients may also help us understand the responses of marine ecosystems to global climate change and other large-scale environmental perturbations such as ocean acidification and spatial and temporal ocean deoxygenation.

The authors of this Research Topic contributed a total of 21 publications covering a wide variety of subjects spanning from microbial metabolic dynamics to biogeochemical cycling of C, N, S, and Fe in micro-, small-, and geographic-scale marine gradients. This Editorial aims to highlight some of the main findings reported in these publications and we would like to take this opportunity to thank all participating editors and reviewers for making this Research Topic a success. In order to cover the broad subjects of the articles published in this Research Topic, we organize our introductions to these publications in the following sequence: (1) microbial ecoenergetic strategies and dynamics in response to energy sources and dynamics; (2) microbial community structure variations and their impacts on marine C, N, S, and Fe cycling in response to natural and anthropogenic gradients; and (3) microbial regulatory processes and mechanisms in response to environmental gradients and variations. We closed this Editorial by referring to the need for ongoing experimental advancement in future studies of marine microorganisms, their communities and ecofunctions, in response to marine environmental changes.

#### MICROBIAL ECOENERGETIC RESPONSES TO ENERGY SOURCES AND DYNAMICS

Sunlight is the major energy source that supports most of the primary production in surface oceans and in sediments under shallow water; hence, light availability controls the vertical distribution of photosynthetic communities in the ocean. However, the photochemical energy conversion efficiency of the phytoplankton communities was recently found to be relatively low and further limited by nutrient scarcity in vast regions of the ocean (Lin et al., 2016; Falkowski et al., 2017). Lichtenberg et al. investigated radiative energy budgets and energy conversion efficiency of benthic phototrophic microbial communities in coral reef sediments and a cyanobacterial biofilm. They found that local photosynthetic efficiencies change as function of physical structure of microbial communities and gradients of diffuse or collimated light further change the pattern of radiative energy conversion. In addition to light energy, energy stored in chemical bonds is also explored by microorganisms for carrying out dark inorganic carbon fixation (Hügler and Sievert, 2011), which may play an important role in marine carbon cycling and climate modulation. For example, in a very recent investigation in the South China Sea, the integrated water column dark carbon fixation rate was estimated to be nearly 4 fold of euphotic zone primary production (Zhou et al., 2017). The review article by Dang and Chen further discussed the eco-energetic strategies of key marine chemolithoautotrophic nitrogen-cycling microorganisms and their tentative responses to marine environmental changes such as those caused by global warming, ocean acidification, deoxygenation, eutrophication, and heavy metal pollutions.

#### MICROBIAL COMMUNITY RESPONSES TO NATURAL AND ANTHROPOGENIC GRADIENTS AND THEIR IMPACTS ON MARINE C, N, S AND FE CYCLING

Cyanobacteria are key primary producers in the ocean. Mackey et al. revealed variations in the oligotypes of Synechococcus strains and thus their microdiversity, relative abundances and niche differentiation in response to changes in season, and salinity in a salt marsh estuary by employing a novel molecular approach called oligotyping. Similarly, Xia et al. verified niche partitioning among Synechococcus in the Pearl River estuary, a salt wedge estuary of the South China Sea. Chen et al. showed that the biogeography of dominant planktonic and benthic microeukaryotic communities (possibly including autotrophs, heterotrophs, and mixotrophs) may be influenced mainly by environmental and spatial factors, while that of the rare subcommunities may be influenced by more complex mechanisms in the coastal environment of Xiamen, China.

Marine Roseobacter clade (MRC) bacteria are abundant as free-living and particle-associated microorganisms; particularly, in coastal waters, and some of them can carry out aerobic anoxygenic photosynthesis (Dang and Lovell, 2002, 2016; Buchan et al., 2014). He et al. investigated the seasonal and spatial distribution of the bacterioplankton communities in highly anthropogenically impacted Qinhuangdao coastal waters and reported that the bacterial abundance had significant positive correlation with seawater total phosphorus content, potentially serving as a key monitoring parameter for anthropogenic impact in the studied aquatic area. These authors also observed an inverse correlation between the dominant Family II Cyanobacteria and Alphaproteobacteria (mainly affiliated with the MRC). It will be worthwhile to further investigate what the ecological mechanism or controlling environmental factors are, if any, that determine the distinct spatial distribution of Cyanobacteria and MRC bacteria.

Shallow-water coral reefs are among the most productive and most diverse symbiotic ecosystems in the oceans (Cunning and Baker, 2014; Blackall et al., 2015; Peixoto et al., 2017). The response of coral microbiomes to environmental disturbance is highly complex (McDevitt-Irwin et al., 2017). Long-term surveys are critical to our ability to differentiate changes in response to anthropogenic disturbances from natural dynamics of the coral microbiomes. The work by Yang et al. highlights the importance of long-term surveys for coral microbial communities in revealing compositional shifts and environmental correlations and reported that the dominant bacterial groups in coral Stylophora pistillata showed differential geographical preference, whereas the composition of the minor bacterial members in S. pistillata fluctuated over time.

Although Archaea have been recognized as an important and diverse group of microorganisms in the ocean, knowledge gaps concerning the ecological and biogeochemical roles of many archaeal lineages remain (Offre et al., 2013; Spang et al., 2017). Ling et al. investigated chemolithoautotrophic ammonia-oxidizing Thaumarchaeota communities, along with communities of ammonia-oxidizing Betaproteobacteria, that were associated with the seagrass Thalassia hemprichii in several coral reef ecosystems of the South China Sea. Liu et al. reported much more abundant heterotrophic MG-II Euryarchaeota than chemolithoautotrophic Thaumarchaeota throughout the water column of the northeastern South China Sea and strong water mixing was inferred to be the cause of this unusual distribution pattern of the marine archaea. Furthermore, Wang et al. found that MG-II Euryarchaeota likely produce a large proportion of GDGTs, potentially important in marine carbon cycling (Zhang et al., 2015) and in revising the interpretation of TEX86, a paleotemperature proxy stored in marine sediments.

The global nitrogen cycle has been experiencing tremendous anthropogenic disturbances (Rockström et al., 2009). Whether nitrogen loss through denitrification and anammox, and nitrogen gain through microbial N<sup>2</sup> fixation are presently still in balance in the anthropogenically-impacted modern ocean has been vigorously debated (e.g., Zhou et al., 2016; Dang and Chen). New diazotrophs and N2-fixing environments have recently been identified, including coastal sediments that harbor diverse and abundant sulfate-reducing bacteria (SRB) that are active in nitrogen fixation (Bertics et al., 2013; Dang et al., 2013; Pedersen et al., 2018; Zhou et al., 2016). The work by Zhang et al. showed the diazotrophic potential of SRB in the rhizospheres of tropical mangroves, which usually constitute highly productive intertidal ecosystems but meanwhile lack sufficient nutrients. Another environmental issue related to the contemporary nitrogen cycle is that N2O has emerged as the top ozone-destructing greenhouse gas (Voss et al., 2013). Microbial reduction is likely the sole biological sink for N2O, and the key enzyme in this process, nitrous oxide reductase, is known for its low oxygen tolerance (Bonin et al., 1989; Körner and Zumft, 1989). Nevertheless, Sun et al. reported in this Research Topic that composition of the active N2Oconsuming microbial assemblages varied with seawater N2O but not O<sup>2</sup> concentration across the oxic/anoxic gradient of the Eastern Tropical South Pacific Ocean. This work also tentatively identified an overlooked N2O sink by showing the presence of active N2O-consuming microorganisms in oxygenated surface seawater.

Many microorganisms participate in the marine sulfur and iron cycles via dissimilatory metabolism (Sievert et al., 2007; Melton et al., 2014). S-cycling bacteria and archaea contribute to either organic carbon consumption (via anaerobic respiration) or inorganic carbon fixation (via chemolithoautotrophy), depending on the in situ redox status and the available energy metabolic substrates. Jiang et al. characterized the versatile physiology and metabolic mechanisms of Hydrogenovibrio thermophilus strain S5, a chemolithomixotrophic hydrogen- and sulfur-oxidizing bacterium isolated from an active hydrothermal vent chimney on the Southwest Indian Ridge. The versatility of this bacterium in energy and carbon source exploitation enables its survival in the highly dynamic and harsh conditions of the deep-sea hydrothermal environments. Tang et al. investigated the microbial communities of the shallow-sea hydrothermal system off Kueishantao Island. They not only detected sulfur oxidation and carbon fixation marker gene sequences in their metagenome datasets, but also identified the signatures of many heterotrophic bacteria that harbored versatile genetic potential to adapt to the shallow-sea hydrothermal environment. Zhang et al. investigated the vertical distribution of SRB and SOXB in natural sediments of the East China Sea, a marginal sea highly impacted by riverine and anthropogenic activities. Qiao et al. investigated the mud deposit bacterial communities of the eastern China marginal seas including the East China Sea and they also quantified the dsrB gene abundance attributed to SRB. The work by Ihara et al. showed the successional dynamics of bacterial communities in marine sediments launched on land by earthquake-induced tsunami and identified campylobacterial SOXB as pivotal microbes during community and functional shift. This work also found the involvement of zetaproteobacterial and betaproteobacterial FeOB in sediment bacterial community succession, verifying the prevalence of FeOB in sedimentary environments of the global coastal seas (McBeth et al., 2011; Laufer et al., 2017). Chiu et al. identified two new pelagic zetaproteobacterial FeOB species from seawater of the Chesapeake Bay oxic-anoxic transition zone and—based on in silico genome sequence analysis—inferred their strategies for adaptation to planktonic and putative particle-associated living in aquatic environments, thereby supporting a previous finding that coastal seawater may commonly harbor biofilm-forming and biocorrosion-causing zetaproteobacterial FeOB (Dang et al., 2011). FeOB have been hypothesized as pioneer species in the initiation of carbon steel biocorrosion in marine environments, while SRB may play more important roles in biocorrosion once the biocorroding microbial communities grow into thick biofilms (Dang et al., 2011). Li et al., indeed, showed the dominance of SRB in the rust microbial communities that formed from long-term steel incubations in coastal waters.

#### REGULATION OF MICROBIAL RESPONSES TO ENVIRONMENTAL GRADIENTS AND VARIATIONS

The microbial responses to environmental gradients and variations are usually highly regulated. Nawaz et al. showed the importance of small regulatory RNAs in the adaptation to deep-sea conditions in Shewanella piezotolerans WP3, an iron-reducing bacterium with identified piezotolerance and psychrotolerance. Furthermore, the work by Zeng et al. showed a novel molecular mechanism of Pseudoalteromonas sp. SM9913, a biofilm-forming marine bacterium, in adaptation to heat stress.

Coevolution of the Earth and its microbiota dictate the capability of individual microorganisms and their communities to respond to environmental changes. Although the mechanisms and processes employed by microbes to respond to environmental changes are highly diverse and complex, established biological and ecological principles are followed and can thus be decoded as suggested by the studies in this Research Topic. The general lack of available representative microbes in culture as well as experimental model systems to simulate environmental gradients presents an ongoing challenge to gaining deeper understanding of the processes, mechanisms and

#### REFERENCES


functions in changing marine ecosystems (Lage and Bondoso, 2012; Thøgersen et al., 2018). Continuing advancement of experimental techniques and protocols, such as those with high sampling frequency and sufficient replicates, long-term surveys, deep sequencing, systematic analyses and modeling will eventually help to reveal the mysteries of the microbial world in aquatic systems.

## AUTHOR CONTRIBUTIONS

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

#### FUNDING

This Research Topic was supported by the National Key Research and Development Program of China grant 2016YFA0601303, China Ocean Mineral Resources R&D Association grant DY135-E2-1-04, China SOA grant GASI-03-01-02-05, NSFC grants 41676122, 91328209, and 91428308, and CNOOC grant CNOOC-KJ125FZDXM00TJ001-2014.

#### ACKNOWLEDGMENTS

We thank the Frontiers in Microbiology and Frontiers in Marine Sciences editorial staff for their initial invitation and professional support throughout. We also thank the peer-review team for their insightful comments and suggestions.


**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 Dang, Klotz, Lovell and Sievert. 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.

# Radiative Energy Budgets of Phototrophic Surface-Associated Microbial Communities and their Photosynthetic Efficiency Under Diffuse and Collimated Light

#### Mads Lichtenberg<sup>1</sup> \* † , Kasper E. Brodersen1 † and Michael Kühl 1, 2

<sup>1</sup> Marine Biological Section, Department of Biology, University of Copenhagen, Helsingør, Denmark, <sup>2</sup> Climate Change Cluster, University of Technology Sydney, Ultimo, NSW, Australia

#### Edited by:

Stefan M. Sievert, Woods Hole Oceanographic Institution, USA

#### Reviewed by:

Markus Huettel, Florida State University, USA Lucas Stal, Royal Netherlands Institute for Sea Research (NWO), Netherlands

\*Correspondence:

Mads Lichtenberg mads.lichtenberg@bio.ku.dk † These authors have contributed equally to this work.

#### Specialty section:

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

Received: 27 January 2017 Accepted: 03 March 2017 Published: 28 March 2017

#### Citation:

Lichtenberg M, Brodersen KE and Kühl M (2017) Radiative Energy Budgets of Phototrophic Surface-Associated Microbial Communities and their Photosynthetic Efficiency Under Diffuse and Collimated Light. Front. Microbiol. 8:452. doi: 10.3389/fmicb.2017.00452 We investigated the radiative energy budgets of a heterogeneous photosynthetic coral reef sediment and a compact uniform cyanobacterial biofilm on top of coastal sediment. By combining electrochemical, thermocouple and fiber-optic microsensor measurements of O2, temperature and light, we could calculate the proportion of the absorbed light energy that was either dissipated as heat or conserved by photosynthesis. We show, across a range of different incident light regimes, that such radiative energy budgets are highly dominated by heat dissipation constituting up to 99.5% of the absorbed light energy. Highest photosynthetic energy conservation efficiency was found in the coral sediment under low light conditions and amounted to 18.1% of the absorbed light energy. Additionally, the effect of light directionality, i.e., diffuse or collimated light, on energy conversion efficiency was tested on the two surface-associated systems. The effects of light directionality on the radiative energy budgets of these phototrophic communities were not unanimous but, resulted in local spatial differences in heat-transfer, gross photosynthesis, and light distribution. The light acclimation index, Ek, i.e., the irradiance at the onset of saturation of photosynthesis, was >2 times higher in the coral sediment compared to the biofilm and changed the pattern of photosynthetic energy conservation under light-limiting conditions. At moderate to high incident irradiances, the photosynthetic conservation of absorbed energy was highest in collimated light; a tendency that changed in the biofilm under sub-saturating incident irradiances, where higher photosynthetic efficiencies were observed under diffuse light. The aim was to investigate how the physical structure and light propagation affected energy budgets and light utilization efficiencies in loosely organized vs. compact phototrophic sediment under diffuse and collimated light. Our results suggest that the optical properties and the structural organization of phytoelements are important traits affecting the photosynthetic efficiency of biofilms and sediments.

Keywords: biofilm, canopy, coral reef sediment, diffuse and collimated light, heat dissipation, light use efficiency, microsensors, photosynthesis

## INTRODUCTION

Photosynthetic sediments and biofilms are characterized by pronounced vertical stratification of the microbial environment as a result of steep light gradients, high metabolic activity and limitations of heat and solute transport by diffusion (Kühl et al., 1996; Kühl and Fenchel, 2000; Al-Najjar et al., 2012). The radiative energy balance in such phototrophic microbial communities is affected by the incident radiative energy from the sun, of which a fraction is backscattered and thus not absorbed, while absorbed light energy is either photochemically conserved via photosynthesis or dissipated as heat via radiative energy transfer and non-photochemical quenching (NPQ; Al-Najjar et al., 2010; Brodersen et al., 2014). The quantity and quality of light are the main controlling factors of photosynthesis, and the microscale distribution of light in microphytobenthic systems has been studied intensively over the last decades (Jørgensen and Des Marais, 1988; Lassen et al., 1992a; Kühl and Jørgensen, 1994; Kühl, 2005). A sub-saturating flux of photons will limit the rate of photosynthesis, as the available light is insufficient to support the maximal potential rate of the light reactions. But as the photon flux increases, photosynthesis saturates, whereby O<sup>2</sup> becomes a competitive inhibitor on the binding-site of CO<sup>2</sup> to Ribulose-1,5-bisphosphate carboxylase oxygenase (Rubisco; Falkowski and Raven, 2007). In addition, when light energy absorption exceeds the capacity for light utilization, excess energy is channeled into heat production via NPQ processes to avoid degradation of pigments and other cell constituents e.g., by reactive singlet oxygen produced by the de-excitation of triplet state chlorophyll ( <sup>3</sup>Chl<sup>∗</sup> ) (Müller et al., 2001).

Photosynthetic organisms deploy different mechanisms to avoid photo-damage, where NPQ is an effective short-term solution to dispose of excess energy (Müller et al., 2001). If a photosynthetic cell experiences high light conditions on a daily basis, long-term regulation can be achieved by regulating light harvesting pigment composition and concentration (Nymark et al., 2009). One strategy is to lower the light harvesting pigment content to decrease the absorption cross section by increasing transmittance, while another strategy involves upregulation of photoprotective pigments such as xanthophylls, that absorb energy-rich blue-green light but quench non-photochemically (Zhu et al., 2010).

Since photosynthetic cells perceive light from all directions, the light field angularity is important for determining the total irradiance experienced by a cell (Kühl and Jørgensen, 1994), and it has e.g., been shown that the incident light geometry can influence photosynthetic light use efficiencies and photoinhibition in terrestrial plant canopies (Gu et al., 2002). In sediments, incident light will be spread by multiple scattering and, while the light field will become entirely diffuse with depth (Kühl and Jørgensen, 1994), the response of benthic photosynthetic organisms to incident diffuse light is unknown. Through evaporation, an increase in cloud-cover has been predicted with global warming (Schiermeier, 2006), which will potentially change the direction of light from relatively collimated beams (∼85% in clear-sky conditions) to a more isotropic diffuse light field (∼100% in cloud covered conditions; Bird and Riordan, 1986; Brodersen et al., 2008; Gorton et al., 2010). In addition, submerged benthic systems will experience temporal and spatial differences in light field isotropy depending on turbidity, water depth, sun angle, and the reflective properties of the surrounding environment (Brakel, 1979; Kirk, 1994; Wangpraseurt and Kühl, 2014).

Increased rates of photosynthesis have been observed in forest communities with an increasing proportion of diffuse light, possibly due to a more even distribution of light in the canopy (Gu et al., 1999; Krakauer and Randerson, 2003; Misson et al., 2005; Urban et al., 2007), whereby light energy is more efficiently harvested from all directions deeper in the canopy. However, at the single leaf scale a 2–3% lower absorptance was found under diffuse light as compared to collimated light at equivalent incident irradiances (Brodersen and Vogelmann, 2007). In corals, it has been observed that gross photosynthesis increased ∼2-fold under collimated compared to diffuse light of identical downwelling irradiance (Wangpraseurt and Kühl, 2014) and the directional quality of light may thus elicit different photosynthetic responses and could potentially change the photosynthetic efficiency. A factor that could contribute to differences in photosynthetic activity under diffuse and collimated light is photoinhibition, which occurs when the electron transport chain is fully reduced and the photosystems are light saturated (Murata et al., 2007). Under high collimated light conditions, chloroplasts in leaves move to periclinal walls, and this might lead to decreased photoinhibition due to shading of other chloroplasts (Gorton et al., 1999). Under diffuse light, chloroplast movement to the periclinal walls is not complete (Williams et al., 2003) and thus distributed more randomly, which could lead to less effective self-shading and photoprotection (Brodersen et al., 2008).

The balance between photosynthesis and respiration and therefore, light use efficiency in benthic phototrophic systems is also influenced by the thickness of the diffusive and thermal boundary layers (Jørgensen and Des Marais, 1990; Jimenez et al., 2011; Brodersen et al., 2014). The diffusive boundary layer (DBL) is a thin water layer over submerged objects through which molecular diffusion is the dominant transport mechanism controlling the exchange of dissolved gases (e.g., O<sup>2</sup> and CO2) and solutes with the ambient water (Jørgensen and Des Marais, 1990; Shashar et al., 1996). The DBL can thus impose a major control on respiration and photosynthesis in aquatic environments. Dissipation of absorbed solar radiation as heat drives an increase in surface temperature that is counterbalanced by heat transfer to the surrounding water via a thermal boundary layer (TBL), where convection dominates the transport of heat and the surface warming increases linearly with the incident irradiance (Jimenez et al., 2008). Heat and mass transfer phenomena through boundary layers are therefore important processes when considering rates of photosynthesis and radiative energy budgets.

In the present study, we present the first radiative energy budget of a heterogeneous coral reef sediment and compare it with the energy budget of a compact photosynthetic biofilm on a coastal sediment. We investigate how diffuse and collimated light fields with identical levels of incident irradiance affect the radiative energy budget of the two microphytobenthic systems. Our analysis is based on a modified experimental approach first described by Al-Najjar et al. (2010).

## MATERIALS AND METHODS

#### Sample Sites and Collection

Coral reef sediment was sampled in April 2012 from a sheltered pseudo-lagoon ("Shark Bay") (Werner et al., 2006) on the reef flat surrounding Heron Island (151◦ 55′E, 23◦ 26′ S) that is located on the southern boundary of the Great Barrier Reef, Australia. Maximal incident solar photon irradiance at the sediment surface of the shallow reef flat during calm mid-day low tides is ∼1,500– 2,000 µmol photons m−<sup>2</sup> s −1 (Jimenez et al., 2012; Wangpraseurt et al., 2014b). The coral sediment (CS) was mainly composed of bright, semi-fine grained particles (mostly in the 200–500 µm size fraction) of deposited CaCO<sup>3</sup> from decomposed corals and other calcifying reef organisms. Diatoms, dinoflagellates, and cyanobacteria were found as dispersed aggregates in the sediment pore space along with amorphic organic material (detritus) throughout the upper few mm of the sediment (Figure S1).

The biofilm (BF) originated from a shallow sand bar at Aggersund, Limfjorden (Denmark) experiencing maximum summer photon irradiance of 1,000–1,500 µmol photons m−<sup>2</sup> s −1 . The biofilm was comprised of a ∼1 mm thick smooth layer of photosynthetically active filamentous cyanobacteria and microalgae (Microcholeus chtonoplastes, Oscillatoria spp., and pennate diatoms) embedded in exopolymers on top of finegrained (125–250 µm) dark sulfidic sandy sediment (Lassen et al., 1992b; Nielsen et al., 2015).

Coral reefs are usually considered oligotrophic but around Heron Island NH<sup>3</sup> and PO<sup>4</sup> concentrations of ∼0.3 and ∼0.1 mg L −1 (corresponding to ∼17 µmol L−<sup>1</sup> NH<sup>3</sup> and ∼1 µmol L−<sup>1</sup> PO4) have been found (Smith and Johnson, 1995), which is lower but in the same order as what is found in Danish waters (see e.g., Figure 16.1 in Henriksen et al. (2001)). In our experiments, we used a recirculating system containing 20 L of seawater (see below) that was changed daily. We therefore do not estimate that nutrient concentrations had a large impact on production between the two systems.

The porosity of the coral sediment and biofilm, φ, was 0.78 and 0.80, respectively, as determined from the weight loss of wet sediment (known initial volume and weight) after drying at 60◦C until a constant weight was reached:

$$\phi = \frac{\frac{M\_W}{D\_W}}{\frac{M\_W}{D\_W} + \frac{M\_S}{D\_S}}\tag{1}$$

where M<sup>W</sup> is the weight of water, D<sup>W</sup> is the density of water, M<sup>S</sup> is the weight of sediment/biofilm, and D<sup>S</sup> is the sediment/biofilm density.

#### Coral Sediment Samples

The CS samples were collected with Perspex corers (inner diameter 5.3 cm), and were maintained under a continuous flow of aerated seawater at ambient temperature and salinity (26◦C and S = 35) under a natural solar light regime for ∼24 h prior to further handling at the Heron Island Research Station (HIRS), Australia. Sediment cores were then mounted in a custom-made flow-chamber flushed with aerated seawater (26◦C and S = 35) for another 24 h prior to measurements. The flow-chamber (interior dimensions: 25 × 8 × 8 cm) had a honeycomb baffle between the water inlet and the sample, ensuring a stable laminar flow (see more details in Lichtenberg et al., 2016). During the acclimation time in the flow-chamber, the sediment cores were kept under a downwelling photon irradiance of ∼1,000 µmol photons m−<sup>2</sup> s <sup>−</sup><sup>1</sup> provided by a fiberoptic tungsten halogen lamp equipped with a collimating lens (KL2500-LCD, Schott GmbH, Germany). Before measurement at each experimental irradiance, the coral sediment core was illuminated for at least 45 min to ensure steady state O<sup>2</sup> and temperature conditions; as confirmed from repeated microprofile measurements. Throughout measurements, the flow-chamber was flushed with a stable laminar flow (∼0.5 cm s−<sup>1</sup> ) of filtered aerated seawater over the sediment surface as generated by a Fluval U1 pump submerged in a 20 L thermostated aquarium (26◦C and S = 35) and connected with tubing to the flowchamber.

#### Biofilm Samples

The BF samples were collected and contained in small rectangular plastic trays (7 × 2 × 5 cm) with the upper surface exposed and flush with the upper edge of the tray wall. After collection, the samples were kept humid and under a 12:12 h light-dark regime (∼100 µmol photons m−<sup>2</sup> s −1 ) in a thermostated room (16–18◦C**)**. The biofilm surface appeared dark green–brownish due to predominance of dense communities of cyanobacteria and diatoms (Lassen et al., 1992b). Prior to measurements, a sample tray was placed for 2 days in a flow-chamber flushed with 0.2 µm filtered aerated seawater (21◦C, S = 30) under a downwelling photon irradiance of ∼500 µmol photons m−<sup>2</sup> s −1 . During measurements, a stable laminar flow (∼0.5 cm s−<sup>1</sup> ) over the biofilm surface was maintained by a water pump (Fluval U1, Hagen GmbH, Germany) immersed in a 20 L aquarium with filtered aerated seawater (21◦C, S = 30) and connected with tubing to the flow-chamber.

#### Experimental Setup

Illumination was provided by a fiber-optic tungsten halogen lamp equipped with a collimating lens (KL-2500 LCD, Schott, Germany) positioned vertically above the flow-chamber. A spectrum of the used halogen lamp can be found in the Suppl. Info. in Lichtenberg et al. (2016) and is compared to typical solar spectrum measured on Heron Island reef flat in the Suppl. Info. in Wangpraseurt et al. (2014a), who found no major spectral effects on gross photosynthesis measurements. The intensity of the lamp could be controlled without spectral distortion by a built-in filter wheel with pinholes of various sizes. The downwelling photon irradiance of photosynthetically active radiation (PAR, 400–700 nm), Ed(PAR), (see definitions of abbreviations in Appendix) was measured with a calibrated irradiance meter (ULM-500, Walz GmbH, Germany) equipped with a cosine collector (LI-192S, LiCor, USA). Defined experimental irradiances (0, 50, 100, 200, 500, and 1,000 µmol photons m−<sup>2</sup> s −1 ) were achieved by adjusting the aperture on the fiber-optic lamp. The downwelling spectral irradiance at the above-mentioned levels was also measured in radiometric energy units (in W m−<sup>2</sup> nm−<sup>1</sup> ) with a calibrated spectroradiometer (Jaz, Ocean Optics, USA).

Collimated light was achieved by attaching a collimating lens to the fiber cable of the lamp. Diffuse light was achieved by inserting a TRIMMS diffuser (Transparent Refractive Index Matched Microparticles; Smith et al., 2003) between the collimator and the sample followed by lamp adjustment to achieve the same absolute levels of downwelling irradiance on the biofilm/sediment surface in collimated and diffuse light treatments.

#### Microscale Measurements of O<sup>2</sup> and Temperature

Oxygen concentrations were measured with a Clark-type O<sup>2</sup> microsensor (tip diameter ∼25µm; OX-25, Unisense A/S, Aarhus, Denmark) with a fast response time (<0.5 s) and a low stirring sensitivity (<1–2%; Revsbech, 1989). The microsensor was connected to a pA-meter (Unisense A/S, Aarhus, Denmark) and was linearly calibrated at experimental temperature and salinity from measurements in the aerated seawater in the freeflowing part of the flow-chamber and in anoxic layers of the sediment.

Temperature measurements were performed with a thermocouple microsensor (tip diameter ∼50 µm; T50, Unisense A/S, Aarhus, Denmark) connected to a thermocouple meter (Unisense A/S, Aarhus, Denmark). The temperature microsensors were linearly calibrated against readings of a high precision thermometer (Testo 110, Testo AG, Germany; accuracy ± 0.2◦C) in seawater at different temperatures. Analogue outputs from the temperature and O<sup>2</sup> microsensor meters were connected to an A/D converter (DCR-16, Pyroscience GmbH, Germany), which was connected to a PC. All microsensors were mounted in a PC-interfaced motorized micromanipulator (MU-1, PyroScience GmbH, Germany) controlled by dedicated data acquisition and positioning software (ProFix, Pyroscience, Germany). The micromanipulator was oriented in a 45◦ angle relative to the vertically incident light to avoid selfshading, especially in the light measurements. Depth profiles of temperature and O<sup>2</sup> concentration were measured in vertical steps of 100 µm. Before profiling, the microsensor tips were manually positioned on the sample surface to define the z = 0 position, determined from visual detection through a stereo microscope. The precisions of this approach is about ± the average grain size of the sediments, i.e., 125–500 µm.

The local volumetric rates of gross photosynthesis (PG(z); in units of nmol O<sup>2</sup> cm−<sup>3</sup> s −1 ) were measured with O<sup>2</sup> microsensors using the light-dark shift method (Revsbech and Jørgensen, 1983). Volumetric rates were measured in vertical steps of 100 µm throughout the sediment until no photosynthetic activity in the given depth was detected. The immediate O<sup>2</sup> depletion rate upon brief (2–4 s) darkening equalled the local rate of photosynthesis just prior to darkening; while no response in the O<sup>2</sup> signal upon darkening indicated a zero rate of photosynthesis. Areal rates of gross photosynthesis (in nmol O<sup>2</sup>

cm−<sup>2</sup> s −1 ) were calculated by depth integration over the euphotic zone with respect to the measuring interval used in the depth profile measurement of PG(z), similar to Al-Najjar et al. (2010, 2012):

$$PG = \Delta z \cdot \sum P\_G(z) \tag{2}$$

#### Temperature and O<sup>2</sup> Calculation

The net upward flux of O<sup>2</sup> from the photic zone of the sediments into the overlaying seawater was calculated (in nmol O<sup>2</sup> cm−<sup>2</sup> s −1 ) from measured steady-state O<sup>2</sup> concentration profiles using Fick's first law of diffusion:

$$J\_{NPP\uparrow} = -D\_0 \frac{\partial C}{\partial z} \tag{3}$$

where D<sup>0</sup> is the diffusion coefficient of O<sup>2</sup> in seawater at experimental temperature and salinity and <sup>∂</sup><sup>C</sup> ∂z is the linear O<sup>2</sup> concentration gradient in the DBL.

The downward O<sup>2</sup> flux from the photic zone of the sediments to the aphotic part of the sediment/biofilm was calculated in a similar manner as:

$$J\_{NPP\downarrow} = -\spadesuit D\_0 \frac{\partial C}{\partial z} \tag{4}$$

The total flux of O<sup>2</sup> out of the photic zone, i.e., the total net photosynthesis in the photic zone (NPP), was subsequently calculated as the difference between the upward and downward O<sup>2</sup> flux (Kühl et al., 1996).

To calculate the radiative energy conserved via photosynthesis (JPS; in J m−<sup>2</sup> s −1 ) we multiplied the areal gross photosynthesis, GPP, with the Gibbs free energy formed in the light-dependent reactions, where O<sup>2</sup> is formed by splitting water, which gains (including the formation of ATP) a Gibbs free energy of E<sup>G</sup> = 482.9 kJ (mol O2) −1 (Thauer et al., 1977).

$$Jps = JGPp\ EG\tag{5}$$

The amount of the absorbed light energy that was not photochemically conserved was dissipated as heat resulting in a local increase of the sediment/biofilm temperature relatively to the ambient seawater and thereby leading to the establishment of a TBL. The heat dissipation, i.e., the heat flux (in J m−<sup>2</sup> s −1 ) from the sediment/biofilm into the water column was calculated by Fourier's law of conduction:

$$J\_{H\uparrow} = k \frac{\partial T}{\partial z} \tag{6}$$

where k is the thermal conductivity in seawater (0.6 W m−<sup>1</sup> K −1 ) and dT/dz is the measured linear temperature gradient in the TBL (Jimenez et al., 2008). The heat flux from the photic zone into the aphotic sediment/biofilm, JH↓, was calculated as in Equation (6) but with the thermal conductivity constant of the sediment, k(b), which was estimated as:

$$k\left(b\right) = k\_s^{(1-\phi)} k\_f^{\phi} \tag{7}$$

where k<sup>s</sup> is the carbonate thermal conductivity (3.1 W m−<sup>1</sup> K −1 ; Clauser and Huenges, 1995), k<sup>f</sup> is the seawater thermal conductivity, and φ is the porosity of the sediment (Lovell, 1985).

The total heat flux, was used as an estimate of the total heat dissipation in the photic zone and was calculated as: J<sup>H</sup> = JH<sup>↑</sup> − JH↓.

#### Microscale Light Measurements

Spectral photon scalar irradiance was measured in units of counts nm−<sup>1</sup> with a fiber-optic scalar irradiance microprobe [integrating sphere diameter ∼100 µm; (Lassen et al., 1992a)] connected to a fiber-optic spectrometer (USB2000, Ocean Optics, Dunedin, FL, USA). A black non-reflective light-well was used to record spectra of the downwelling photon scalar irradiance, Ed(λ), (in units of counts nm−<sup>1</sup> ) with the tip of the scalar irradiance microsensor positioned in the light path at the same distance from the light source as the sediment surface. Using identical light settings, the absolute downwelling irradiance, EABS(λ) (in W m−<sup>2</sup> ) was also quantified with a calibrated spectroradiometer (Jaz-ULM, Ocean Optics, Dunedin, Florida, USA).

#### Irradiance Calculations

The spectral scalar irradiance, E0(λ), was measured in vertical steps of 0.1–0.2 mm in the sediment and was calculated as the fraction of the incident downwelling irradiance, i.e., E0(λ)/Ed(λ), and plotted as transmittance spectra in % of Ed(λ). The relative measurements of scalar irradiance in different depths in the biofilm/sediment were converted to absolute scalar irradiance spectra in units of W m−<sup>2</sup> nm−<sup>1</sup> as EABS(λ) <sup>×</sup>E0(λ)/Ed(λ). Absolute scalar irradiance spectra were converted to photon scalar irradiance spectra (in units of µmol photons m−<sup>2</sup> s −1 nm−<sup>1</sup> ) by using Planck's equation:

$$E\_{\lambda} = h \frac{c}{\lambda} \tag{8}$$

where E<sup>λ</sup> is the energy of a photon with wavelength, λ, h is Planck's constant (6.626 × 10−<sup>34</sup> W s<sup>2</sup> ), and c is the speed of light in vacuum (in m s−<sup>1</sup> ).

Spectral attenuation coefficients of scalar irradiance, K0(λ), were calculated as (Kühl, 2005):

$$\mathcal{K}\_0\left(\lambda\right) = -\ln \frac{\left(E\_0(\lambda)\_1/E\_0(\lambda)\_2\right)}{z\_2 - z\_1} \tag{9}$$

where E0(λ)<sup>1</sup> and E0(λ)<sup>2</sup> are the spectral scalar irradiances measured at depth z<sup>1</sup> and z2, respectively.

Light attenuation was also calculated by integrating the spectral quantum irradiance over PAR (420–700 nm) yielding the PAR scalar irradiance (E0(PAR), in µmol photons m−<sup>2</sup> s −1 ), i.e., the light energy available for oxygenic photosynthesis at each measurement depth. The diffuse attenuation coefficient of E0(PAR), K0(PAR), was obtained by fitting the measured E0(PAR) vs. depth profiles with an exponential model:

$$E\_0\left(z\right) = E\_0\left(0\right)e^{-K\_0(PAR)\left(z - z(0)\right)}\tag{10}$$

#### Reflectance Measurements

The PAR irradiance reflectance (R) of the sediment/biofilm surface was calculated as

$$R\left(PAR\right) = \int\_{420}^{700} \frac{E\_{\mu}\left(\lambda\right)}{E\_{d}\left(\lambda\right)} d\lambda \tag{11}$$

where Eu(λ) is the upwelling irradiance at the sediment surface, here estimated as the diffuse backscattered spectral radiance measured at the sediment surface (Kühl, 2005) and Ed(λ) is the downwelling irradiance estimated as the backscattered spectral radiance measured over a white reflectance standard (Spectralon; Labsphere, North Sutton, NH, USA); both measured with a fiber-optic field radiance microprobe (Jørgensen and Des Marais, 1988). The R(PAR) measurements assumed that the light backscattered from the sediment/biofilm surface was completely diffused (Kühl and Jørgensen, 1994).

#### Absorbed Light Energy

The absorbed light energy (JABS; in W m−<sup>2</sup> = J m−<sup>2</sup> s −1 ) in the sediment/biofilm was estimated by subtracting the downwelling and upwelling irradiance at the surface:

$$J\_{ABS} = \int\_{420}^{700} E\_d(\lambda)(1 - R\_\lambda(\lambda))d\lambda \tag{12}$$

where Ed(λ) and R(λ) are the downwelling spectral irradiance and irradiance reflectance, respectively. This parameter is equivalent to the so-called vector irradiance, which is a measure of the net downwelling radiative energy flux.

#### Energy Budget and Photosynthetic Efficiency Calculations

A balanced radiative energy budget of the sediment/biofilm was calculated according to (Al-Najjar et al., 2010) with slight modifications (**Figure 1**) as:

$$J\_{ABS} = J\_H + J\_{PS} \tag{13}$$

assuming that autofluorescence from the sediment/biofilm was negligible. Consequently, εPS + ε<sup>H</sup> = 1, where, εPS and ε<sup>H</sup> represent the efficiency of photosynthetic energy conservation and heat dissipation, respectively, for a given absorbed light energy JABS in the entire euphotic zone (Al-Najjar et al., 2010):

$$\left(\varepsilon\_{\rm PS} = \frac{J\_{\rm PS}(J\_{\rm ABS})}{J\_{\rm ABS}}\right) \text{ and } \varepsilon\_{\rm H} = \frac{J\_{\rm H}(J\_{\rm ABS})}{J\_{\rm ABS}} \tag{14}$$

Areal gross photosynthesis rates as a function of JABS, were fitted with the saturated exponential model (Webb et al., 1974) to estimate the maximum conserved energy flux by photosynthesis (JPS,max) (in J m−<sup>2</sup> s −1 ):

$$J\_{\rm PS} \left( J\_{\rm ABS} \right) = J\_{\rm PS, max} \left( 1 - e^{-J\_{\rm ABS}/E\_k} \right) \tag{15}$$

This yielded an estimate of the maximum photochemically conserved energy flux JPS,max. The respective efficiencies

under light-limiting conditions, i.e., for JABS→0, were then calculated as:

$$
\varepsilon\_{\text{PS,max}} = \frac{I\_{\text{PS,max}}}{E\_k} and \text{ } \varepsilon\_{H,min} = 1 - \varepsilon\_{\text{PS,max}}\tag{16}
$$

where E<sup>k</sup> is the photochemical light acclimation index, i.e., the irradiance at the onset of photosynthetic saturation, calculated as E<sup>k</sup> = JPS,max/α, where α is the initial slope of the fitted photosynthesis vs. JABS curve.

## RESULTS

#### Light Environment

At all incident irradiances, the photon scalar irradiance, E0(PAR), decreased with increasing sediment depth (**Figure 2**). Light attenuation was strongly enhanced around wavelengths 625 and 670 nm, corresponding to absorption maxima of phycocyanin and Chl a, respectively (**Figure 3**). Surface reflection from the biofilm surface was on average 1.8 and 1.7% of the incident PAR under diffuse and collimated light, respectively, while it was >15 times higher in the coral sediment, i.e., 30.2 and 28.1% for diffuse and collimated light, respectively. Reflection did not change with increasing irradiance (Figure S2). The profiles of scalar irradiance showed non-uniform attenuation with depth and could be influenced by local enhancement of photon pathlength (Kühl and Jørgensen, 1994; Kühl et al., 1997) in the uppermost layers (**Figure 2**). At the highest incident photon irradiances (500 and 1,000 µmol photons m−<sup>2</sup> s −1 ), the exponential attenuation of collimated light within the biofilm was observed below 0.2 mm, whereas diffuse light was attenuated exponentially from the biofilm surface under all investigated irradiance levels (**Figure 2**). In the coral sediment, the exponential attenuation occurred deeper (below 0.5–0.7 mm) due to enhanced scattering, redistribution and trapping of photons in the upper sediment layers (**Figure 2**). In the biofilm, PAR attenuation was stronger in the top layer than in the bottom layer both for diffuse and collimated light (**Figure 2**). Additionally, attenuation of collimated light in the top layer was stronger than for diffuse light at all irradiances except 1,000 µmol photons m−<sup>2</sup> s −1 , whereas light attenuation in the lower sediment dominated layers was similar for diffuse and collimated incident light. In the coral sediment no distinct

differences in light attenuation was observed between topand bottom layers other than a deeper onset of exponential attenuation (0.5–0.7 mm). The top layer of the biofilm showed ∼10 times stronger light attenuation than the coral sediment with average PAR attenuation coefficient of α = 9.52 and α = 10.54 mm−<sup>1</sup> for diffuse and collimated light, respectively, compared to α = 1.18 mm−<sup>1</sup> in the coral sediment (both light types).

In both sediments, attenuation of light corresponded to absorption maxima of Chl a (440 and 670 nm) and phycocyanin (620 nm; **Figure 3**). A third attenuation maximum was observed around 575 nm indicative of phycoerythrin, commonly found in cyanobacteria (Colyer et al., 2005). In the biofilm, attenuation of visible light was strongest in the top 0.3 mm of the biofilm, except under the highest collimated irradiance (1000 µmol photons m−<sup>2</sup> s −1 ), where the strongest attenuation occurred over the 0.3–0.6 mm zone (**Figure 3**). Below 0.6 mm, the enhanced attenuation around wavelengths 575, 625, and 670 nm decreased and the attenuation of light in the PAR region became more uniform in the underlying layers (**Figure 3**). Again, attenuation of collimated light was slightly higher than diffuse light.

In the coral sediment, the highest light attenuation was 1–2 mm below the sediment surface (∼1.6 mm−<sup>1</sup> at 670 nm at all incident irradiances) while the lowest attenuation was found in the upper 0–1 mm, consistent with the scalar irradiance profiles (**Figures 2**, **3**).

## Temperature and O<sup>2</sup> Microenvironment

In the biofilm, a ∼0.8 mm thick diffusive boundary layer (DBL) developed between the biofilm and the surrounding water (**Figures 1**, **4**). In dark, O<sup>2</sup> was depleted within the upper 1.5 mm and the areal dark respiration rate was calculated to 0.039 nmol O<sup>2</sup> cm−<sup>2</sup> s −1 . The fluxes of O<sup>2</sup> increased with irradiance until saturation was reached at a downwelling photon irradiance of ∼100 µmol photons m−<sup>2</sup> s −1 , where the top of the biofilm experienced O<sup>2</sup> concentrations >450% of air saturation (**Figure 4**). The O<sup>2</sup> concentration profiles for diffuse and collimated light were similar, although O<sup>2</sup> penetrated deeper under diffuse light, especially at the highest photon irradiances (500 and 1,000 µmol photons m−<sup>2</sup> s −1 ; **Figure 4**). The coral sediment had a ∼1–1.4 mm thick DBL; dark respiration was similar to the biofilm (0.037 nmol O<sup>2</sup> cm−<sup>2</sup> s −1 ), while saturation of photosynthesis was reached at a higher downwelling photon irradiance of ∼200 µmol photons m−<sup>2</sup> s −1 (**Figure 3**). The similar dark respiratory O<sup>2</sup> uptake in sediment and biofilm indicated that the combined respiration of autotrophic and heterotrophic organisms was of similar magnitude in the two systems. The more variable DBL thickness in the coral sediment varied independently of irradiance and was most likely a result of the heterogeneous surface topography (**Figure 4**). A detailed mapping of the DBL landscape was beyond the scope of this study but, we estimate that the mass transfer between the sediment and overlying water was not influenced by turbulences which would have been evident as non-linear concentration gradients between sediment surface and bulk water (Lichtenberg et al., 2017). At incident irradiances >200 µmol photons m−<sup>2</sup> s −1 the O<sup>2</sup> productive zone was stratified under both diffuse and collimated light, with an O<sup>2</sup> concentration maximum of ∼600% air saturation ∼1.7 mm below the sediment surface (**Figure 4**). Photosynthesis was apparently distributed in two major layers, a ∼0.5 mm thick layer at the sediment surface, and a ∼1 mm thick layer peaking 2 mm below the sediment surface (**Figure 4** and Figure S3). The O<sup>2</sup> concentration profiles for diffuse and collimated light were similar at low to moderate irradiance, then showed a deeper O<sup>2</sup> penetration depth under diffuse light at incident irradiance >500 µmol photons m−<sup>2</sup> s −1 in comparison to O<sup>2</sup> profiles measured under collimated light (**Figure 4**). The O<sup>2</sup> profiles in the coral sediment showed high standard deviations, possibly due to a more patchy distribution of the photosynthetic organisms within the sediment and overall variability in the sediment grain size and surface topography.

In both biofilm and coral sediment, the surface temperature increased relative to the overlaying seawater with increasing irradiance. The local heating was dissipated by heat transfer over a ∼3 mm thick TBL into the overlaying seawater and into deeper sediment layers (**Figures 5**, **6**). Robust measurements of biofilm/sediment heating could only be obtained at incident photon irradiances of ≥200 µmol photons m−<sup>2</sup> s −1 (≥500 µmol photons m−<sup>2</sup> s −1 for the coral sediment under collimated light). At the highest irradiance (1,000 µmol photons m−<sup>2</sup> s −1 ), the biofilm surface was 0.51 ± 0.036 and 0.41 ± 0.008◦C warmer than the overlaying water, while the coral sediment surface was 0.53 ± 0.031 and 0.48 ± 0.040◦C warmer than the surrounding water for diffuse and collimated light, respectively. Similar temperature profiles were observed between collimated and diffuse light, although a slightly enhanced surface heating and thus a higher efflux of heat was observed under diffuse light (**Figure 5**). Comparing the slope of the surface warming vs. vector irradiance under diffuse and collimated light, respectively, diffuse light had a greater impact on surface warming by 30 and 27% in the biofilm and in the coral sediment, respectively (**Figure 6**).

#### Photosynthesis

Maximal volume-specific gross photosynthesis rates of the biofilm ranged between 7.0 and 8.7 nmol O<sup>2</sup> cm−<sup>3</sup> s −1 (collimated and diffuse light, respectively) under low irradiance (50–200 µmol photons m−<sup>2</sup> s −1 ), while rates decreased at photon

FIGURE 5 | Vertical depth profiles of temperature change, <sup>1</sup>T (in ◦C) measured in biofilm (upper panels) and coral sediment (lower panel) at downwelling photon irradiances of 0, 200, 500, and 1,000 µmol photons m−<sup>2</sup> <sup>s</sup> <sup>−</sup><sup>1</sup> under collimated (A,C) and diffuse light (B,D). Symbols represent means, while dashed lines indicate ± 1 S.D. (n = 3). The dotted line in y = 0 indicates the sediment surface, while the dotted line in x = 0 indicates a 0◦C temperature change.

irradiances of >200 µmol photons m−<sup>2</sup> s −1 (Figure S3A). The thickness of the photic zone generally increased with increasing photon irradiance and varied from 0.4 to 1.2 mm in the biofilm under diffuse light and from 0.2 to 0.9 mm under collimated light.

In the coral sediment, the highest volume-specific rates of photosynthesis were measured within the upper 1 mm, with maximal gross photosynthesis rates of 11.97 nmol O<sup>2</sup> cm−<sup>3</sup> s −1 at the sediment surface under collimated light and 3.05 nmol O<sup>2</sup> cm−<sup>3</sup> s −1 at a depth of 0.6 mm under diffuse light (Figure S3B). The photic zone in the coral sediment increased with increasing irradiance and ranged in thickness from 1.5 to 3 mm under diffuse light and from 2 to 3.5 mm under collimated light. The apparent stratification in O<sup>2</sup> concentration found in the coral sediment was confirmed in the profiles of gross photosynthesis with peaks in gross photosynthesis in the upper 1 mm and 1.5–2.5 mm from the surface at photon irradiances >50 µmol photons m−<sup>2</sup> s −1 (Figure S3B).

Under low photon irradiance <200 µmol photons m−<sup>2</sup> s −1 in the biofilm, the area specific gross photosynthesis rate (PG) was higher under diffusive illumination, while PG under diffuse and collimated illumination were similar at higher irradiances (**Figure 6**). In contrast, PG in the coral sediment was generally in the range of 3–4 times lower under diffuse- compared to collimated light (Figure S3B; **Figure 6B**). We note that the gross photosynthesis measurements in the coral sediment under diffuse light were performed at the University of Technology Sydney (UTS) rather than on HIRS, where the rest of the measurements took place. We speculate that the transport from Heron Island created prolonged anoxic conditions throughout the sediment and this might have caused a change in community composition and structure of the sediment. These measurements were therefore excluded when calculating the light energy budget for diffuse light in the coral sediment.

#### Energy Budgets

The photosynthesis-irradiance (PE) curve of the coral sediment measured in diffuse light increased with increasing light intensity with an initial slope of 0.05 ± 0.01, until reaching an asymptotic saturation level at JPS,max = 1.72 ± 0.20 J m−<sup>2</sup> s −1 at a downwelling photon irradiance of ∼300 µmol photons m−<sup>2</sup> s −1 (**Figure 6B**). In contrast, the PE-curve of the coral sediment in collimated light increased with the with a slope of 0.26 ± 0.04, reaching a maximum saturation value of JPS,max = 4.24 ± 0.23 J m−<sup>2</sup> s −1 , at downwelling photon irradiance ∼110 µmol photons m−<sup>2</sup> s −1 (**Figure 6B**). In the biofilm, the onset of photosynthesis saturation occurred already at a downwelling photon irradiance of ∼50 µmol photons m−<sup>2</sup> s −1 , where JPS,max reached an asymptotic saturation level of 0.87 J m−<sup>2</sup> s −1 for both diffuse and collimated light (**Figure 6A**).

Sediment surface warming increased linearly with irradiance under both diffuse and collimated light with average slopes of CSαdiff = 4.33·10−<sup>3</sup> ◦C (J m−<sup>2</sup> s −1 ) −1 and CSαcoll = 2.14·10−<sup>3</sup> ◦C (J m−<sup>2</sup> s −1 ) −1 in the coral sediment, as compared to BFαdiff = 2.77·10−<sup>3</sup> ◦C (J m−<sup>2</sup> s −1 ) −1 and BFαcoll = 2.0·10−<sup>3</sup> ◦C (J m−<sup>2</sup> s −1 ) −1 in the biofilm (**Figures 6C,D**). Surface warming was stronger under diffuse light as compared to collimated light in both sediments (**Figures 5**, **6C,D**).

FIGURE 6 | Energy conversion by photosynthesis, heat dissipation and the sum of photosynthesis and heat dissipation vs. downwelling irradiance in biofilm (left panels) and corals sediment (right panels). Red symbols and lines show data for diffuse illumination, while black symbols and lines show data for collimated illumination. (A,B) Areal gross photosynthesis rates (in J m−<sup>2</sup> s −1 ) measured at downwelling photon irradiances of 0, 50, 100, 200, 500, and 1,000 µmol photons m−<sup>2</sup> s −1 , and then fitted with a saturated exponential model (Webb et al., 1974; CS: R 2 diff = 0.92, R 2 coll = 0.97; BF: R <sup>2</sup> = 0.88 for both diffuse and collimated; n = 3). (C,D) Temperature gradients (in ◦C) between the ambient seawater and the sediment surface (flow = 0.5 cm s−<sup>1</sup> ), measured at vector irradiances of 30, 75, and 149 J m−<sup>2</sup> s <sup>−</sup><sup>1</sup> or 40, 100, and 200 J m−<sup>2</sup> s −1 for the coral sediment and biofilm, respectively. Data points show means ± SD (n = 3); CS: R 2 diff = 0.99, R 2 coll = 0.96; BF: R <sup>2</sup> = 0.99 for both diffuse and collimated light. (E,F) The summed energy dissipation of the system (in J m−<sup>2</sup> s −1 ), i.e., the sum of energy conserved by photosynthesis and energy dissipated as heat, measured at vector irradiances of 30, 75, and 149 J m−<sup>2</sup> s <sup>−</sup><sup>1</sup> and 40, 100, and 200 J m−<sup>2</sup> s −1 for the coral sediment and biofilm, respectively. The dashed line represents a 1:1 relationship between the incoming and outgoing energy of the system (i.e., the theoretically expected relationship). CS: R 2 diff = 0.99, R 2 coll = 0.96; BF: R <sup>2</sup> = 0.99 for both diffuse and collimated light; (n = 3).

The summed flux of energy conserved by photosynthesis and dissipated as heat (JPS + JH) serves as a control to determine the potential deviations between absorbed and dissipated energy (**Figures 6E,F**). Dissipation of energy from the system increased linearly with increasing vector irradiance with slopes in the coral sediment of 0.89 ± 0.003 and 0.89 ± 0.120, for diffuse and collimated light respectively, and slopes in the biofilm of 0.93 and 1.03, for diffuse and collimated light respectively. When all outgoing/used energy equals the incoming light energy the slope of the used- vs. incoming energy curve would be =1, and thus the method used here apparently accounted for the majority of the incident light energy.

About 29% of the incident light energy was back-scattered from the coral sediment surface and thus not absorbed, whereas the surface reflection was only ∼2% of the incident irradiance in the biofilm (**Figure 7**; Figure S2). The fraction of energy conserved by photosynthesis decreased with increasing irradiance in both biofilm and sediment (**Figures 7**, **8**). Over the investigated incident irradiances (200–1,000 µmol photons m−<sup>2</sup> s −1 ), photosynthetic energy conservation in the coral sediment illuminated with diffuse light decreased from 6.7 to 2.0% of the incident light energy, favoring heat dissipation (which increased from 63.1 to 67.8%), and from 9.3 to 2.1% of the incident light energy under collimated light (where

heat dissipation increased from 62.6 to 69.8%; **Figure 7**; Table S2).

Under an incident photon irradiance of 200µmol photons m−<sup>2</sup> s −1 , the proportion of incident light energy that was conserved via photosynthesis was much lower in the biofilm where 1.9 and 2.3% (diffuse and collimated light, respectively) of the incident light energy was conserved, whereas 96.3 and 96.0% of the incident light energy was dissipated as heat, respectively (**Figure 7**; Table S1). At an incident irradiance of 1,000 µmol photons m−<sup>2</sup> s −1 , only 0.6 and 0.5% of the incident energy was conserved by photosynthesis, while 97.6 and 97.8% was dissipated as heat under diffuse and collimated light, respectively (**Figure 7**; Table S2).

The maximum photochemical energy conservation in the coral sediment was observed at an incident irradiance of ∼100 µmol photons m−<sup>2</sup> s −1 (18.1% of the absorbed light energy), whereas the biofilm had maximum energy conservation through photosynthesis (14.7% of the absorbed light energy) at the lowest measured incident irradiance (50 µmol photons m−<sup>2</sup> s −1 ; **Figure 8**). In addition, the biofilm had higher photosynthetic efficiencies under diffuse light compared to collimated light at low light intensities (<200 µmol photons m−<sup>2</sup> s −1 ; **Figure 8**).

The photosynthetic efficiencies of biofilm and coral sediment under light-limiting conditions (JABS→0), εPS,max, were calculated from the initial slope of the areal PG vs. vector irradiance curve to 26.2% of the absorbed light energy (CS, collimated light) compared to 16 and 9.0% of the absorbed light energy (BF, diffuse and collimated light, respectively).

#### DISCUSSION

We present a closed radiative energy budget of a heterogeneous coral reef sediment and compare it to the radiative energy budget of a flat dense biofilm (**Figure 6**; Figure S4). The closed light energy budgets were measured under both diffuse and collimated illumination to test potential effects of the directionality of light

on the photosynthetic efficiencies of the phototrophs. We found that a higher fraction of the absorbed light energy was conserved by photosynthesis in the heterogenous loosely organized coral sediment, while the radiative energy budgets of both sediment types were highly dominated by dissipation of heat.

#### Light

The thin highly pigmented cyanobacterial biofilm was growing on the surface of a fine-grained (125–250 µm) dark sandy sediment, whereas the photosynthetic microorganisms exhibited a more patchy distribution within the large-grained (100–500 µm) bright and highly scattering coral sediment. This structural difference between the two systems led to a ∼15 times higher surface reflection and a decreased energy absorption in the coral sediment compared to the biofilm that displayed >8 times higher light attenuation coefficients. As previously shown (Lassen et al., 1992b; Kühl and Fenchel, 2000) the scalar irradiance at, or immediately below, the surface increased, and the spectral composition was altered relative to the incident irradiance (**Figures 2**, **3**). Such increase in scalar irradiance in the near surface layer is due to intense multiple scattering by particles (biotic and abiotic) causing a local photon path-length increase and thus a prolonged residence time of scattered photons in the surface layers that also receive a continuous supply of incident photons from the light source (Kühl and Jørgensen, 1994). This effect can be further enhanced in the presence of exopolymers with a slightly higher refractive index than the surrounding seawater leading to photon trapping effects (Kühl and Jørgensen, 1994; Decho et al., 2003). Furthermore, the structural difference between the loosely organized CaCO<sup>3</sup> particles compared to the flat biofilm could possibly result in differences in the reflection characteristics from the uppermost layers. In the biofilm, the flat homogeneous surface reflects light relatively uniformly, with some ratio between specular vs. diffuse reflection. However, in the heterogeneous coral sediment a higher degree of forward scattering will most likely be present as the angle of reflection will be more complex due to the roughness of the surface, resulting in a deeper penetration of light in the coral sediment.

#### Temperature

We directly measured both the upward and downward heat dissipation of radiative energy (**Figure 5**). Previous studies of energy budgets ignored the downward heat flux (Al-Najjar et al., 2010, 2012), and although Jimenez et al. (2008) estimated the downward heat dissipation from a theoretical model considering the physical properties of heat transfer in coral skeleton, this study presents energy budgets of phototrophic systems for which the complete heat balance was directly measured. Over a range of incident irradiances, the downward heat flux was the same order of magnitude as the upward heat flux in both biofilm and coral sediment and thus is an important parameter when compiling light energy budgets for the photic zone in benthic systems (**Figure 5**).

The majority of the absorbed light energy was dissipated as heat (**Figure 7**; Table S1) with a linear relationship between increasing incident irradiance and heat dissipation under both diffuse and collimated light, albeit with a ∼30% enhanced surface warming under diffuse light as compared to collimated light (**Figures 5**, **6**). Apparently, diffuse light was absorbed more efficiently in the uppermost layers, increasing the local photon density and residence time in these layers resulting in increased energy deposition and surface temperatures. This was supported by a higher heat flux into the water column under diffuse light, and a higher heat flux into the sediment under collimated light (data not shown). At increasing irradiances the surface temperature of the sediments exceeded the surrounding water temperature and convective heat transport occurred over the TBL (**Figure 5**; Jimenez et al., 2011). While we cannot dissect the observed heat dissipation into particular mechanisms and their relative magnitude, part of such dissipation in optically dense biofilms and sediments involves NPQ processes that protect the photosynthetic apparatus under high irradiance by channeling excess light energy into heat dissipation (Falkowski and Raven, 2007; Al-Najjar et al., 2012). The heat fluxes from the photic zone were generally higher in the biofilm when compared to the coral sediment, due to the lower reflection and thus higher absorption in the biofilm (Figure S4). However, when normalizing the heat fluxes to the absorbed light energy (which was 33% higher in the biofilm than in the coral sediment) the heat dissipation was of similar magnitude, and variations in heat flux values between the sediment and biofilm became <15%. The degree of heat dissipation therefore seems tightly correlated to the quantity of absorbed energy.

#### Photosynthesis

The overall photosynthetic efficiency of the biofilm and coral sediment decreased with increasing incident irradiance, with photosynthesis exhibiting saturation at higher irradiance under both diffuse and collimated light (**Figure 6**). The highest energy storage efficiency of the coral sediment was observed under lightlimiting conditions (<200 µmol photons m−<sup>2</sup> s −1 ; **Figures 7**, **8**), and the coral sediment generally exhibited high light use efficiencies that were comparable to those observed in corals at equivalent incident photon irradiances (Brodersen et al., 2014). The photosynthetic activity in the coral sediment was stratified at incident irradiances >50 µmol photons m−<sup>2</sup> s <sup>−</sup><sup>1</sup> under both diffuse and collimated light (**Figure 4**). This stratification could be a result of different factors influencing the photosynthetic activity such as steep light attenuation over depth, locally high volume-specific rates of metabolic activity, higher local biomass of phototrophs, and diffusion limitation of metabolic products and substrates (Kühl et al., 1996; Kühl and Fenchel, 2000; Al-Najjar et al., 2012). Such vertical stratification has also been associated with phototactic responses to light (Whale and Walsby, 1984; Lassen et al., 1992b), where motile photosynthetic organisms migrate to an optimal depth for photosynthesis at a given irradiance, where the available light is neither limiting nor inhibiting the rate of photosynthesis (Al-Najjar et al., 2012). These migration patterns are well documented both as photoand aero-tactic responses and to escape from e.g., toxic levels of sulfide (Kühl et al., 1994; Bebout and Garcia-Pichel, 1995). The two photosynthetic active layers were situated at the sediment surface and ∼2 mm below (∼0.5 and 1 mm thick layers, respectively; **Figure 3**).

The area-specific rates of gross photosynthesis of the coral sediment were ∼4 times higher than in the biofilm, due to a ∼3 times deeper euphotic zone and slightly higher volumespecific photosynthesis rates in the coral sediment than in the biofilm (**Figures 6**, **7**; Figure S3). Consequently, the coral sediment reached an asymptotic maximum in PG in terms of energy dissipation via photosynthesis of ∼4.2 J m−<sup>2</sup> s −1 as compared to only ∼0.9 J m−<sup>2</sup> s −1 in the biofilm (**Figure 6**). The E<sup>k</sup> value, i.e., the irradiance at the onset of photosynthesis saturation, was >2 higher in the coral sediment compared to the Danish biofilm, which reflects the different in situ light conditions experienced by the two systems in their respective geographic locations (Denmark: 55◦N, Heron Island: 23◦ S). Thus, the dense biofilm appeared acclimated to low irradiances as previously shown (Kühl et al., 1996; Kühl and Fenchel, 2000; Al-Najjar et al., 2012) where highly reduced quantum efficiencies are seen at increasing irradiance due to the employment of e.g., NPQ processes. Accordingly, the coral sediment maintained higher photosynthetic efficiencies, even at high irradiance. This could in part be explained by the high scattering in the sediment particles that creates a more even spread of the light field over the sediment matrix and a more dispersed photic zone; a factor that have been speculated to be responsible for the high photosynthesis in coral tissues (Brodersen et al., 2014; Wangpraseurt et al., 2014a). A more homogenous distribution of light would create a larger region where light is neither limiting nor inhibiting photosynthesis. Thus, the loosely organized more heterogenous coral sediment apparently exhibit a more open, canopy-like organization compared to the dense biofilm.

Community photosynthesis is generally higher than that of individual phytoelements (Binzer and Sand-Jensen, 2002; Binzer and Middelboe, 2005; Binzer et al., 2006) and in addition, higher community photosynthesis has been found under diffuse illumination in open canopy systems which was explained by a more even light field inside the canopy (Gu et al., 2002; Brodersen et al., 2008). In spite of this difference in overall acclimation to light, a decrease in the surface layer photosynthesis was seen in the coral sediment at an incident irradiance of 500 µmol photons m−<sup>2</sup> s −1 , which could either reflect the heterogeneity and patchiness of the phototrophs found in the sediment, or could point to a possible migration of motile phototrophic organisms. Migration as a phototactic response is recognized as an effective mechanism for controlling photon absorption across different systems such as terrestrial plants (Wada et al., 2003) and microphytobenthic assemblages (Serodio et al., 2006; Cartaxana et al., 2016a,b), and similar phototactic response has been shown in coral tissues where the in hospite light environment can be modulated by host tissue movement (Wangpraseurt et al., 2014a, 2017). Downward migration at high irradiances is probably correlated with increasing photic stress e.g., by the formation of reactive oxygen species that can damage photosystem II by preventing the synthesis of the D1 protein in these layers (Hihara et al., 2001; Nishiyama et al., 2001; Aarti et al., 2007; Latifi et al., 2009; Al-Najjar et al., 2010). Several ways to counter such photic stress exists. One of the most effective short-term responses to photic stress is to employ NPQ in which photons are emitted as heat when cells experience over-saturating photon fluxes. Another strategy to avoid photo-damage is to upregulate the expression of sun-protective pigments such as β-carotenes (Zhu et al., 2010), which were found in significant amounts by HPLC analysis of the coral sediment (Figure S1).

Photosynthetic energy conservation was higher under collimated light as compared to illumination with diffuse light at moderate irradiance (200 µmol photons m−<sup>2</sup> s −1 ; **Figure 6**). This finding correlates with previous studies of individual terrestrial leaves reporting 10–15% higher energy conservation via photosynthesis under collimated- relative to diffuse light (Brodersen et al., 2008) and in corals it has been shown that gross photosynthesis was 2-fold higher under direct vs. diffuse light (Wangpraseurt and Kühl, 2014). In terrestrial leaves, the more efficient utilization of collimated light has been ascribed to specialized tissue structures such as columnar palisade cells (Vogelmann and Martin, 1993), that increase the absorptance of direct light over diffuse light (Brodersen and Vogelmann, 2007). Furthermore, light-induced chloroplast movement has been shown to be less effective under diffuse illumination (Gorton et al., 1999; Williams et al., 2003). In corals the higher photosynthesis at the tissue surface was explained by optical properties enhancing the scalar irradiance near the surface under direct illumination (Wangpraseurt and Kühl, 2014). This tendency changed dramatically in the dense photosynthetic biofilm at light-limiting conditions (≤100 µmol photons m−<sup>2</sup> s −1 ) favoring effective light utilization under diffuse light (**Figure 7**). Thus, the optical properties and the structural organization of phytoelements seem tightly linked to the photosynthetic quantum efficiencies across different systems and light angularity may therefore elicit differential photosynthetic responses depending on the system and on the scale at which it is studied.

#### CONCLUSION

Our results show that a higher fraction of the absorbed light energy was conserved by photosynthesis in the heterogeneous coral sediment due to a deeper photic zone and slower saturation of photosynthesis with increasing irradiance in the more open structure of the sediment microcanopy as compared to the flat and highly absorbing biofilm. The balanced radiative energy budget of biofilm and coral sediment was highly dominated by heat dissipation and the efficiency of photosynthetic energy conservation decreased with increasing irradiance. Although the two systems exhibited similar heat dissipation, the photic zones wherein such dissipation took place was very different e.g., by a three times deeper photic zone in the coral sediment than in the biofilm. In addition, several variances were found between illumination with diffuse or collimated light: (i) diffuse light enhanced dissipation of heat (∼30%) in the upper sediment layers as compared to collimated light; (ii) at low incident irradiance (200 µmol photons m−<sup>2</sup> s −1 ) photosynthetic energy conservation was higher (3–4% of the absorbed light energy) in collimated light as

#### REFERENCES


compared to diffuse light; a tendency that dramatically changed in the photosynthetic biofilm at low and light-limiting incident irradiances (≤100 µmol photons m−<sup>2</sup> s −1 ) favoring effective light utilization under diffuse light (up to a ∼2-fold increase). However, cyanobacterial and diatom dominated mats have been shown to migrate vertically employing photo- and/or chemotactic responses (Richardson and Castenholz, 1987; Bhaya, 2004; Serodio et al., 2006; Coelho et al., 2011; Cartaxana et al., 2016a) and the motility of the phototrophs was not considered here. Thus, there is a need to further explore how vertical migration affects the radiative energy balance and thereby the light use efficiency in microbenthic systems such as sediments and biofilms.

#### AUTHOR CONTRIBUTIONS

ML, KB, and MK designed the experiments; ML and KB performed experiments; ML, KB, and MK analyzed data; ML wrote the paper with editorial inputs from KB and MK.

#### ACKNOWLEDGMENTS

We thank P. J. Ralph, P. Brooks, M. Zbinden and other colleagues at University of Technology Sydney (C3, UTS) for access to laboratory facilities, technical support and help with HPLC analysis of the coral sediment. We thank the staff at Heron Island Research Station for technical assistance during the field work. V. Schrameyer, D. Wangpraseurt and D. A. Nielsen are thanked for thoughtful discussions. The research was conducted under research permits for field work on the Great Barrier Reef, Australia (G11/34670.1 and G09/31733.1) and was funded by the Danish Council for Independent Research|Natural Sciences (MK), the Knud Højgaards Fund, the Oticon Foundation, Thorsons Travel Grant and Københavns Universitets Fœlleslegat (ML, KB).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.00452/full#supplementary-material


surface for cloudless atmospheres. J. Clim. Appl. Meteorol. 25, 87–97. doi: 10.1175/1520-0450(1986)025<0087:SSSMFD>2.0.CO;2


light in the marine diatom Phaeodactylum tricornutum. PLoS ONE 4:e7743. doi: 10.1371/journal.pone.0007743


R. J. Thomas, G. J. Wilmink, and B. L. Ibey (San Francisco, CA: SPIE), 894117-1–894117-6.


**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 Lichtenberg, Brodersen and Kühl. 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.

## APPENDIX

#### **Definition of abbreviations and parameters.**


# Ecological Energetic Perspectives on Responses of Nitrogen-Transforming Chemolithoautotrophic Microbiota to Changes in the Marine Environment

Hongyue Dang<sup>1</sup> \* and Chen-Tung A. Chen<sup>2</sup>

<sup>1</sup> State Key Laboratory of Marine Environmental Science, Institute of Marine Microbes and Ecospheres, College of Ocean and Earth Sciences, Xiamen University, Xiamen, China, <sup>2</sup> Department of Oceanography, National Sun Yat-sen University, Kaohsiung, Taiwan

Transformation and mobilization of bioessential elements in the biosphere, lithosphere, atmosphere, and hydrosphere constitute the Earth's biogeochemical cycles, which are driven mainly by microorganisms through their energy and material metabolic processes. Without microbial energy harvesting from sources of light and inorganic chemical bonds for autotrophic fixation of inorganic carbon, there would not be sustainable ecosystems in the vast ocean. Although ecological energetics (eco-energetics) has been emphasized as a core aspect of ecosystem analyses and microorganisms largely control the flow of matter and energy in marine ecosystems, marine microbial communities are rarely studied from the eco-energetic perspective. The diverse bioenergetic pathways and eco-energetic strategies of the microorganisms are essentially the outcome of biosphere-geosphere interactions over evolutionary times. The biogeochemical cycles are intimately interconnected with energy fluxes across the biosphere and the capacity of the ocean to fix inorganic carbon is generally constrained by the availability of nutrients and energy. The understanding of how microbial eco-energetic processes influence the structure and function of marine ecosystems and how they interact with the changing environment is thus fundamental to a mechanistic and predictive understanding of the marine carbon and nitrogen cycles and the trends in global change. By using major groups of chemolithoautotrophic microorganisms that participate in the marine nitrogen cycle as examples, this article examines their eco-energetic strategies, contributions to carbon cycling, and putative responses to and impacts on the various global change processes associated with global warming, ocean acidification, eutrophication, deoxygenation, and pollution. We conclude that knowledge gaps remain despite decades of tremendous research efforts. The advent of new techniques may bring the dawn to scientific breakthroughs that necessitate the multidisciplinary combination of eco-energetic, biogeochemical and "omics" studies in this field.

Keywords: carbon cycle, chemolithoautotrophy, energy metabolism, global change, global warming, nitrogen cycle, ocean acidification, ocean deoxygenation

#### Edited by:

Karla B. Heidelberg, University of Southern California, United States

#### Reviewed by:

Zhe-Xue Quan, Fudan University, China Martin Koenneke, MARUM – Center for Marine Environmental Sciences, University of Bremen, Germany

> \*Correspondence: Hongyue Dang danghy@xmu.edu.cn

#### Specialty section:

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

Received: 07 December 2016 Accepted: 20 June 2017 Published: 14 July 2017

#### Citation:

Dang H and Chen C-TA (2017) Ecological Energetic Perspectives on Responses of Nitrogen-Transforming Chemolithoautotrophic Microbiota to Changes in the Marine Environment. Front. Microbiol. 8:1246. doi: 10.3389/fmicb.2017.01246

**Abbreviations:** ANME, anaerobic methane oxidation; CBB, Calvin-Benson-Bassham cycle; rTCA, reductive tricarboxylic acid cycle; Comammox, complete oxidation of ammonia to nitrate; Anammox, anaerobic ammonium oxidation; DC/4-HB, dicarboxylate/4-hydroxybutyrate cycle; WL, Wood-Ljungdahl pathway (i.e., reductive acetyl-CoA pathway); 3-HP/4-HB, 3 hydroxypropionate/4-hydroxybutyrate cycle.

## INTRODUCTION

fmicb-08-01246 July 12, 2017 Time: 15:25 # 2

Ecological energetics (eco-energetics) is the study of energy flow and transformations in an ecosystem or through a population in a specific environment (Odum, 1968). Energy flow is a basic property of any ecosystem. From sunlit seawater to dark deep ocean and marine sediments, microorganisms employ various energy-transducing strategies to carry out ecological and biogeochemical functions (Kolber, 2007; Falkowski et al., 2008). The diverse microbial eco-energetic strategies are essentially a result of evolution during long-term biosphere-geosphere interactions (Nitschke et al., 2013; Sousa et al., 2013; Jelen et al., 2016), as vitally put by Lane et al. (2013): "if nothing in biology makes sense except in the light of evolution, nothing in evolution makes sense except in the light of energetics."

Microorganisms constitute the most abundant, diverse and metabolically active component of biomass in the marine environment (Azam, 1998; Kallmeyer et al., 2012). The microbial communities largely control the flow of energy from abiotic forms to higher trophic levels in the ocean (Azam et al., 1983; DeLong, 2009; Brown et al., 2014). For example, marine photolithoautotrophic and chemolithoautotrophic microorganisms harvest and transform energy from otherwise largely bio-inaccessible sources (e.g., light and inorganic chemical bonds) to forms useable by chemoorganoheterotrophic consumers such as protists and animals (Brown et al., 2014). With this primary eco-energetic service, microorganisms set and control the reduction-oxidation (redox) and energy states of their environment, provide ecological services and influence the climate-mediating potential of the ocean (Azam and Malfatti, 2007; Falkowski et al., 2008; Carpenter et al., 2012). Energy is the ultimate limiting factor in determining the structure and function of the Earth ecosystem (Odum, 1968). While this makes eco-energetics "the core of ecosystem analysis" (Odum, 1968), marine microbial communities are seldomly studied from the eco-energetic perspective (Kolber, 2007; Vallino and Algar, 2016).

The marine biogeochemical cycles are driven by the microbial engines (Falkowski et al., 2008), which are mainly fuelled by energy conserved through microbial metabolic processes (**Figure 1**) (Falkowski and Godfrey, 2008; Orcutt et al., 2011). However, modeling studies of ecosystem metabolism including most recent ones usually ignored marine bacteria and archaea completely or considered them solely as decomposers (Heymans et al., 2014). This contradicts the diverse ecofunctions including the widespread autotrophy of the bacterial and archaeal communities in the ocean (Berg et al., 2010; Fuchs, 2011; Hügler and Sievert, 2011). In stark contrast to higher organisms such as plants and animals, bacteria and archaea employ diverse and complex energy metabolic pathways (Kolber, 2007), which are adapted to and effective in diverse environments. Microbial communities select energetically favorable electron donors and acceptors from their environment for energy transduction (Bar-Even et al., 2012; Eggleston et al., 2015). Even so, energy may be a limited resource for certain marine ecosystems (Burgin et al., 2011; Moore et al., 2013; Vallino and Algar, 2016). The sources and sustainability of energy supplies largely control the diversity and actual rates of the energy metabolic pathways and thus the composition, structure, and function of the microbial communities (Kolber, 2007; Dahle et al., 2015).

Microbial chemolithotrophic metabolism was discovered in the 1880s by Sergei Winogradsky, a pioneer in microbial ecology (Dworkin, 2012). Similar to photolithoautotrophs, chemolithoautotrophs contribute to primary production, which is, however, fueled by energy conserved from aerobic or anaerobic oxidation of inorganic electron donors (e.g., NH3, NH<sup>4</sup> +, NO<sup>2</sup> <sup>−</sup>, S2−, S, H2, CO, and Fe2+). Some microorganisms are chemoorganoautotrophs that oxidize organic chemicals (e.g., CH4) to conserve energy for carbon fixation, such as the anaerobic methane-oxidizing (ANME) archaea (Kellermann et al., 2012).

A series of exciting discoveries of new chemoautotrophic microorganisms and their bioenergetic pathways were made around the turn of the 21st century. In 1999, for example, chemolithoautotrophic anaerobic ammonium-oxidizing (anammox) bacteria were discovered and subsequently found to be widely distributed in oxygen-deficient and oxygendepleted seawater and sediments, carrying out an important biogeochemical process in biological nitrogen removal from the ocean (Strous et al., 1999; Dalsgaard et al., 2003; Kuypers et al., 2003; Oshiki et al., 2016). The turn of the last century also witnessed the discovery of the ANME archaea and their consortial association with sulfate-reducing bacteria (Hinrichs et al., 1999; Boetius et al., 2000). ANME play an important role in the removal of methane, a potent greenhouse gas, from deep-sea cold-seep sediments and many other methane-rich environments (Knittel and Boetius, 2009; Marlow et al., 2016). Certain ANME archaea also harbor the genetic and biochemical inventory for N<sup>2</sup> fixation (Pernthaler et al., 2008; Dang et al., 2009a, 2013a; Dekas et al., 2009; Miyazaki et al., 2009), thereby contributing to coupled carbon, sulfur, and nitrogen cycling in methane-rich, sulfate-rich but nitrogen-poor environments (Fulweiler, 2009). In 2002, neutrophilic, chemolithoautotrophic iron-oxidizing bacteria (FeOB) were found to be abundant in deep-sea hydrothermal vent environments (Emerson and Moyer, 2002) and they were later classified as a new class of the Proteobacteria: the Zetaproteobacteria (Emerson et al., 2007). Subsequent investigations indicate that zetaproteobacterial FeOB exist in coastal seawater and sediments as well, where they participate in coastal iron cycling and biocorrosion of man-made iron constructs (Dang et al., 2008a, 2011; McBeth et al., 2011; McBeth and Emerson, 2016). Recently, zetaproteobacterial FeOB were also discovered to utilize ferrous iron (Fe2+) from deep-sea basaltic rocks and basaltic glasses as an energy source, thereby contributing to carbon fixation in ultra-oligotrophic abyssal plains (Orcutt et al., 2015; Henri et al., 2016).

Since their discovery in the 1880s, chemolithoautotrophic ammonia-oxidizing bacteria (AOB) were believed to be the sole microorganisms responsible for biological ammonia oxidation in oxic environments. This long-lasting misconception was refuted in 2005 by the discovery of chemolithoautotrophic ammoniaoxidizing archaea (AOA) (Francis et al., 2005; Könneke et al., 2005), which are affiliated with the newly defined phylum, Thaumarchaeota (Brochier-Armanet et al., 2008; Spang et al.,

2010). The most recent addition to the metabolic diversity of chemolithoautotrophic microorganisms was the bacterial strains in the nitrite-oxidizing genus Nitrospira capable of carrying out complete oxidation of ammonia to nitrate (comammox) (Daims et al., 2015; van Kessel et al., 2015). The history of chemoautotroph studies indicates that the ocean is full of surprises and opportunities of unknown microorganisms and novel bioenergetic strategies.

The discovery of chemolithoautotrophy by Winogradsky ended the long-lasting misconception that photoautotrophic organisms such as plants and algae are the sole primary producers on Earth (Dworkin, 2012). The discoveries of diverse chemolithic bioenergetic pathways contributed greatly to the understanding of the complexity of energy flow in Earth ecosystems and the interdependency of the biogeochemical cycles involving carbon, nitrogen, sulfur, iron, and other bioessential elements (**Table 1**). The discoveries of chemosynthetic ecosystems in the deep-sea hydrothermal vent and cold-seep environments as "oases" in the vast deep ocean "deserts" were true scientific thrills in the 70s and 80s of the 20th century, spotlighting the cornerstone species role in community structure and the primary producer role in trophic transfer the chemoautotrophic bacteria and archaea play in these sunlight-independent marine ecosystems (Felbeck and Somero, 1982; Paull et al., 1985; Jørgensen and Boetius, 2007). They provided the first evidences about the importance of microbial chemolithoautotrophy for energy and matter flows in nature and stimulated the search of life's origin on Earth and beyond (Nisbet and Sleep, 2001; Martin et al., 2008).

Chemolithoautotrophic microorganisms may contribute substantially to primary production in non-extreme marine environments as well. For example, dark carbon fixation in the redox transition zone of the Cariaco Basin, mainly carried out by chemolithoautotrophic sulfur-oxidizing bacteria (SOB) fueled with seawater reduced sulfur species, was equivalent to 10–333% of the local surface ocean photosynthetic primary production (Taylor et al., 2001). Microbial chemolithoautotrophs also contribute substantially to primary production in oxygenated dark oceans. Carbon fixation in meso- and bathypelagic waters of the North Atlantic, presumably by chemolithoautotrophic AOA, could amount to 15–53% of phytoplankton export production from surface water (Reinthaler et al., 2010). Dark carbon fixation in Tyrrhenian deep waters of Central Mediterranean Sea by chemolithoautotrophic microorganisms (mainly affiliated with AOA) was comparable to photosynthetic production (Yakimov et al., 2011). On the global scale, seawater AOA may fix approximately 400 Tg C y−<sup>1</sup> (Hügler and Sievert, 2011). In the middle (i.e., the twilight zone) and deep ocean, chemolithoautotrophs also contribute to the production and accumulation of quite an amount (5–10 µM) of semi-labile dissolved organic carbon (DOC) (Follett et al., 2014), which may be further transformed by microorganisms to produce recalcitrant DOC (RDOC) (Jiao et al., 2014). The average

turnover time of deep ocean RDOC reaches millennial timescales (Hansell, 2013). Dilution and structural recalcitrance preclude microbial consumption, constituting the major mechanisms for long-term sequestration of marine RDOC (Jiao et al., 2014; Arrieta et al., 2015; Moran et al., 2016). The deep-ocean DOC concentrations maintain small (∼40 µM) and relatively constant (Chen, 2011; Hansell, 2013), which, however, sustain active microbial communities (Arrieta et al., 2015). The in situ primary production of chemolithoautotrophs may provide an important source of organic matter, in addition to that released from sinking particles, hydrothermal vents and cold seeps (Chen, 2011), to fuel the activities of the deep-ocean microbiota. However, it is currently unknown whether the different sources of DOC (e.g., produced by in situ chemolithoautotrophs or released from sinking particles, hydrothermal vents or cold seeps) may have different molecular structures and bio-utilizabilities and thus different residence times in the ocean. Furthermore, although on average the contribution of the chemolithoautotrophic microbiota to ocean's carbon fixation may be substantial, the in situ chemolithoautotrophic carbon fixation rates are highly variable among different marine environments (Taylor et al., 2001; Reinthaler et al., 2010). The in situ energy sources (e.g., bioavailable inorganic reductants and oxidants) may exert substantial influences on the abundance, diversity, activity, distribution, and dynamics of marine chemolithoautotrophs, and thus the energy environment may play important roles in chemolithoautotrophic carbon fixation and carbon sequestration (Dang and Jiao, 2014). Further systematic investigations are needed to quantitatively understand the roles of the microbial chemolithoautotrophs in ocean's carbon budget and dynamics and in microbe-environment interactions.

Global change caused by anthropogenic activities may alter the physical, chemical and energy environment of the marine ecosystems and thus alter the spatiotemporal dynamics and functions of the microbiota (Kolber, 2007; Hutchins et al., 2009; Middelburg and Levin, 2009; Dang and Jiao, 2014). To understand how the microbial eco-energetic processes influence the structure and function of the marine ecosystems and how they respond to and exert impacts on the changing marine environment is fundamental to a mechanistic and predictive understanding of the global carbon cycle and the ocean's climate-modulating capacity. Because of the tremendous diversity of the marine microorganisms, even just considering the chemolithoautotrophs (Berg et al., 2010; Berg, 2011; Hügler and Sievert, 2011), it is practical and meaningful to divide a community into distinct functional groups in microbial eco-energetic studies. The nitrogen cycle is probably the most perturbed biogeochemical cycle due to human activities (Rockström et al., 2009). Therefore in this review, we focus on the major functional groups of chemolithoautotrophic bacteria and archaea that are involved in marine nitrogen cycling (**Figure 2**) to tentatively illustrate their energetic strategies, ecological processes, contributions to carbon cycling, and responses to and impacts on global change associated with global warming, ocean acidification, eutrophication, deoxygenation, and pollution.

## DIVERSE CHEMOLITHOAUTOTROPHIC PATHWAYS IN MARINE NITROGEN-CYCLING BACTERIA AND ARCHAEA

Dissolved inorganic nitrogen (DIN) compounds are used as either nutrients for assimilatory biomass production or electron donors or electron acceptors that are transformed by dissimilatory cellular redox reactions for microbial energy transduction (**Figure 2**). Activity of chemolithoautotrophic nitrogen-cycling bacteria and archaea is generally inhibited by light (for AOB and AOA) or by oxygen (for anammox bacteria) and outcompeted by phytoplankton for ammonium uptake in the marine photic zone (Arrigo, 2005; Merbt et al., 2012; Smith et al., 2014). Thus, DIN dissimilatory utilization for energy transduction happens mainly in the twilight and dark zones of the ocean, where aerobic oxidation of ammonia and nitrite occurs under oxic and hypoxic conditions and anammox (and denitrification) occurs under suboxic and anoxic conditions.

## Diverse Eco-energetic Strategies of Marine Chemolithoautotrophic Nitrifying Microorganisms

Nitrification is carried out mainly by chemolithoautotrophs, in two separated steps either by AOB and AOA for aerobic oxidation of ammonia to nitrite and NOB for aerobic oxidation of nitrite to nitrate or by comammox bacteria for the joint aerobic oxidation of ammonia and nitrite (Arp et al., 2007; Kuypers, 2015; Daims et al., 2016). Although most seawater environments in the surface ocean are oligotrophic, nitrification occurs throughout the water column, with the only exception likely in the core of the anoxic marine zones (Ulloa et al., 2012). Formation of the primary nitrite maximum (PNM) at the base of the marine euphotic zone in stratified water columns may be caused by phytoplankton excretion (Lomas and Lipschultz, 2006; Beman et al., 2012). However, recent studies have shown that ammonia oxidation by AOA may actually produce the major source of nitrite in PNM (Beman et al., 2013; Buchwald and Casciotti, 2013; Santoro et al., 2013). Ammonia oxidation by AOA also contributes, to variable degrees, to the formation of the secondary nitrite maximum frequently observed in the oceanic oxygen minimum zones (OMZs) (Lam et al., 2011). AOA constitute the most abundant functional group of microorganisms in the ocean's mesopelagic and bathypelagic zones (Kirchman et al., 2007; Bristow et al., 2015), and they play a major role in nitrification and dark carbon fixation in the interior of the ocean (Herndl et al., 2005; Ingalls et al., 2006; Follett et al., 2014; Berg et al., 2015a). The deep oceans maintain a large reservoir of nitrate, associated mainly with the in situ AOA abundance (Herndl et al., 2005; Yakimov et al., 2011).

Ammonia-oxidizing archaea are usually more abundant and active than AOB in the ocean, particularly in oligotrophic environments (Stahl and de la Torre, 2012; Corinaldesi, 2015). The prevalence of nitrifying activity by marine AOA is mainly due to their extremely high specific affinity for ammonia and their


TABLE 1| Typical bacterial and archaeal chemolithoautotrophic metabolic pathways found in marine environments.

fmicb-08-01246 July 12, 2017 Time: 15:25 # 5

environmental adaptivity to low concentrations of ammonium and oxygen (Martens-Habbena et al., 2009; Hatzenpichler, 2012; Stahl and de la Torre, 2012; Horak et al., 2013; Offre et al., 2014; Qin et al., 2014; Corinaldesi, 2015). Two distinct marine ecotypes ("shallow" clade vs. "deep" clade) of AOA exist (Hatzenpichler, 2012; Luo et al., 2014), each may be adapted to distinct light and nutrient regimes of the water column. Some marine AOA may utilize ammonium instead of ammonia as the preferred energy substrate (Qin et al., 2014). Many AOA can also hydrolyze urea to utilize the ureolytic products (i.e., ammonia and CO2) for coupled ammonia oxidation and carbon fixation (Yakimov et al., 2011; Alonso-Sáez et al., 2012; Connelly et al., 2014; Offre et al., 2014; Qin et al., 2014; Bayer et al., 2016; Tolar et al., 2016). Urea utilization thus represents a "short cut" and eco-energetically economic pathway between nitrification and carbon fixation in environmental AOA (Kirchman, 2012). Unlike AOB that use the Calvin-Benson-Bassham (CBB) cycle for carbon fixation, AOA use the most energetically efficient 3-hydroxypropionate/4 hydroxybutyrate (3-HP/4-HB) cycle pathway for carbon fixation (Hügler and Sievert, 2011; Könneke et al., 2014). AOA contribute substantially to nitrification and dark carbon fixation even in hypoxic seawater, and their activities, albeit reduced, can still be detected under sulfidic conditions (Berg et al., 2015b). Moreover, some AOA may have the capacity of coping with phosphorus scarcity in marine environments. They harbor the pst gene that encodes for the high-affinity, high-activity phosphate ABC transporters (Dang et al., 2013b). Some AOA also produce inorganic phosphite and organic phosphorus compounds such as phosphonates, potentially serving as phosphorus storage mechanisms for metabolic sustainability under phosphorusstarving environmental conditions (Metcalf et al., 2012; Stahl and de la Torre, 2012; Dang et al., 2013b; Van Mooy et al., 2015). The production, processing, and uptake of these phosphorus compounds (in an oxidation state of +3) are highly energetically expensive, putatively indicating the importance of phosphorus to AOA in oligotrophic environments (Dang et al., 2013b). Alternatively, these ecophysiological traits may be evolutionary relics of ancient AOA, which experienced severe scarcity of phosphorus in the pre-anthropogenic ocean (Benitez-Nelson, 2015; Van Mooy et al., 2015). Furthermore, some environmental AOA assemblages were predicted to be mixotrophic (Dang et al., 2008b, 2010c) and certain AOA isolates are able to achieve maximum bioenergetic and growth efficiency with the availability of labile organic matter (Qin et al., 2014). The AOA mixotrophic potential has been challenged by the finding that organic matter is used by certain AOA isolates for non-enzymatic detoxification of hydrogen peroxide rather than as assimilable carbon source (Kim et al., 2016). However, a recent study indicates that the genomes of certain AOA do harbor key genes that encode peroxidases and catalases for coping with oxidative stress (Sauder et al., 2017). The diverse ecotypes and ecophysiological potentials of the numerically dominant AOA warrant further systematic investigations of their contributions to

the ocean's carbon, nitrogen and phosphorus cycling and energy flux.

In marine environments, although AOA are ubiquitous and usually the dominant ammonia oxidizers, AOB may occupy particular niches and play important biogeochemical roles as well. Marine particles harbor abundant and active AOB (Karl et al., 1984). Most AOB in true marine environments are affiliated with the Nitrosospira genus, while Nitrosomonas AOB usually prevail in terrestrially impacted marine environments such as estuaries and coastal bays (Dang et al., 2010b). In stark contrast to this general AOB distribution pattern, nevertheless, marine particle-associated AOB are mainly affiliated with the Nitrosomonas genus (Karl et al., 1984; Phillips et al., 1999). Microbial hydrolysis of marine particleassociated organic nitrogenous compounds may produce high concentrations of NH<sup>4</sup> <sup>+</sup> and NH<sup>3</sup> (Shanks and Trent, 1979; Gotschalk and Alldredge, 1989). Locally enriched ammonia may meet the need of those AOB (e.g., certain Nitrosomonas bacteria) that require high ammonia availability for energy transduction. Both Nitrosomonas and Nitrosospira AOB are affiliated with Betaproteobacteria. In marine environments, there exists another genus of AOB, the Nitrosococcus that is affiliated with Gammaproteobacteria (Arp et al., 2007). Interestingly, Nitrosococcus AOB are found exclusively in marine environments and the optimal growth conditions of all known Nitrosococcus isolates in culture need at least 50 mM NH<sup>4</sup> <sup>+</sup> at pH 7.5 to 8.0 (Campbell et al., 2011; Wang et al., 2016). This implies that the niche of Nitrosococcus AOB may be mainly associated with estuarine and coastal sedimentary environments (Wang et al., 2016), where high concentrations of NH<sup>4</sup> <sup>+</sup> are usually found (e.g., Dang et al., 2008b, 2010b). It is also likely that the marine Nitrosococcus AOB are continuously under the environmental pressure of low energetic substrate availability, partially explaining their low abundance commonly found in seawater. The distinct distribution patterns of Nitrosomonas, Nitrosospira, Nitrosococcus, and AOA in the ocean may well reflect the different energy environments (e.g., the availability of NH3) they are dwelling.

To complete the nitrification process, nitrite produced by AOA and AOB needs to be further oxidized to nitrate, which is carried out by NOB. The oxidation of nitrite provides very low energy gain. Furthermore, the activity and growth of NOB may be limited by substrate availability in marine environments. These may be the major reasons that the abundance of NOB in most marine environments is usually very low (Füssel et al., 2012; Beman et al., 2013), except in the brine-seawater interface layer of the Red Sea where Nitrospina-like NOB may constitute up to onethird of the bacterial community (Ngugi et al., 2016). In contrast, AOA usually constitute a much higher fraction, up to or even greater than 40% of the total bacterial and archaeal community in the mesopelagic and bathypelagic zones of the ocean (Herndl et al., 2005; Kirchman et al., 2007; Bristow et al., 2015).

In spite of these bioenergetic and eco-energetic disadvantages, NOB are widespread in seawater and sediment environments. The strategy of NOB to overcome the constraint of low energy gain from nitrite oxidation is to produce high amounts of nitrite oxidoreductase, the key enzyme for nitrite oxidation (Spieck et al., 2014). This channels more cellular metabolic energy to maintenance rather than to growth. Nitrite availability has recently been identified as a key factor driving niche differentiation in NOB (Nowka et al., 2015). In addition, to overcome the problem of low and varying nitrite concentrations in oxic seawater, some NOB can degrade certain simple dissolved organic nitrogen (DON) compounds such as urea and cyanate and reciprocally feed AOB with the degradation product NH<sup>3</sup> for greater nitrite production (Koch et al., 2015; Palatinszky et al., 2015). Metagenomic screening has shown that ureaseand cyanase-harboring NOB may be prevalent in environments (Koch et al., 2015; Palatinszky et al., 2015). It is reasonable to hypothesize that metabolic collaborations between NOB and ammonia oxidizers (i.e., AOB and AOA) in seawater may be facilitated in marine particle- and biofilm-associated microenvironments where cross-feeding is favored (Damashek et al., 2016; Dang and Lovell, 2016).

For more than a century, nitrification was accepted as a two-step biogeochemical process carried out sequentially by ammonia oxidizers and NOB. However, it was predicted that there exist some bacteria that can catalyze comammox, a process that is energetically feasible (Costa et al., 2006). Comammox bacteria have recently been discovered as unique sublineage II Nitrospira (Daims et al., 2015; van Kessel et al., 2015). Nitrospira are globally distributed, and similarly, functional gene biomarkers of comammox bacteria have been found to be prevalent in engineered and natural environments including marine sediments (Daims et al., 2015; van Kessel et al., 2015). The combination of the divided labors of ammonia oxidizers and NOB in the comammox bacteria bestows certain eco-energetic advantages, including facilitated acquisition of nitrite as an energy substrate and enhanced energy yield in terms of adenosine triphosphate (ATP) production (Costa et al., 2006; Daims et al., 2016). However, comammox also confers certain disadvantages to the bacteria. ATP is produced for catalytic purposes rather than for energy storage in cells (Pfeiffer and Schuster, 2005). The long metabolic pathway of comammox lowers the ATP production rate and thus the maximal growth rate the bacteria can achieve (Costa et al., 2006). The comammox process is predicted to be favored when ammonia and nitrite as energetic substrates are limiting and replenished slowly and when bacteria grow in clonal patches such as in biofilms (Costa et al., 2006). So far, no comammox bacteria have been found in marine waters (Kuypers, 2015). However, particles in seawater may present unique niches for comammox bacteria.

#### The Unique Eco-energetic Mechanism of Chemolithoautotrophic Anammox Bacteria

In oxygen-deficient and oxygen-depleted sediments and marine waters such as those occurring permanently in oceanic OMZs and seasonally in eutrophic coastal areas, microorganisms carrying out anammox and denitrification contribute to fixed nitrogen removal (Devol, 2015). Anammox bacteria employ the redox reaction of coupled nitrite reduction and ammonium oxidation for energy transduction to fix carbon (Kartal et al., 2013;

Oshiki et al., 2016), while bacterial and archaeal denitrifiers are usually heterotrophs that use organic matter as electron donors for stepwise reduction of nitrate, nitrite, NO, and N2O to produce N<sup>2</sup> (Carolan et al., 2015; Devol, 2015). Anammox bacteria harbor anammoxosomes, unique organelles functionally analogous to eukaryotic mitochondria to perform the energetic reaction (van Niftrik and Jetten, 2012; Jogler, 2014). Anammox bacteria are affiliated with a narrow bacterial clade, the Candidatus Brocadiales order of Planctomycetes (Kartal et al., 2013). In addition to physiological specialization and phylogenetic segregation, niche separation is also prevalent in anammox bacteria. Ca. Scalindua mainly occur in marine environments and all the other anammox bacterial genera are adapted to low-salinity habitats (Kartal et al., 2013). Estuaries are an exception where Ca. Scalindua and some freshwater anammox bacteria may both exist (Dang et al., 2010a, 2013c). In contrast, marine denitrifiers are common in Bacteria and Archaea, a phenomenon likely being facilitated by horizontal gene transfer for denitrifying trait spreading (Jones et al., 2008). Furthermore, most denitrifiers are facultative anaerobes and can rapidly switch among different energetic pathways in response to changing environmental conditions (Dang and Jiao, 2014), while anammox bacteria are obligate anaerobes and they may prefer stable environmental conditions.

As anammox bacteria and denitrifying microorganisms occupy similar environments, they may compete for nitrate and nitrite (as electron acceptors) for energy transduction. The bioavailability of organic carbon and organic matter stoichiometry may be key factors determining the relative contributions of anammox and denitrification to fixed nitrogen removal in the ocean (Thamdrup and Dalsgaard, 2002; Engström et al., 2005; Babbin et al., 2014; Chang et al., 2014; Babbin et al., 2016). Most organic compounds that are microbiologically utilizable as electron donors can be more easily oxidized and thus more energy-favorable than ammonium. This energetic difference influences the distribution of the anammox bacteria and denitrifiers in the ocean and their relative contributions to the marine nitrogen and carbon cycling (Dang et al., 2009b, 2010a; Ulloa et al., 2012). In suboxic and anoxic environments that are rich in organic matter such as some eutrophic coastal waters and sediments, heterotrophic denitrifiers usually outperform chemolithoautotrophic anammox bacteria for fixed nitrogen removal, while in certain oceanic OMZs the contribution of anammox bacteria may match or even outperform that of the heterotrophic denitrifiers (Ward et al., 2009; Ulloa et al., 2012). Similarly, organicpoor deep-sea sediments usually favor anammox activity over denitrification (Jaeschke et al., 2010; Devol, 2015). However, the influence of organic carbon on the partitioning of nitrogen loss between anammox and denitrification may be more complicated than previously thought. It was found recently that organic matter enrichment may stimulate dissimilatory nitrate reduction to ammonium (DNRA), competing against denitrification for nitrate acquisition (Brin et al., 2017). The complex microbial nitrogen transformation processes and their distinctly different responses to specific suboxic and anoxic marine environments constitute an obstacle to a simple and predictive understanding of the microbe-environment interactions.

#### ECO-ENERGETIC RESPONSES OF NITROGEN-CYCLING CHEMOLITHOAUTOTROPHS TO GLOBAL CHANGE IMPACTS

Chemolithoautotrophic microorganisms have been playing critical roles in shaping the Earth's environment and planetary evolution (Fuchs, 2011; Martin et al., 2014). Although photoautotrophs (i.e., cyanobacteria, algae, and plants) are the most dominant carbon fixers on the surface layer of land and ocean, chemolithoautotrophic microorganisms have been carrying out carbon fixation long before the advent of oxygenic photosynthetic organisms (Fuchs, 2011; Braakman and Smith, 2012; Jelen et al., 2016). Some chemolithoautotrophs were probably among the first organisms on earth and have played a key role in driving the transition of earth from its primordial inorganic state to a state rich in biogenic organic matter. They may have made substantial impacts on the geochemical condition and redox status of the primitive Earth, leading to the origination and evolution of the planet ecosystems and biogeochemical cycles.

Anthropogenic activities have significantly changed the Earth's environments and ecosystems as well (Duarte, 2014; Brondizio et al., 2016). Rapid increase in CO<sup>2</sup> emission as a result of fossil fuel consumption has led to various environmental problems such as global warming, ocean acidification, and ocean hypoxia. Carbon perturbation induced by ongoing global change is further compounded by other harmful human activities such as environmental pollution and eutrophication (Bijma et al., 2013). How these anthropogenic environmental disturbances influence the eco-energetics and biogeochemical functions of the marine microbiota warrants in-depth examinations, particularly for the chemolithoautotrophic microorganisms that play important roles in carbon fixation and carbon sequestration of the ocean.

#### Chemolithoautotrophic Responses to Global Warming

Temperature is an important environmental factor that influences microbial ecophysiology and biogeochemical functioning in multiple and profound ways. For example, the permeability of proton across cytoplasmic membranes, which plays a critical role in microbial bioenergetic processes, increases with environmental temperature (Tolner et al., 1997; Berry, 2002). Proton leakage lowers the microbial bioenergetic efficiency. Therefore, microorganisms continuously monitor surrounding temperature and abrupt temperature changes usually induce rapid microbial stress responses (Schumann, 2009). It was predicted that a 1◦C increase of seawater temperature in the bathypelagic ocean may cause a 55% increase of heterotrophic production by the in situ bacterial and archaeal community (Lønborg et al., 2016). However, autotrophic and heterotrophic microorganisms may respond to ocean warming

differently, leading to changes in the ocean's metabolic balance (Lønborg et al., 2016).

Nitrogenous nutrient scarcity tends to limit primary production of the phytoplankton communities throughout much of the low-latitude surface ocean (Moore et al., 2013). Eco-energetically, chemolithoautotrophic nitrogen-cycling microorganisms carry out dissimilatory transformations of nitrogenous compounds, resulting in the compositional changes and/or loss of nitrogenous nutrients from environments (Stein and Klotz, 2016). They contribute to marine carbon cycling not only by direct carbon fixation but also by their control on the speciation, concentration, and distribution of nitrogenous nutrients in the ocean, thus influencing the composition, structure, activity, and spatiotemporal dynamics of the photosynthetic communities and their primary productivity. The interactions of nitrogen-cycling microorganisms with photosynthetic communities via the linkage of nutrients are highly complex and influenced by diverse factors. For example, the deep ocean is usually rich in nitrate, resulted mainly from the activities of chemolithoautotrophic ammonia- and nitriteoxidizers. However, the deep ocean is devoid of sun light and thus phototrophic activities. This creates a spatial decoupling of nutrients with solar light energy, limiting substantially the primary productivity and carbon sequestration capacity of the ocean. With the impact of global warming, the surface ocean is becoming warmer and more stratified, aggravating the spatial decoupling of deep ocean nutrients and surface ocean photosynthetic CO<sup>2</sup> fixation.

Aerobic ammonia oxidation by AOA has been found as one of the highest energy-yielding chemolithotrophic processes in high temperature environments such as hot springs (Dodsworth et al., 2012; Hatzenpichler, 2012). Physiological, genomic, and phylogenetic analyses suggest that the ancestor of AOA was thermophilic and a number of studies support this inference by showing the prevalence of AOA in terrestrial hot spring and coastal and deep-sea hydrothermal vent environments (Hatzenpichler, 2012; Stahl and de la Torre, 2012; Wang L. et al., 2015). The ubiquity of abundant AOA in mesopelagic and bathypelagic seawater and deep-sea sediments suggests that their mesophilic and psychrophilic physiology may be the result of secondary adaptations (Hatzenpichler, 2012; Stahl and de la Torre, 2012). On the contrary, AOB may have a mesophilic origin. A microcosm experiment showed that the composition of soil AOB usually changes very little with increasing temperature, while the abundance and ammonia oxidation potential activity of certain AOA phylotypes increase significantly with increasing temperature (Tourna et al., 2008). In line with this, a recent study showed that the biochemical processes of ammonia oxidation may be thermodynamically different between soil AOA and AOB, with AOA having a significantly higher minimum temperature than AOB for ammonia-oxidizing activity (Taylor et al., 2017). Although it is unknown if marine AOA and AOB bear a similar thermodynamic difference, temperature was putatively identified as a key environmental factor affecting the composition and distribution of the AOA assemblages in sediments of the South China Sea, a large marginal sea in the subtropical and tropical Pacific Ocean (Dang et al., 2013b). Does this mean that the marine AOA may be more responsive than AOB to ocean warming? The polar ecosystems are currently facing strong global warming effects (Schofield et al., 2010; Spielhagen et al., 2011). However, a recent investigation in western coastal Arctic seawater shows that nitrification rates, likely contributed mainly by AOA, are resistant to short-term temperature elevations (Baer et al., 2014). A common trend of the warming effects on the diverse marine nitrifying microbiota cannot be concluded yet. How and to what extent elevated seawater temperatures caused by global warming may directly affect the ecophysiology of marine AOB, AOA, and NOB are still difficult to predict.

In addition to the direct effects, global warming may also exert certain indirect effects on the marine nitrifying microbiota and their eco-energetic and carbon-fixing activities. Global warming causes ocean stratification that weakens the vertical mixing of seawater and makes surface water of the open oceans to be more oligotrophic. This effect may lower phytoplankton primary productivity and thus the export of particulate organic matter (POM) from the surface ocean to the deep ocean and sediments (Bijma et al., 2013; Turner, 2015). Marine particles and sinking phytoplankton aggregates are the hotspots of extracellular enzyme activities, contributing to microbial degradation of POM polymeric organic matter and release of nutrients and energy substrates such as ammonium, phosphate and labile DOC in the middle and deep waters of the ocean (Azam and Malfatti, 2007; Wright et al., 2012; Ploug and Bergkvist, 2015; Dang and Lovell, 2016; Krupke et al., 2016). Certain studies have suggested enhanced nitrification activities in marine waters that are rich in particles and phytoplankton aggregates (Karl et al., 1984; Klawonn et al., 2015a; Damashek et al., 2016). Diminished biological pump (BP) function caused by the global warming effects reduces not only the ocean's carbon sequestration capacity but also the nutrient replenishment rate in the interior of the ocean. Diminished ammonium supply due to weakened BP exports may lead to reductions of the nitrifiers' activities such as nitrate production and carbon fixation in the marine mesopelagic and bathypelagic zones.

A number of studies have shown that the anammox bacteria are well adapted to low-temperature environments such as deep-sea and polar region sediments (Rysgaard et al., 2004; Jaeschke et al., 2010; Russ et al., 2013; Canion et al., 2014a,b; Shao et al., 2014; Sonthiphand et al., 2014). Anammox bacteria can alter the composition of their cell membranes and thus enhance the membrane fluidity by increasing the content of short chain ladderane fatty acids in response to low-temperature conditions (Rattray, 2008). In cold environments, anammox bacteria are found to carry out psychrophilic anammox process, while denitrifying bacteria usually carry out psychrotrophic nitrogen removal process (Rysgaard et al., 2004; Canion et al., 2014a,b). Colder temperature thus may be an important environmental factor that favors anammox over denitrification in deep-ocean and polar sea sedimentary environments (Rysgaard et al., 2004; Canion et al., 2014a,b; Shao et al., 2014). The different temperature adaptations between anammox bacteria and denitrifiers suggest that increased temperatures caused by global warming may favor denitrification over anammox in cold marine environments. However, in a recent study, the

psychrophilic physiology of anammox bacteria could not be verified and anammox and denitrification were found to have similar temperature responses, which are not influenced by warming in temperate coastal environments (Brin et al., 2017). Another study found little changes of community structure and activity rate of anammox bacteria and denitrifiers in response to increased temperatures, suggesting both microbial groups may be ecophysiologically tolerant to climate warming disturbances (Canion et al., 2014b). More systematic ecology studies covering broader environmental conditions are necessary for revealing the true ecophysiological characteristics of the marine anammox bacteria and denitrifiers.

The anammox and denitrification microbial assemblages may be influenced by global warming via its indirect effects. Although both denitrification and anammox activities may be enhanced by marine particles and phytoplankton aggregates, seawater anammox bacteria are frequently found as free-living microorganisms except in a few cases (Woebken et al., 2007; Ganesh et al., 2015; Stief et al., 2016). Dissolved organic matter (DOM) and POM usually stimulate denitrification rates by providing the necessary energy and carbon sources for the denitrifiers (Canion et al., 2014b; Chang et al., 2014; Babbin et al., 2016), while the chemolithoautotrophic anammox bacteria rely less on organic matter for energy and material metabolisms. Therefore, reductions in POM flux and DOC availability induced by global warming may affect more negatively on denitrification than on anammox in the ocean's mesopelagic and bathypelagic zones and the deep-sea sediments (Canion et al., 2014b). However, in temperate estuarine and coastal sediment environments, a recent study showed that the role of organic matter in altering nitrogen removal partitioning between anammox and denitrification cannot be verified (Brin et al., 2017).

The controversial effects of temperature and organic matter on the activities of anammox and denitrification indicate the complexity of the microbial responses to the global warming effects. Microbial communities in different marine environments may have developed different ecophysiologies and habitat adaptivities. The key limiting environmental factors that control the anammox and denitrifying activities in different environments may be different as well. For example, microbial assemblages in temperate estuarine and coastal environments experience obvious seasonal changes in temperature, nutrients and various sources of organic matter inputs. In addition, estuarine and coastal environments usually experience eutrophication and pollutions. The microbiota here may have developed adaptations to these varying factors. Moreover, the different microbial nitrogen transforming processes may also be influenced by the scarcity of various trace metals as enzyme cofactors and/or by the toxicity of diverse heavy metals to enzymes, a critical scientific question in microbial eco-energetics that is not yet clearly solved (Klotz and Stein, 2008; Simon and Klotz, 2013; Glass et al., 2015; Löscher et al., 2016). This implies that some environmental conditions, other than temperature and organic matter, may be the most important factors controlling the denitrification and anammox activities in certain marine environments. Furthermore, complex interactions among the different microbial functional groups and between microorganisms and other organisms in the ecosystems may all influence the outcomes of the global warming effects. For example, addition of extra DOM or POM may stimulate the DNRA activity rather than the denitrification activity (Brin et al., 2017). The effects of protozoan grazing and viral lysis on the partitioning of nitrogen loss between denitrification and anammox in marine environments are not resolved, either (Löscher et al., 2016). The high degrees of complexity and uncertainty indicate that there is the need of more systematic investigations for a comprehensive and accurate understanding of the responses of marine anammox and denitrifying microorganisms to the impacts of ocean warming.

#### Chemolithoautotrophic Responses to Ocean Acidification

Increased anthropogenic CO<sup>2</sup> emission causes not only global warming, but also increased partial pressure of CO<sup>2</sup> (pCO2) and hence decreased pH in seawater. The most significant drops of pH are usually associated with estuarine and coastal seawater, caused additionally by terrestrial, anthropogenic, mixing and upwelling inputs of nutrients and organic matter that lead to enhanced primary production and microbial respiration (Cai et al., 2011; Lui et al., 2015). Increased atmospheric deposition of nitrogen and sulfur in coastal regions resulting from fossil fuel combustion and agricultural fertilizer application also lowers seawater pH (Doney et al., 2007; Hagens et al., 2014). Acidification may influence the ocean's primary productivity and carbon sequestration capacity (Doney et al., 2009). Acidification also changes the equilibrium of the ocean's carbonate chemistry system, leading to stresses and damages to certain sensitive ecosystems such as the shallow coral reefs (Andersson and Gledhill, 2013; O'Brien et al., 2016). Ocean acidification may reduce marine biodiversity and fisheries as well, due to acidification-induced animal physiological stresses and/or acidification-induced changes in the ecosystem's trophic transfer efficiency (Widdicombe and Spicer, 2008; Branch et al., 2013; Cripps et al., 2016; van Leeuwen et al., 2016).

Ocean acidification may exert significant impacts on marine biogeochemical cycles. For example, microbial photosynthesis and nitrogen fixation have been found to be enhanced under acidification conditions in the surface ocean, which may be related directly to enhanced inorganic carbon assimilation due to increased seawater pCO<sup>2</sup> (O'Brien et al., 2016). However, ocean acidification exerts a negative impact on the chemolithoautotrophic ammonia oxidation process. Ocean acidification changes the seawater NH3/NH<sup>4</sup> <sup>+</sup> equilibrium by ionizing more ammonia molecules to form ammonium cations. A 0.3 pH decrease projected to happen by the year 2100 (Caldeira and Wickett, 2005) may cause a 50% decrease of the seawater NH<sup>3</sup> concentration (Zeebe and Wolf-Gladrow, 2001). Studies have shown that this shift in NH3/NH<sup>4</sup> <sup>+</sup> equilibrium may directly reduce the ocean's ammonia oxidation rate (Beman et al., 2011). Although marine sediments have certain buffering effect against porewater pH changes (Kitidis et al., 2011), decreases of benthic nitrification

rate have also been reported in some investigations (Braeckman et al., 2014). However, the change of nitrite oxidation activity in response to ocean acidification may be quite different from that of ammonia oxidation activity. The response of the two-step nitrification process to ocean acidification may be more complicated that previous thought. A recent study showed that the nitrite oxidation rates of coastal seawater correlate positively with [H+] and thus negatively with pH (Heiss and Fulweiler, 2016). Currently it is not clear if this phenomenon is common in the world oceans, nor is known about the mechanism for this NOB response.

Furthermore, global warming may exert a compound effect along with ocean acidification on the change of marine nitrification. It was found that high temperature in summer had an inhibitory effect on the activity and growth of NOB, leading to the decoupling of ammonia oxidation and nitrite oxidation and thus the accumulation of nitrite in seawater (Bristow et al., 2015; Schaefer and Hollibaugh, 2017). Some NOB such as Nitrotoga spp. prefer environments with a slightly acidic pH and lowtemperatures (<20◦C) in physiological experiments (Hüpeden et al., 2016). Global warming increases the temperature of both the surface ocean and the ocean's interior (Masuda et al., 2010; Mora et al., 2013; Levin and Le Bris, 2015), thus it may negatively influence the rate of nitrite oxidation in certain environments of the ocean. Under the combined influences of ocean acidification and global warming, the chemical composition of the ocean's nitrogenous nutrient reservoir may be altered. It is well known that nitrate is the primary inorganic nitrogen source for marine diatoms, which contribute substantially to the BP-mediated particulate organic carbon (POC) export and storage in the ocean's interior (Bowler et al., 2010; Beman et al., 2011; Diner et al., 2016). On the contrary, marine dinoflagellates were found to be favored by ammonium, which also enhances algal bloom formation and toxin production (Leong et al., 2004; Hattenrath-Lehmann et al., 2015). The lowered nitrate pool due to ocean acidification- and warming-induced decrease of nitrification may lower diatoms-mediated primary production and POC-mediated carbon sequestration of the ocean but increase the incidences of harmful algal blooms by dinoflagellates.

Although certain soil AOB and AOA strains such as Ca. Nitrosotalea devanaterra have been found to be obligate acidophiles (Hayatsu, 1993; Lehtovirta-Morley et al., 2016), it is doubtable that the majority of the marine AOB and AOA are acidophiles because the seawater pH is usually above 7. The long-term lack of acidic conditions makes it difficult for marine AOB and AOA to evolve genetic and physiological mechanisms to become acidophilic. However, not all the bacterial and archaeal ammonia oxidizers respond to ocean acidification in the same way. Ocean acidification may change the composition of the ammonia-oxidizing communities, in which urease-harboring AOA and AOB may gain more importance as they can use urea as a source of ammonia and CO<sup>2</sup> for both energy transduction and carbon fixation (Koper et al., 2004; Klotz et al., 2006; Kirchman, 2012; Bowen et al., 2013). Urea is the most abundant chemical species of labile low-molecular-weight DON in the ocean (Solomon et al., 2010). Rapid autochthonous production by bacteria, algae, protists, and animals (e.g., zooplankton, mollusks, crustaceans, fish, and mammals) and various allochthonous inputs such as from agricultural fertilizers make urea an important constituent of the marine nitrogen cycle (Berman and Bronk, 2003; Glibert et al., 2006; Solomon et al., 2010). The urea molecule is uncharged and its chemistry does not change with pH. Urea is prevalent in seawater and sediment porewater (Glibert et al., 2006; Solomon et al., 2010), likely serving as a preferred and important nitrogen source for ureolytic ammonia oxidizers under acidified conditions (Pommerening-Röser and Koops, 2005; Lu and Jia, 2013). Indeed, ureolytic AOA have been found as the major ammonia oxidizers contributing significantly to the nitrification activity in certain marine environments (Alonso-Sáez et al., 2012; Tolar et al., 2016). Urea as an alternative substrate source for microbial ammonia oxidation may compensate, to currently unknown degrees, for the reduction of the marine nitrification activities caused by ocean acidification. In addition, some studies showed that the nitrite oxidation rates are greater than the ammonia oxidation rates in coastal seawater (Heiss and Fulweiler, 2016). Some sources of nitrite, other than that provided by in situ ammonia oxidation, may provide extra nitrite for the activity of NOB. Seawater cyanate may be used by NOB for "reciprocal feeding" of ammonia oxidizers to obtain extra nitrite (Palatinszky et al., 2015; Heiss and Fulweiler, 2016). However, currently the quantitative contributions of ureolytic ammonia oxidizers and cyanate-degrading NOB to nitrification in the global ocean under acidification conditions have not been in-depth and systematically investigated.

Ocean acidification may enhance anammox activity, partially resulting from increased ammonium concentrations due to the shift of the NH3/NH<sup>4</sup> <sup>+</sup> equilibrium under acidification conditions (Widdicombe and Needham, 2007; Gazeau et al., 2014; Tait et al., 2014). Ocean acidification may enhance the activity of denitrification as well, because the NO<sup>3</sup> <sup>−</sup>, NO<sup>2</sup> −, and NO reductases are more active at neutral or lower pH (Richardson et al., 2009). Therefore, it is likely that fixed nitrogen loss from the ocean will increase due to ocean acidification. Furthermore, ocean acidification may cause an increase of greenhouse gas N2O production as N2O reductase is sensitive to pH and less active at pH < 7 (Richardson et al., 2009). A recent study suggests that low pH interferes with the N2O reductase assembly, putatively revealing a molecular mechanism of the acidification effect on N2O dynamics (Liu et al., 2014). Acidification may increase N2O production by aquatic ammoniaoxidizers, as well (Frame et al., 2017). Seawater N2O production is further compounded by many macroscopic processes of the ocean, such as vertical mixing and upwelling (Tseng et al., 2016). In summary, ocean acidification may cause altered (likely decreased) nitrification, increased N2O emission and increased loss of nitrogenous nutrients via enhanced denitrification and anammox, among many other significantly altered biogeochemical processes in the ocean. It is necessary to develop a mechanistic and quantitative understanding of the various marine nitrogen-cycling processes and the ocean acidification effects, which are fundamentally important for better modeling and predication of the behavior and function of the future marine nitrogen and carbon cycles.

## Chemolithoautotrophic Responses to Ocean Eutrophication and Deoxygenation

The trophic and oxygenation states are two critical and interconnected factors that have significant influences on the ecological processes and biogeochemical functions of the ocean's ecosystems. Anthropogenic eutrophication and upwelling are the major contributors to coastal hypoxia and anoxia, which may be additionally enhanced by incoming offshore seawater in certain marginal seas (Lui et al., 2014). Coastal eutrophication is predicted to increase, due to continuing increase of human activities (Doney, 2010; Nogales et al., 2011; Statham, 2012; Bijma et al., 2013). Under the impact of ongoing global change, the duration and intensity of most of the large-scale upwelling systems are predicted to increase as well (Sydeman et al., 2014; Wang D. et al., 2015). Global warming exacerbates the frequency, extent and impacts of coastal "dead zones," which usually occur during warm seasons. The oceanic OMZs will also intensify and expand due to warming-induced oxygen solubility reduction and seawater stratification (Keeling et al., 2010; Wright et al., 2012). Therefore, oceanic OMZs and coastal hypoxia and anoxia will undoubtedly exacerbate in the future. These situations may alter the major metabolic pathways and functional services of the affected ecosystems.

Hypoxic and anoxic environments may facilitate microbial nitrogen fixation (Zhou et al., 2016). Under O2-rich environments, bacterial nitrogen acquisition through nitrogen fixation is eco-energetically less favorable than assimilatory nitrate uptake (Eichner et al., 2014; Jiang et al., 2015). However, in hypoxic environments such as the marine OMZs, nitrogen fixation may be eco-energetically more favorable and less inhibited by high nitrate concentrations (Großkopf and Laroche, 2012). Because most of the diazotrophs in the aphotic marine hypoxic and anoxic environments are heterotrophs, the availability of metabolizable organic substrates as energy sources is an important factor influencing the abundance and activity of the in situ diazotrophs. Alternatively, chemolithoautotrophic ANME-2c archaea may contribute substantially to nitrogen fixation in methane-rich environments (Pernthaler et al., 2008; Dang et al., 2009a; Dekas et al., 2009; Miyazaki et al., 2009). Furthermore, the diazotrophic abundance and activity may also be controlled by the availability of phosphate and/or iron, which may vary in different marine environments (Moore et al., 2013; Dang and Lovell, 2016). Diazotrophic microorganisms and activity have been confirmed in studied marine hypoxic and anoxic environments (Fernandez et al., 2011; Hamersley et al., 2011; Jayakumar et al., 2012; Loescher et al., 2014). The study of non-cyanobacterial diazotrophic contribution to the ocean's reactive nitrogen pool constitutes a new research paradigm of marine nitrogen cycling (Bombar et al., 2016). Breakthroughs in this field are indubitably instrumental to developing better understandings of the marine carbon cycle and its interactions with global change.

Although nitrogen fixation may be enhanced in hypoxic and suboxic seawater, loss of fixed nitrogen via denitrification and anammox is more eco-energetically favorable and thus nitrogen loss is the major microbial process in these environments (Lam and Kuypers, 2011). It has been estimated that oceanic OMZ seawater accounts for one third or more of fixed nitrogen loss on a global scale (Canfield et al., 2010b). Under hypoxic and anoxic conditions, enhanced availability of reduced inorganic chemicals such as ammonia/ammonium and sulfide as energy sources facilitates microbial carbon fixation that is coupled to nitrogen and sulfur cycling processes, including chemolithoautotrophic ammonia oxidation, nitrite oxidation, sulfur oxidation, and anammox. Some of these microbial processes are intensified particularly at or near the oxic-anoxic interfaces in the water columns (Füssel et al., 2012; Capone and Hutchins, 2013). These processes may help to restore the nitrogen balance by removal of excess nitrogen originated from riverine and terrestrial inputs in eutrophic estuarine and coastal waters. However, they may aggravate the scarcity of nitrogenous nutrients and exert further limitation on the ocean's capacity of primary productivity and carbon sequestration in oceanic waters. A recent study has found that AOA and NOB in OMZs have high affinities for oxygen and nitrification (even at 5 nM O2) may control fixed nitrogen loss that is subsequently performed by denitrification and anammox (Bristow et al., 2016).

The scarcity of fixed nitrogen limits primary production and BP-mediated carbon sequestration in vast regions of the ocean (Moore et al., 2013). To make the situation even worse, the current fixed nitrogen pool of the ocean may be unbalanced and dwindling, caused by nitrogen loss via denitrification and anammox being significantly greater than nitrogen gain via nitrogen fixation (Codispoti et al., 2001; Mahaffey et al., 2005; Codispoti, 2007). Although uncertainties and debates remain about the conundrum of this imbalance (Gruber and Galloway, 2008; Canfield et al., 2010a; Zehr and Kudela, 2011; Großkopf et al., 2012; Voss et al., 2013; Devol, 2015; Klawonn et al., 2015b; Zhou et al., 2016), enhanced nitrogen loss by denitrification and anammox as a result of aggravated hypoxia, anoxia and acidification in coastal seas and oceanic OMZs may indeed diminish the marine fixed nitrogen reservoir under the impacts of increasing anthropogenic activities and global warming. This diminishment may constitute a positive feedback mechanism that speeds up global change by further limiting the ocean's carbon sequestration capacity.

The marine hypoxic and anoxic environments are also the hotspots for the production of biogenic greenhouse gases such as N2O, CH4, and occasional H2S (Naqvi et al., 2010; Wright et al., 2012; Capone and Hutchins, 2013; Carolan et al., 2015; Murray et al., 2015; Kock et al., 2016). N2O is produced by many microbial processes and it is also the precursor of NO radicals that cause ozone destruction in the stratosphere (Carpenter et al., 2012; Schreiber et al., 2012; Voss et al., 2013; Mellbye et al., 2016). Hypoxic and anoxic environments that are rich in labile organic matter facilitate heterotrophic denitrification and the production of N2O as a metabolic intermediate (Gilly et al., 2013; Townsend-Small et al., 2014; Babbin et al., 2015; Castro-González and Farías, 2015).

Autotrophic denitrification that couples denitrification with anaerobic chemolithoautotrophic sulfide oxidation also contributes to fixed nitrogen removal and N2O production in these environments (Shao et al., 2010; Ulloa et al., 2012). Anammox bacteria may produce N2O as well, via enzymatic NO detoxification (Kartal et al., 2007). N2O is also produced by AOA and AOB, though the detailed mechanisms involved in these two groups of aerobic microorganisms may be quite different (Kozlowski et al., 2016). Hypoxic conditions strongly stimulate AOB N2O production via the enzymatic nitrifier denitrification process, while N2O production in AOA may result from abiotic reactions (Codispoti, 2010; Zhu et al., 2013; Kozlowski et al., 2016).

Submicromolar O<sup>2</sup> has been found to reversibly suppress anammox and denitrification, likely at the enzymatic level (Dalsgaard et al., 2014). Due to intracellular anammoxosomes, anammox bacteria may be more resistant to O<sup>2</sup> suppression than denitrifying bacteria (Dalsgaard et al., 2014). Relative to anammox, denitrification is likely to be a more important N2O production process in hypoxic environments. It has also been found that sulfide, which accumulates in extremely anoxic environments or exists cryptically under hypoxic conditions (Canfield et al., 2010b; Glaubitz et al., 2013), strongly stimulates denitrifying N2O production without affecting the anammox process (Dalsgaard et al., 2014). Hypoxic and anoxic conditions are usually accompanied by environmental acidification, which may cause increased denitrifier N2O production (see above "Chemolithoautotrophic Responses to Ocean Acidification" section). Marine hypoxic and anoxic conditions influence the speciation and abundance of many trace elements and heavy metals, which may also influence the microbial production of greenhouse gasses via influencing the synthesis and activity of the involved enzymes (Glass and Orphan, 2012). Furthermore, many denitrifiers harbor truncated denitrifying pathways lacking the gene of N2O reductase for reducing N2O to N2, contributing to N2O production and accumulation in the environments (Jones et al., 2008; Richardson et al., 2009; Graf et al., 2014). Although shorter pathways of energy metabolism lowers ATP yield, they increase the ATP production rate and thus the maximal growth rate of the denitrifiers (Costa et al., 2006; Jones et al., 2008; Simon and Klotz, 2013), providing an eco-energetic advantage in environments rich in labile organic matter and nitrate.

The diverse microbial N2O production pathways and environmental controlling factors dictate the dynamics of the marine N2O reservoir. With the increasing expansion of both the oceanic OMZs and coastal hypoxic and anoxic water bodies, it is reasonable to predict that the microbial processes in these environments may contribute more to the production of N2O. This may constitute a positive feedback on global change. Currently, the relative contributions of the different microbial N2O production processes to the marine N2O reservoir and dynamics is still not reliably solved, particularly at the regional and global scales. This situation may be tackled by future investigations and modeling, in which the incorporation of the microbial eco-energetic constraints may be helpful.

## Chemolithoautotrophic Responses to Environmental Trace Element Variations and Heavy Metal Pollutions

Ecosystem energy flow interconnects with the interweaved biogeochemical cycles of carbon, nitrogen, phosphorus, sulfur, and many other elements such as biological trace metals (Falkowski et al., 2008; Jelen et al., 2016). Many of the microbial energetic processes are catalyzed by metalloenzymes (Nitschke et al., 2013). Although exergonic bioenergetic reactions are thermodynamically favorable, they are usually hindered kinetically by high activation barriers and need enzymes to speed up (Nitschke et al., 2013). Activities of metalloenzymes rely on various redox-active metal cofactors. Due to their plasticity in adopting different oxidation states and coordination environments in diverse enzyme molecules, redox-sensitive transition metals such as Fe, Ni, Cu, Zn, Co, Mo, W, V, and Mn are the key elements that constitute the metal cofactors in metalloenzymes (Klotz and Stein, 2008; Nitschke et al., 2013; Simon and Klotz, 2013; Gómez-Consarnau and Sañudo-Wilhelmy, 2015). The valence, speciation, solubility, adsorption, organic complexation and rates of redox processes of transition metals are subject to influences by physicochemical conditions such as environmental pH, O<sup>2</sup> content and redox potentials (Byrne et al., 1988; Banks et al., 2012; Sunda, 2012; Scholz et al., 2014). The properties of transition metals in marine environments are influenced by diverse biological factors as well (Morel and Price, 2003; Gerringa et al., 2016). The recently proposed "Ferrojan Horse Hypothesis" highlights a newly discovered viral mechanism for the behaviors of Fe in marine environments (Bonnain et al., 2016). Under global change impacts such as those from ocean warming, acidification, and deoxygenation, the bioavailability and bioactivity of some of these transition metals may be altered (Hoffmann et al., 2012; Gledhill et al., 2015; Emerson, 2016; Stockdale et al., 2016). This may influence the synthesis and activity of certain key metalloenzymes that are involved in microbial energy metabolism, leading to changes of the composition and activity of the marine microbiota and further changes of the marine environment and its functions. For example, AOB employ the Fe-based electron transfer system for ammonia oxidationmediated bioenergetic process, while AOA employ the Cu-based electron transfer system for ammonia oxidation-mediated energy metabolism (Walker et al., 2010; Santoro et al., 2015). Vast areas of marine environments are Fe-limited, particularly in the open oceans. The reliance on Cu other than Fe may provide AOA an eco-energetic advantage over AOB and contribute to the dominance of AOA in many marine environments (Walker et al., 2010). However, seawater Cu and Fe concentrations and speciation are subject to variation. The scarcity of bioavailable Cu in certain marine environments may impose a limitation on AOA abundance and activity (Jacquot et al., 2014; Shiozaki et al., 2016). There exist steep concentration gradients of Cu and Fe in marine OMZs, where suboxic and anoxic conditions decrease dissolved Cu concentrations but increase dissolved Fe concentrations and these trace metal profiles concur with the in situ Cu metalloenzyme gene profile of ammonia-oxidizing

Thaumarchaeota and the Fe metalloenzyme gene profile of anammox Planctomycetes, respectively (Glass et al., 2015). The temporospatial distribution and dynamics of transition metals may be an important factor determining the temporospatial distribution and dynamics of the various microbial energetic pathways and functions in the ocean. Researches on this aspect are rare and certainly need to be strengthened.

Most of the microbial metalloenzymes are sensitive to the inhibitory effects of heavy metals. Genes encoding heavy metal resistance are much more abundant in the genome of Ca. Nitrososphaera gargensis isolated from a heavy metal-containing hot spring than in AOA isolated from marine environments (Spang et al., 2012). This indicates that AOA need specific genetic and biochemical mechanisms for heavy metal resistance and marine AOA may be sensitive to the inhibitory effects of heavy metals. Ongoing marine environmental changes caused by ocean warming, acidification, deoxygenation, eutrophication, and pollution may change the concentration, speciation, solubility and mobility of heavy metals as well, particularly in estuarine and coastal environments where heavy metal contaminations are usually prevalent (Atkinson et al., 2007; Millero et al., 2009; Gao et al., 2014; Zeng et al., 2015). The global change problem may be worsened by the compounding effects of heavy metals on the marine microbial processes and functions. The complex interactions between the energy metabolic processes and element biogeochemical cycles indicate that the eco-energetics of the marine microbiota need to be studied with a multidisciplinary effort (Klotz, 2010).

#### FUTURE PERSPECTIVES ON MARINE CHEMOLITHOAUTOTROPHIC ECO-ENERGETIC RESEARCH

Microbial metabolism is driven by thermodynamic favorability, which is determined by the availability of free energy in the involved biochemical reactions. Energy is therefore an important constraint, along with nutrients, on the physiology of any organism and the structure and function of any ecosystem. For example, low light availability exerts an energy limitation on the photosynthetic productivity though nutrients are abundant in eutrophic estuaries with high seawater turbidity (Dang and Jiao, 2014). On the contrary, the lack of reduced inorganic chemicals such as NH<sup>4</sup> <sup>+</sup>/NH<sup>3</sup> and H2S as energy substrates exerts a limitation on the chemolithoautotrophic primary production in the oxic deep oceans where nutrients such as nitrate and phosphate are usually available.

Environmental physicochemical conditions may exert important constraints on Gibbs energy yields and activity rates of the marine microbiota. For example, it has recently been found that the temperature-pH-salinity extremes exert a much stronger effect on the growth of anaerobically respiring and fermentative bacterial and archaeal strains than on the growth of aerobically respiring strains (Harrison et al., 2015). The difference in living parameter spaces between anaerobic microorganisms and aerobic microorganisms is likely related to their distinct eco-energetic properties. ATP yields of aerobically respiring microorganisms can be an order of magnitude higher than those of the anaerobically respiring or fermentative microorganisms, enabling better performances of aerobes in stress resistance, growth and activity over a broader range of physicochemical extremes (Harrison et al., 2015). This rule may be applicable to the chemolithoautotrophic microorganisms as well. The eco-energetic differences between the aerobic and anaerobic chemolithoautotrophic microorganisms may also affect their respective performances and functions under physicochemical extremes. Although it has not yet been systematically investigated, this hypothesis is important for understanding some of the mechanisms that lead to the compositional and functional shift of the marine microbiota under the various global change impacts.

Ecosystem energy flow involves diverse metabolic pathways (usually harbored by different microorganisms) and their interactions at various temporal and spatial scales. A community perspective is needed for the understanding and study of ocean biogeochemistry and eco-energetics (**Figure 1**) (Strom, 2008). Many ecophysiological activities and biogeochemical functions of the marine microbiota are carried out through microbial interactions including both cooperation and competition (Litchman et al., 2015; Dang and Lovell, 2016). For example, marine AOA usually have very small cell sizes and genomes (Martens-Habbena et al., 2009; Bayer et al., 2016), providing certain eco-energetic advantages particularly in oligotrophic environments (Batut et al., 2014; Martínez-Cano et al., 2015). Due to genome reduction, some marine AOA have lost the genes that encode for the catalase-peroxidase proteins (Kim et al., 2016). This energetic economy necessitates the dependency of these AOA on other co-occurring microorganisms for oxidative damage protection. The dependence of genome-reduced microorganisms on other microorganisms facilitates the development of metabolic collaborations and other mutualistic interactions in microbial communities. In addition to producing phosphonates for sharing with other microorganisms as a phosphorus source (Metcalf et al., 2012), marine AOA also harbor the genes for synthesis of vitamin B<sup>12</sup> (Doxey et al., 2015), an essential cofactor required by many marine organisms (Gómez-Consarnau and Sañudo-Wilhelmy, 2015). Auxotrophy and physiological complementation, among many other microbial interactions, may help establish metabolic interconnectedness in natural microbial communities (Giovannoni et al., 2014; Garcia et al., 2015; Kouzuma et al., 2015). Microbial interactions are one of the key intrinsic properties of natural microbial communities that defines not only the composition and structure but also the activity and function of the communities (Hunt and Ward, 2015). Therefore, a systems ecology approach is needed for eco-energetic analyses of the marine ecosystems, in particular for a mechanistic understanding of the ecosystems' driving force, the energy flow along the electron transfer pathways and redox exchange-induced matter fluxes within the marine microbiomes (Kolber, 2007).

In the past, investigations based on matter metabolism and fluxes became the mainstream of marine ecology research, though energy metabolism plays an equally important role on

the structure and function of marine ecosystems. A few attempts have been made on eco-energetic analyses for certain marine environments (e.g., Amend et al., 2003; Akerman et al., 2011; Dahle et al., 2015; LaRowe and Amend, 2015; Bach, 2016). Most of these analyses focused on energylimited environments such as the subseafloor deep biosphere where only the maintenance energy of the studied microbial community needs to be considered. However, community growth-related temporospatial variation is common in many marine environments, which need more sophisticated and dynamic eco-energetic processes to be taken into consideration for the analyses (Vallino and Algar, 2016). With the advance of the "omics" approaches such as metagenomics, metatranscriptomics, and metaproteomics for community metabolic network analyses, a comprehensive understanding of the mechanisms, processes and environmental responses may be obtained about the functions of marine microbial communities (Morris et al., 2010; Bodrossy, 2015; Reed et al., 2015; Perez-Garcia et al., 2016). These may also help to understand the genetic, biochemical and physiological constraints on the coupling or uncoupling of metabolic processes among different microorganisms or functional groups in an environment. Just like people need two legs to walk, the combination of the "omics" techniques with in situ energy and matter flux measurements or calculations may help to develop advanced biogeochemical models for better understanding and prediction of the processes and functions

#### REFERENCES


of the marine ecosystems and their responses to the global change impacts (Kolber, 2007; Soh and Hatzimanikatis, 2010; Vallino and Algar, 2016).

#### AUTHOR CONTRIBUTIONS

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

#### FUNDING

This work was supported by the China SOA grant GASI-03- 01-02-05, National Key Research and Development Program of China grant 2016YFA0601303, China MOST 973 program grant 2013CB955700, NSFC grants 91328209, 41676122, and 91428308, and CNOOC grants CNOOC-KJ 125 FZDXM 00TJ 001-2014 and CNOOC-KJ 125 FZDXM 00ZJ 001-2014.

#### ACKNOWLEDGMENT

We thank professor Martin G. Klotz (City University of New York) for valuable discussions and manuscript improvement and the two reviewers for their valuable comments.


thaumarchaeal strains indicates niche differentiation. ISME J. 10, 1051–1063. doi: 10.1038/ismej.2015.200


ocean acidification. PLoS ONE 9:e108153. doi: 10.1371/journal.pone. 0108153



potential chemolithotrophic reactions in geothermal springs. FEMS Microbiol. Ecol. 81, 446–454. doi: 10.1111/j.1574-6941.2012.01369.x




gene transfer from a Nitrococcus-like γ-proteobacterium. BMC Evol. Biol. 7:85. doi: 10.1186/1471-2148-7-85




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

# Seasonal Succession and Spatial Patterns of Synechococcus Microdiversity in a Salt Marsh Estuary Revealed through 16S rRNA Gene Oligotyping

#### Edited by:

*Martin G. Klotz, Washington State University Tri-Cities, United States*

#### Reviewed by:

*Ryan Paerl, University of Copenhagen, Denmark Lucas Stal, Royal Netherlands Institute for Sea Research, Netherlands*

#### \*Correspondence:

*Katherine R. M. Mackey kmackey@uci.edu*

#### † Present Address:

*Julie A. Huber, Marine Chemistry and Geochemistry Department, Woods Hole Oceanographic Institution, Woods Hole, MA, United States*

#### Specialty section:

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

Received: *29 January 2017* Accepted: *25 July 2017* Published: *09 August 2017*

#### Citation:

*Mackey KRM, Hunter-Cevera K, Britten GL, Murphy LG, Sogin ML and Huber JA (2017) Seasonal Succession and Spatial Patterns of Synechococcus Microdiversity in a Salt Marsh Estuary Revealed through 16S rRNA Gene Oligotyping. Front. Microbiol. 8:1496. doi: 10.3389/fmicb.2017.01496*

Katherine R. M. Mackey <sup>1</sup> \*, Kristen Hunter-Cevera<sup>2</sup> , Gregory L. Britten<sup>1</sup> , Leslie G. Murphy <sup>2</sup> , Mitchell L. Sogin<sup>2</sup> and Julie A. Huber 2 †

*<sup>1</sup> Earth System Science, University of California Irvine, Irvine, CA, United States, <sup>2</sup> Marine Biological Laboratory, Josephine Bay Paul Center for Comparative Molecular Biology and Evolution, Woods Hole, MA, United States*

*Synechococcus* are ubiquitous and cosmopolitan cyanobacteria that play important roles in global productivity and biogeochemical cycles. This study investigated the fine scale microdiversity, seasonal patterns, and spatial distributions of *Synechococcus* in estuarine waters of Little Sippewissett salt marsh (LSM) on Cape Cod, MA. The proportion of *Synechococcus* reads was higher in the summer than winter, and higher in coastal waters than within the estuary. Variations in the V4–V6 region of the bacterial 16S rRNA gene revealed 12 unique *Synechococcus* oligotypes. Two distinct communities emerged in early and late summer, each comprising a different set of statistically co-occurring *Synechococcus* oligotypes from different clades. The early summer community included clades I and IV, which correlated with lower temperature and higher dissolved oxygen levels. The late summer community included clades CB5, I, IV, and VI, which correlated with higher temperatures and higher salinity levels. Four rare oligotypes occurred in the late summer community, and their relative abundances more strongly correlated with high salinity than did other co-occurring oligotypes. The analysis revealed that multiple, closely related oligotypes comprised certain abundant clades (e.g., clade 1 in the early summer and clade CB5 in the late summer), but the correlations between these oligotypes varied from pair to pair, suggesting they had slightly different niches despite being closely related at the clade level. Lack of tidal water exchange between sampling stations gave rise to a unique oligotype not abundant at other locations in the estuary, suggesting physical isolation plays a role in generating additional microdiversity within the community. Together, these results contribute to our understanding of the environmental and ecological factors that influence patterns of *Synechococcus* microbial community composition over space and time in salt marsh estuarine waters.

Keywords: oligotyping, Synechococcus, salt marsh estuary, succession, temperature, salinity

## INTRODUCTION

The genus Synechococcus is a remarkably diverse group that inhabits aquatic environments ranging from open ocean and coastal waters (Bouman et al., 2006; Zwirglmaier et al., 2007, 2008; Post et al., 2011; Flombaum et al., 2013; Sohm et al., 2015; Hunter-Cevera et al., 2016) to brackish estuaries (Chen et al., 2006; Choi and Noh, 2009), and freshwater lakes (Ivanikova et al., 2008). Despite occupying this extensive range of environments, little is known about the patterns of abundance, diversity, and distribution of Synechococcus in estuary waters relative to coastal waters and the open ocean. Estuaries are dynamic environments where freshwater and seawater mix, generating hydrographic gradients that vary spatially and over tidal and seasonal cycles. These habitats therefore have distinctive gradients, making them attractive sites to investigate Synechococcus community dynamics.

The diversity of Synechococcus based on 16S rRNA gene phylogeny spans three sub-clusters (5.1, 5.2, and 5.3), and further resolution of the genus is possible using additional genetic markers such as the internally transcribed spacer (ITS) region and the genes cpeA, narB, ntcA, petB, rbcL, and rpoC1 (Scanlan et al., 2009; Ahlgren and Rocap, 2012). These markers partition the genus into at least 20 distinct clades (Scanlan et al., 2009; Ahlgren and Rocap, 2012; Huang et al., 2012; Mazard et al., 2012; Sohm et al., 2015). The majority of marine strains fall within sub-cluster 5.1. Efforts to link clade identity to ecological or physiological characteristics suggest that Synechococcus has diversified based on temperature (Mackey et al., 2013; Pittera et al., 2014), nutrient type and availability (Glover et al., 1988; Moore et al., 2002; Saito et al., 2005; Palenik et al., 2006; Mackey et al., 2015; Sohm et al., 2015), and light intensity and spectrum (Palenik, 2001; Everroad et al., 2006; Mackey et al., 2017). For example, clade III is found in warm, oligotrophic waters, while clades I and IV tend to dominate colder, nutrient rich waters, and clade II is more broadly distributed throughout the tropics and subtropics (Mazard et al., 2012; Sohm et al., 2015). However, niches occupied by Synechococcus strains are generally more flexible and have more overlap compared to ecotypes of the closely related genus Prochlorococcus. This may be due to extensive lateral gene transfer or independent parallel evolution of traits that occurred after the marine Synechococcus lineages diverged (Ahlgren and Rocap, 2012; Sohm et al., 2015).

A common observation within Synechococcus populations is the co-occurrence of different clades. Co-occurrence patterns appear to be particularly dynamic in space and time in estuaries. A comparison of Synechococcus populations from estuarine and coastal waters in the Pearl River Estuary in Hong Kong showed that clades II and IX dominated both sites in the winter, whereas in summer the coastal population comprised clades II and VI, and the estuarine population comprised co-occurring marine and freshwater taxa (Xia et al., 2015). This shift was attributed to warmer temperatures and freshwater input during the summer monsoon. In contrast, freshwater taxa were rare in the Chesapeake Bay estuary even at low salinity sites, and populations were instead dominated by co-occurring members of clades CB4 an CB5 (Chen et al., 2006). Coastal waters also show co-occurrence patterns. In the California Current, clades II and III co-occurred in the period leading up to the spring bloom, and clades I and IV dominated during the bloom itself (Tai and Palenik, 2009). In the Red Sea, co-existence of clades V, VI, and X was observed during transitional periods between mixing and stratification (Post et al., 2011). Very little is known about the factors that support coexistence among these closely related organisms. The greater genetic diversity of the aggregate Synechococcus population may effectively expand the overall niche filled by this genus by out-competing other genera. In some cases, less abundant Synechococcus clades may persist at low, but stable, levels by avoiding top down pressures like grazing and viral infection, both of which are more impactful when cell numbers are higher (Pedrós-Alió, 2006; Sogin et al., 2006; Fuhrman, 2009). Low-abundance strains can influence biogeochemical processes (Gilbert et al., 2012) and may serve as important reservoirs of genetic diversity (Sogin et al., 2006). The mechanisms that give rise to coexistence of high- and lowabundance strains is not well-understood, although their cooccurrence patterns are routinely observed (Ahlgren and Rocap, 2012; Sohm et al., 2015). Methods that examine small-scale diversity among co-occurring strains can provide a window into the factors that influence dynamics of rare as well as closely related but distinct populations.

To explore the patterns of estuarine Synechococcus diversity in space and time, we used an oligotyping approach to examine populations in estuarine waters of the Little Sippewissett salt marsh located in Cape Cod, MA. Oligotyping is a method that differentiates between closely related members of a microbial community (Eren et al., 2013) and enables fine scale diversity within microbial communities to be identified. The technique uses entropy analysis of variable sites in related sequences identified via taxonomic classification or clustering of high throughput sequencing data (Nicholson et al., 1987; Eren et al., 2013; Wilbanks et al., 2014). We hypothesized that oligotypes would report seasonal community dynamics and co-occurrence patterns as has been done in previous studies at the clade (Chen et al., 2006; Choi and Noh, 2009) and subclade (Tai and Palenik, 2009; Paerl et al., 2011; Robidart et al., 2012) levels. We identified 12 Synechococcus oligotypes, show that their relative abundances vary seasonally and spatially within the estuary, and explore their patterns of co-occurrence and succession in light of their fine-scale diversity.

#### MATERIALS AND METHODS

## Site Description and Sample Collection

The estuary of Little Sippewissett salt marsh is located on the southwestern tip of Cape Cod in the town of Falmouth, MA (41.576◦N, 70.636◦W). The estuary empties into Buzzards Bay on the western side of the Cape. Samples were collected and processed as described in Eren et al. (2013). Briefly, surface water was collected during low tide from seven stations within the marsh and nearby coastal waters (**Figure 1**). Two stations were in coastal waters (1 and 7), four were within the main channel of the estuary (2, 3, 4, and 5) and one was in a spatially segregated shallow site within the marsh (station 6). Depending on the

tidal cycle and amplitude, station 6 at times only had several centimeters of water depth. Sampling was done weekly from May 31, 2007 until September 4, 2007, and then monthly until September 2008.

One-liter water samples were stored on ice until filtration onto 0.22 µm pore size Sterivex filter cartridges. Dissolved oxygen and temperature were measured for each sample with a hand-held probe (Yellow Spring Instrument, YSI), and daily precipitation values were obtained from the Global Historical Climatology Network (GHCN) online database (https://climexp.knmi.nl/) of rain gauge measurements in Vineyard Haven, MA (41.39◦N, 70.61◦W) ∼8 miles offshore of Sippewissett Marsh. DNA was extracted and purified using a modified salt precipitation method as described previously (Sinigalliano et al., 2007; Eren et al., 2013). Bacterial 16S rRNA gene amplicons spanning the V4 through V6 regions were amplified using fusion primers, sequenced from the V6 end on a Roche GS-FLX 454 instrument using Titanium protocols, and quality-filtered and trimmed as described previously (Filkins et al., 2012). Raw sequences are deposited in the NCBI Sequence Read Archive (SRP017571: PRJNA183456).

#### Oligotyping

Within-genus diversity of Synechococcus was resolved using the oligotyping procedure as described previously (Eren et al., 2013), following recommended procedures described at http://merenlab.org/2013/11/04/oligotyping-best-practices/. Oligotyping is a computational method that identifies fine-scale diversity based on subtle variations in 16S RNA gene sequences. Oligotyping improves the resolution with which diversity can be identified because oligotype identification is not based on the availability of existing sequences within reference databases (as in taxonomic classification), and resolution of fine-scale diversity is not restricted by the choice of similarity threshold value (as in cluster analysis). The technique uncovers microdiversity within a given taxon or operational taxonomic unit (OTU), and is particularly useful when applied to taxonomic assignments or OTUs that occur throughout a dataset and may respond to changing environmental conditions (Eren et al., 2013).

Sequences identified as Synechococcus based on GAST taxonomic assignments were aligned with PyNAST aligner (version 0.1) against a GreenGenes OTU alignment template (version 6 Oct2010). Of the 13,427 Synechococcus sequences, only one failed to align. Mean length of Synechococcus reads was 482 base pairs with a standard deviation of 1.74 base pairs. Uninformative gap regions were removed and the entropy of each nucleotide position was calculated within the oligotype package. Nucleotide positions used to define oligotypes were selected from an iterative process; positions were included if they contributed to converged entropy within an oligotype and excluded if they contained redundant information of another position or did not contribute to convergence. A total of 19 positions were used to define oligotypes, and each oligotype was required to have a minimum substantive abundance of 20 ("M" parameter), such that an oligotype was not included if the most common sequence for that type occurred <20 times. Oligotypes were not required to comprise a certain percentage of reads, represent a minimum number of reads in a sample, or appear in more than one sample. Twelve oligotypes were identified and represented ∼80% of the total Synechococcus reads. Reads that were not identified as belonging to an oligotype are collectively referred to as "other Synechococcus" in the discussion below.

The sequence read counts for oligotypes presented here are normalized to either total microbial read counts or total Synechococcus read counts. As such, these data show changes in relative abundance, and do not necessarily reflect actual changes in cell concentrations of a given oligotype. For example, an increase in the relative abundance of a given oligotype could be a result of a true increase in the cell concentration of that oligotype, but could also result if the concentration of another oligotype decreases. Therefore, the data presented here indicates whether an oligotype becomes a larger or smaller fraction of the community, but do not necessarily indicate increases or decreases in absolute cell abundance.

#### Clade Identification

For each representative oligotype sequence, we inferred a clade designation by matching the representative V4–V6 sequences for each oligotype to a reference database of 16S rRNA gene sequences from cultured Synechococcus. The database consisted of sequences for which unambiguous clade assignments have been determined with a higher resolution diversity marker (i.e., ITS, petB, rpoC1, ntcA). All sequences were downloaded from Genbank and clade classifications were obtained from the following sources (Fuller et al., 2003; Choi and Noh, 2009; Mazard et al., 2011, 2012; Ahlgren and Rocap, 2012; Hunter-Cevera et al., 2016). Representative V4–V6 sequences for each oligotype and database sequences were aligned with Clustal W (embedded within BioEdit, version 7.2.0, Hall 1999), and exact matches between oligotype and database sequences were identified. When no direct match was found, the closest sequence was used to infer a clade designation.

#### Statistical Analyses

To statistically group oligotypes according to their cooccurrence, we computed the principal components (PCs) with respect to a sample matrix of Synechococcus oligotype reads normalized to total Synechococcus reads for that sample. We then projected each oligotype sample vector onto the first two PCs of the matrix (Anderson, 2003). The relative strength of oligotype groupings was then quantified as the Euclidean distance of each group's centroid to the centroid of other groups in the space of the first two PCs. To investigate environmental correlates of oligotype grouping, we computed the multiple regression of each of the first two PCs against the following environmental variables: (i) water temperature, (ii) salinity, (iii) dissolved oxygen and (iv) weekly precipitation. The most important set of environmental correlates for each PC was defined as the set that minimizes the Bayesian Information Criterion (BIC; Kass and Raftery, 1995) for all possible multiple regressions with respect to the four environmental variables.

#### RESULTS

#### Synechococcus Oligotype Patterns

A total of 3,142,957 PCR amplicon sequence reads were generated from microbial populations of 189 samples collected at seven stations (**Figure 1**) on 27 different dates throughout 2007 and 2008. Of these sequences, 13,358 reads were classified as Synechococcus. The number of Synechococcus reads relative to the total microbial population varied spatially and seasonally, ranging from 0% on some winter sampling dates/locations up to 5.9% at coastal station 7 during the summer (**Figure 2**).

We identified 12 Synechococcus oligotypes, which together accounted for an average of 73% of the Synechococcus reads over the entire dataset and comprised 0–100% of the total Synechococcus population depending on sampling location and date. These oligotypes were affiliated with several different Synechococcus clades based on sequence similarity of the representative V4–V6 sequence for each oligotype to known cultured isolates, and each of the 12 representative V4–V6 sequence for each oligotype mapped perfectly or within several base pairs to a range of known Synechococcus clades (**Figure 3**; **Table 1**). The 12 oligotypes shared >95% V4-V6 sequence similarity (**Table 2**). **Figure 3** shows the placement of these 12 oligotypes within the known Synechococcus phylogeny. The remaining reads not assigned to an oligotype were enumerated as "other Synechococcus" in our analysis. Clear seasonal and spatial patterns were apparent; Synechococcus comprised a greater fraction of the microbial population in the summer than in the winter, and had higher relative abundances in the coastal ocean compared to stations within the estuary (**Figure 4**).

Within the weekly summer samples, succession of oligotypes occurred between early and late summer (**Figure 5**). In the early summer beginning in June, oligotype O2 dominated the Synechococcus community. Oligotypes O4 and O5 also reached maximum values early in the summer, although at relative abundances below oligotype O2. Toward the end of the summer the Synechococcus community consisted of oligotypes O1, O3, O6, O7, O8, O10, and O11. The seasonal succession was apparent at each station, though oligotype 9 was only strongly represented in the shallow station 6 where it comprised a larger segment of the

FIGURE 3 | Phylogenetic tree constructed from 16S rRNA sequences (∼1,200 bp) of *Synechococcus* clade representatives and representative V4–V6 sequences for each oligotype. Reconstruction was performed with the ARB software package (version 5.3, Ludwig et al., 2004) with a maximum likelihood approach using RAxML and a GTR GAMMA rate substitution model. Due to their shorter sequence length, oligotype sequences (indicated in red) were added to this base tree using the ARB parsimony method. Bootstrap analysis for support of tree branches was also carried out in ARB with rapid bootstrap analysis and 500 sample trees. Values for major branches appear in light gray.

overall Synechococcus community (up to 100% of Synechococcus reads) than at other sites. Oligotypes 6, 8, 10, 11, and 12 were rare overall, and their relative abundances were not closely linked to any one station throughout the year (Figure S1).

TABLE 1 | Clade affiliations of the 12 V4–V6 representative sequences for each *Synechococcus* oligotype.


Different patterns were observed between co-occurring oligotypes in the early and late summer communities. Oligotypes that comprised the majority of the early summer community tended to be well-correlated with each other (Figure S2). Relative abundance of these types showed consistent ratios: the relative abundance of O2 (clade I) was usually two-fold >O4 (clade I) and four-fold >O5 (clade IV). In contrast, correlations among oligotypes that comprised the late summer community were more variable. Strong correlations were observed among the most abundant oligotypes (O1 and O3, R2 = 0.87, O1 and 06, R2 = 0.96 and O3 and O6 = 0.83). Lower abundant oligotypes did not show strong correlations, despite belonging to the same clades (i.e., CB5).

#### Environmental Variables

Salinity values along the main channel of the estuary remained above 15 psu with the exception of station 4, the location furthest upstream toward the freshwater end member (**Figure 2D**). Station 4 also showed the greatest seasonal variation in salinity, with fresher waters occurring during the winter months. Temperature ranged from ∼4 ◦C to >35◦C seasonally, with the highest temperatures observed in the shallow station 6 during summer months (**Figure 2C**). Dissolved oxygen varied among stations, with the highest concentrations observed in the coastal stations 1 and 7, and the lowest values observed in the freshwater end member station 4 (**Figure 2E**).

From principle component analysis (PCA) of oligotype relative abundances, we found that physical and chemical conditions in the estuary correlated with the patterns of oligotype co-occurrence (**Figure 6**). The first principle component captured 46% of the variance in oligotype relative abundance and discriminated oligotypes according to dissolved oxygen and temperature based on the multiple regression model selection using BIC. The second principle component explained an additional 21% of the variation and discriminated oligotypes according to salinity alone.

The statistical relationships among the dominant PCs and the environmental variables were consistent with the composition

TABLE 2 | Percent sequence similarity between V4 and V6 representative sequences for each oligotype.


shift observed during the summer (**Figure 5**). Specifically, oligotypes 2, 4, and 5 correlated with lower temperatures and higher dissolved oxygen content based on the PCA. The shift in late summer to oligotypes 1, 3, and 7 correlated with higher temperatures. A third group of co-occurring late summer oligotypes (the comparatively lower relative abundance "rare" oligotypes 6, 8, 10, and 11) correlated with both PC1 and PC2. Oligotype 12 was not affiliated with either principle component or site, but became more abundant in the fall after the dominance of oligotypes 1, 3, and 7 in late summer had begun to dissipate. Oligotype 9 was also not strongly correlated with PC1 or PC1.

#### DISCUSSION

#### Temporal Diversity Patterns

Synechococcus oligotypes comprised a larger proportion of the microbial community in the spring and summer months of June through September compared to winter months (**Figure 4**). Two distinct Synechococcus communities emerged that dominated in the early summer (May/June) and late summer (July onward), each composed of different, co-occurring Synechococcus oligotypes (**Figure 5**).

The Synechococcus community in early summer consisted of clade I oligotypes O2 and O4, and clade IV oligotype O5. Clades I and IV commonly co-occur in coastal waters, such as the cool waters off the coast of California (Tai and Palenik, 2009), as well as the open ocean at higher latitudes (Zwirglmaier et al., 2007, 2008). Tai and Palenik observed that while clade IV was typically more abundant than clade I in the open ocean, the relative dominance of the two clades fluctuated along the California coast, with clade I dominating for part of the bloom cycle (Tai and Palenik, 2009). In contrast, in this study, clade I outnumbered clade IV throughout the summer months. Similar patterns were observed at the nearby coastal water Martha's

Vineyard Coastal Observatory (MVCO), located southeast of Sippewissett Marsh, where the spring bloom of Synechococcus was mainly comprised of clade I representatives (Hunter-Cevera et al., 2016), and is correlated with warming water temperatures in the spring.

In contrast to the early summer community, the late summer community correlated with higher temperature and, for some oligotypes, higher salinity levels (**Figure 6**). This late summer community included members of clades CB5, I, IV, and VI. The greater diversity of oligotypes in the late summer relative to early summer is consistent with the late summer communities that occur offshore; at the MVCO site, oligotypes comprising the late summer community were from clades other than clade I, but could not be completely resolved (included matches to CB5, CRD1, II, III, IV, VII, and WPC2; Hunter-Cevera, 2014).

Clade CB5 that dominated in the late summer in Little Sippewisset Marsh is an estuarine clade that was first identified and described in samples from the Chesapeake Bay (Chen et al., 2006). The observation that this clade emerged strongly under warmer late summer conditions but was not prominent in the cooler early summer period could indicate a preference for warmer water temperatures. While the late summer community shift is correlated with warming temperatures overall (**Figure 6**), this shift in diversity nevertheless lagged the seasonal warming in that the increase in relative abundance was most pronounced only after temperatures exceeded ∼15◦C (**Figure 7**). Similar patterns were observed in LSM for Pelagibacter, where the population transitioned from being dominated by a polar oligotype to a tropical oligotype between June and July, and this transition also lagged the seasonal temperature increase (Eren et al., 2013). Other factors not considered in this study, such as nutrient availability, irradiance, viral lysis, and grazing could also influence the observed seasonal dynamics.

The influence of water temperature on dissolved oxygen levels likely explains their anti-correlation (with respect to PC1), and hence greater retention of photosynthetically generated oxygen in the cooler early summer waters compared to warmer late summer temperatures.

FIGURE 6 | Principle component analysis of *Synechococcus* oligotype relative abundance. T, temperature; P, precipitation; S, salinity; DO, dissolved oxygen.

#### Spatial Diversity Patterns

Coastal marine waters strongly influenced the physical, chemical, and biological characteristics of Little Sippewissett Marsh waters. The salinity and temperature patterns within the main channel of the estuary closely reflected the seawater end member, as the ocean source is larger than the freshwater source in this estuary. The cyanobacterial community composition along the main channel transect likewise reflected that of the marine source water. This pattern suggests that cyanobacterial populations within the main channel of the estuary were transported from the coastal waters, rather than from the freshwater source, and were either diluted, grazed, or declined due to stress within the estuary.

Greater spatial diversity in Synechococcus community structure has been observed in the Chesapeake Bay (Chen et al., 2006) and Pearl River Estuaries (Xia et al., 2015), where communities in coastal and estuarine waters differ considerably. The relatively smaller size of Little Sippewissett Marsh compared to these larger estuaries is likely the reason that Synechococcus population structure is more conserved throughout this estuary. While certain characteristics of the Synechococcus community are similar to other estuaries, such as having higher diversity and relative abundance in the summer (Xia et al., 2015), spatial differentiation within an estuary may be more sensitive to site characteristics (channel length, residence time, site geomorphology, etc.).

The influence of coastal seawater on the marsh was less pronounced at Station 6, the shallow water station that was more isolated from the main channel. This station supported a diverse population of oligotypes that did not resemble the community composition at the other sites (**Figure 5**). Certain oligotypes that were dominant in the seawater, such as oligotypes O1 and O2, were also present at station 6. However, their relative abundances were smaller compared to O9 on many of the sampling dates. Oligotype O9 is a member of clade 5.2, which are halotolerant Synechococcus more distantly related to marine Synechococcus that are commonly found in estuaries and coastal environments (Scanlan et al., 2009). The salinity and dissolved oxygen levels at station 6 were generally similar to other stations within the main channel, although temperature was up to 10◦C warmer (**Figure 2**). Variance in the relative abundance of O9 was not explained by temperature (PC1, **Figure 6**), despite the fact that higher temperature was a salient feature of the station. This suggests that other factors were more important is driving the relative abundance of O9.

Within the estuary, the community composition at a given site is governed by the net growth rate of each population and the residence time of the water, which determines how quickly populations are diluted by incoming seawater. Limited tidal influence may have enabled the population at station 6 to change relative to the main channel of the estuary by allowing growth to outpace dilution. Less frequent mixing with seawater may have provided sufficient time for the community at station 6 to diverge. Greater exposure to high irradiances, including UV light, in this shallow site would also serve to reshape the microbial community relative to other locations in the estuary and coastal ocean where mixing would mitigate exposure to high light.

These divergent populations could act as reservoirs of diversity for the nearby coastal ocean, particularly following rain events or king tides that would inundate more spatially isolated sites and wash members of the community into the coastal ocean. Indeed, certain rare Synechococcus strains have been brought into culture from the nearby MVCO, including phycocyanin-containing strains that lack phycoerythrin (such as sub-cluster 5.2; Hunter-Cevera, 2014). These strains are not abundant at MVCO based on flow cytometry measurements, yet they clearly persist in low numbers in coastal waters and thrive when brought into culture (Hunter-Cevera et al., 2016). Rare microbial populations have the potential to become dominant following local or global changes that favor their growth, hence although they represent only a small fraction of the overall community, their relative abundances could increase in time (Sogin et al., 2006). Spatially isolated sites within local estuaries, like station 6 from within LSM, may be a reservoir of these rare isolates for nearby coastal waters.

Spatial isolation of a site that is more remote from the tidal influence of seawater may also be an important first step in recruiting and enriching unique cyanobacterial communities within the estuary. Prior studies that have characterized microbial communities in LSM have shown that cyanobacteria other than Synechococcus (e.g., Lyngbya, Nostoc, Phormidium, and Oscillatoria) dominate the top layer of microbial mats (Nicholson et al., 1987). Microbial mats are complex communities where co-occurring populations have a high degree of biogeochemical interactivity. Areas like station 6, where unique communities emerge, may therefore also serve as important reservoirs for cyanobacterial diversity in the estuary.

## Using Oligotyping to Understand Interactions among Co-occurring Synechococcus

An important next step in understanding marine microbial community dynamics is to identify factors that drive the cooccurrence of closely related microbes within an environment, leading certain taxa to co-vary in space and time (Horner-Devine et al., 2007; Fuhrman, 2009). While this is already being done for Synechococcus at the sub-clade (Paerl et al., 2011; Robidart et al., 2012) and clade level (Choi and Noh, 2009; Tai and Palenik, 2009; Ahlgren and Rocap, 2012; Sohm et al., 2015), oligotyping allows diversity and succession patterns within microbial communities to be investigated with fine-scale resolution at the nucleotide level. This resolution makes it possible to investigate the underlying dynamics between Synechococcus oligotypes that could potentially underpin clade co-occurrence patterns observed in the environment (e.g., clades I and IV) by showing whether oligotypes within the same clade share similar population dynamics.

Our analysis revealed that oligotypes belonging to the same clade often showed similar relative abundance patterns, along with oligotypes from different clades. However, for early and late summer communities, the strength of correlation between oligotypes of the same or different clades varied. For the early summer community, the three most abundant oligotypes (O2, O4, and O5) followed a predictable pattern in which the relative abundances of each oligotype were more or less constant over time (Table S1, Figure S2). This stable proportionality occurred irrespective of whether the oligotypes were from the same clade (as for clade I oligotypes O2 and O4) or different clades (clade IV member O5).

In contrast, the late summer community showed more complex patterns of co-occurrence among oligotypes. The correlations between oligotypes belonging to the same clade were not always strong, and correlations between oligotypes in the CB5 clade from the late summer community provide a good example of this (Table S1; Figure S2). Many studies have shown that the relative proportions of co-occurring Synechococcus populations to each other at the clade and subclade level vary in space and time based on environmental factors like seasonal temperature fluctuations (Choi and Noh, 2009; Tai and Palenik, 2009; Xia et al., 2015; Hunter-Cevera et al., 2016), nutrient availability and upwelling (Choi and Noh, 2009; Robidart et al., 2012), circulation patterns (Choi and Noh, 2009; Paerl et al., 2011), and abundance of other phytoplankton (Robidart et al., 2012). This study shows that co-occurrence patterns are also evident at the oligotype level in Little Sippewissett Marsh, and that they change throughout the summer. Greater variability in oligotype co-occurrence behavior was observed in the late summer community compared to the early summer community. This could be due in part to the greater number and diversity of oligotypes comprising the late summer community compared to the early summer community, or the comparatively lower number of sequences for some oligotypes, which complicates correlation analysis.

Rare taxa are recognized as potential reservoirs of genetic diversity (Sogin et al., 2006), and episodic blooms of rare taxa can lead to significant shifts in the biogeochemical and ecological characteristics of the environment (Gilbert et al., 2012). Relative to the more abundant taxa within a microbial community, rare taxa can be closely related (e.g., belonging to the same population or pan-genome) or distantly related (e.g., belonging to different phyla). In both cases rare taxa can confer genetic diversity, but by different mechanisms. Distantly related rare tax may increase genetic diversity in the abundant populations by contributing to their genomes via lateral gene transfer. In contrast, the genomes of rare taxa may differ from closely related abundant taxa by virtue of gene mutation or horizontal gene transfer, which may enable rare taxa to outcompete abundant taxa when environmental conditions shift. Although microbial microdiversity is essential for maintaining proper ecosystem function, our understanding of the environmental factors that drive spatial and temporal changes in rare taxa is still in the early stages (Alonso-Sáez et al., 2015). The oligotyping approach applied here revealed that the late summer Synechococcus community included the four least abundant oligotypes, O6, O8, O10, and O11 (Figure S1). The factor that appears to set these rare oligotypes apart from other members of the late summer community is their correlation with higher salinity (PC2; **Figure 6**), and this was reflected in their distributions within the estuary, where they were never identified in samples from the freshwater source station 4 (Figure S1). It is not clear what factors served to constrain the relative abundances of these rare oligotypes at other sites with higher salinities, but it could stem from nutrient limitation, grazing, or viral susceptibility.

The rare members of the community did not follow any consistent pattern in terms of their correlations with each other or the more dominant oligotypes. Of the four oligotypes, O10 was generally the least correlated with other oligotypes in the late summer community (Table S1, Figure S2) despite being a member of clade CB5. Hence, despite being closely related at the clade level to many other co-occurring oligotypes in the community, this rare oligotype displayed different relative abundance patterns. This microdiversity may effectively permit clade CB5 to fill a slightly larger niche than if it contained fewer oligotype members.

We observe that the relative abundance of co-occurring Synechococcus is not constant over time; the timing with which each oligotype reaches its peak relative abundance during the late summer is slightly offset. In the late summer community, oligotype O7 shows this pattern, where its correlation with oligotypes O8, O10, and O11 shows a bifurcated pattern (Figure S2) that is driven by O7 reaching its peak relative abundance earlier in the season than the other oligotypes.

This study investigated the seasonal community dynamics of Synechococcus oligotypes in a small New England estuary. Patterns that have been observed at the clade and subclade levels, such as correlation between temperature and Synechococcus relative abundance, and the co-occurrence of groups from different clades, were shown to occur among oligotypes. This analysis still leaves open the question of whether cooccurring oligotypes affect each other in beneficial ways, as is common in microbial mats within Little Sippewissett Marsh. The phenomenon of co-occurrence appears to be a case of "Paradox of the Plankton (Hutchinson, 1961);" how are very similar organisms able to coexist simultaneously? Synechococcus microdiversity is an excellent case study of such microbial diversity questions, and how such diversity is maintained. Do clades differ slightly in physiology, or do ecological selection factors make this co-occurrence possible? While at times oligotypes within the same and different clades do co-exist, their relative abundances change over the season. As hypothesized by Hutchinson (1961), changes in the environment (such as the spring warming) appear to prevent any one clade from excluding all others. Further application of Synechococcus oligotyping in other environments may shed light on these long standing questions by providing new insights on (1) the potential competitive, commensal, and mutualistic interactions that may be occurring among closely related Synechococcus oligotypes, (2)

#### REFERENCES


the role of rare oligotypes within the community, and (3) the drivers of Synechococcus biogeography under current and future changing environmental conditions.

#### AUTHOR CONTRIBUTIONS

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

#### ACKNOWLEDGMENTS

Precipitation data was accessed via the KNMI Climate Explorer database. We thank H. Morrison for assistance with sequencing, R. Paerl and L. Stal for reviewing the manuscript, and P. Lescault for early discussions of the data set. This work was supported through a subcontract from the Woods Hole Center for Oceans and Human Health, from the National Institutes of Health and the National Science Foundation (NIH/NIEHS 1 P50 ES012742- 01 and NSF/OCE 0430724), a National Research Council Research Associateship Award and L'Oreal USA Fellowship (JH), an Alfred P. Sloan Research Fellowship in Ocean Sciences and the Clare Boothe Luce Program (KM), NASA Astrobiology Institute Cooperative Agreement NNA04CC04A (MS), the Alfred P. Sloan Foundation's ICoMM field project, and the W. M. Keck Foundation.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.01496/full#supplementary-material

adaptation in marine Synechococcus spp. J. Bacteriol. 188, 3345–3356. doi: 10.1128/JB.188.9.3345-3356.2006


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

# Synechococcus Assemblages across the Salinity Gradient in a Salt Wedge Estuary

Xiaomin Xia\*, Wang Guo, Shangjin Tan and Hongbin Liu\*

Division of Life Science, The Hong Kong University of Science and Technology, Hong Kong, Hong Kong

#### Edited by:

Hongyue Dang, Xiamen University, China

#### Reviewed by:

Yonghui Zeng, Aarhus University, Denmark Hanna Maria Farnelid, Linnaeus University, Sweden

\*Correspondence:

Hongbin Liu liuhb@ust.hk Xiaomin Xia xxia@connect.ust.hk

#### Specialty section:

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

Received: 30 March 2017 Accepted: 21 June 2017 Published: 06 July 2017

#### Citation:

Xia X, Guo W, Tan S and Liu H (2017) Synechococcus Assemblages across the Salinity Gradient in a Salt Wedge Estuary. Front. Microbiol. 8:1254. doi: 10.3389/fmicb.2017.01254 Synechococcus are the most abundant and widely distributed picocyanobacteria in the ocean. The salt-wedge type of estuary possesses the complete horizontal and vertical gradient of salinity together with other physical and chemical parameters. In order to reveal whether such a complex environmental gradient harbors a high diversity of Synechococcus, we investigated the abundance, taxonomic composition and pigment genetic diversity of Synechococcus in surface and bottom waters across the salinity gradient in a salt-wedge estuary by flow cytometric analysis and pyrosequencing of the rpoC1 gene and cpcBA operon (encoding phycocyanin). Synechococcus were ubiquitously distributed in the studied region, with clear spatial variations both horizontally and vertically. The abundance and diversity of Synechococcus were low in the freshwater-dominated low salinity waters. By pyrosequencing of the rpoC1 gene, we have shown that with the increase of salinity, the dominant Synechococcus shifted from the freshwater Synechococcus to the combination of phylogenetic subcluster 5.2 and freshwater Synechococcus, and then the strictly marine subcluster 5.1 clade III. Besides, the composition of Synechococcus assemblage in the deep layer was markedly different from the surface in the stratified waters (dissimilarities: 40.32%-95.97%, SIMPER analysis). High abundance of clade III Synechococcus found in the brackish waters may revise our previous understanding that strains of this clade prefers oligotrophic environment. Our data also suggested that both the phylogenetic subcluster 5.3 Synechococcus, a lineage that was not well understood, and subcluster 5.1 clade I, a typical cold water lineage, were widely distributed in the bottom layer of the estuary. Clade I detected in the studied region was mainly contributed by subclade IG. Analysis of the cpcBA operon sequences revealed niche partitioning between type 1 and type 3 Synechococcus, with type 2 distributed broadly across the whole environmental gradients. Our results suggest that the salt wedge estuary provides various niches for different lineages of Synechococcus, making it an environment with high Synechococcus diversity compared with adjacent freshwater and shelf sea environments.

Keywords: salt wedge estuary, pyrosequencing, rpoC1 gene, cpcBA operon, salinity gradient

#### INTRODUCTION

fmicb-08-01254 July 4, 2017 Time: 16:4 # 2

Synechococcus is one of major components of the phytoplankton community in both marine (Glover et al., 1986; Partensky et al., 1999) and freshwater (Callieri and Stockner, 2002; Sarmento et al., 2008) ecosystems. Compared with Prochlorococcus, another important pico-cyanobacteria which dominate in the oligotrophic open ocean, Synechococcus have higher nutrient requirements and are therefore more abundant in coastal (Li, 1998; Flombaum et al., 2013) and upwelling waters (Partensky et al., 1999; Cuevas and Morales, 2006). For instance, the highest Synechococcus abundance was recorded in the Costa Rica Dome where strong upwelling occurs (Saito et al., 2005), varying between 1.2 × 10<sup>6</sup> and 3.7 × 10<sup>6</sup> cells mL−<sup>1</sup> . A high abundance of Synechococcus was also reported in the Red Sea (Veldhuis and Kraay, 1993), Baltic Sea (Kuosa, 1991), and Chesapeake Bay (Wang et al., 2011). In Hong Kong coastal waters, Synechococcus are also important primary producers in the summer, with the maximum abundance reaching 5.7 × 10<sup>5</sup> cells mL−<sup>1</sup> (Liu et al., 2014).

Synechococcus are divided into three major pigment types by their different phycobiliprotein compositions: type 1 binds only phycocyanobilin (PCB), type 2 binds both PCB and phycoerythrobilin (PEB), while type 3 contains PCB, PEB and phycourobilin (PUB) (Six et al., 2007). Hence, type 1 is also called PC-only Synechococcus while type 2 and 3 are PE-containing Synechococcus. Type 3 is further divided into four subtypes, 3a (low PUB), 3b (medium PUB), 3c (high PUB) and 3d (variable PUB), according to the compositional proportion of PUB relative to PEB. Studies have widely reported different geographical distributions of Synechococcus pigment types – type 1 is abundant in high nutrient and turbidity coastal and estuarine waters, type 2 prefers relatively clean coastal waters, while type 3 dominates in oligotrophic open ocean (Olson et al., 1988, 1990; Wood et al., 1998). Besides in situ fluorometer (Cowles et al., 1993) and FCM (Olson et al., 1988), recently molecular methods have been applied to study Synechococcus pigment diversity in marine waters (Crosbie et al., 2003; Haverkamp et al., 2009; Everroad and Wood, 2012; Liu et al., 2014; Xia et al., 2017). The cpeBA operon and cpcBA operon are two gene markers commonly used to identify Synechococcus pigment types (Crosbie et al., 2003; Haverkamp et al., 2008; Liu et al., 2014; Xia et al., 2017).

Taxonomically, cluster 5 marine Synechococcus is further divided into three subclusters, S5.1, S5.2 and S5.3 according to the gene markers, such as the 16S rRNA and rpoC1 (Herdman et al., 2001; Fuller et al., 2003; Mazard et al., 2012). These three subclusters are further composed of at least 19 phylogenetical lineages (Farrant et al., 2016; Xia et al., 2017). Studies that adopted culture independent methods have revealed niche differentiation in Synechococcus lineages (Zwirglmaier et al., 2008; Huang et al., 2012; Xia et al., 2017). For example, clade I is known as cold water Synechococcus, while clade II is dominant in tropical/subtropical warm waters. Previous studies have also reported that distinct Synechococcus communities were present in the oligotrophic oceanic waters and nutrient rich coastal waters (Scanlan and West, 2002). Environmental factors such as concentration and type of inorganic nitrogen (Ahlgren and Rocap, 2006), phosphate concentration (Tetu et al., 2009), temperature (Pittera et al., 2014), salinity (Rajaneesh and Mitbavkar, 2013; Xia et al., 2015), and trace metal (Ahlgren et al., 2014) are all known to influence the distribution of Synechococcus lineages. However, the niches of some Synechococcus lineages remain unknown.

Previous studies suggested that Synechococcus pigment genes, such as PE-encoding genes, have undergone horizontal gene transfers between Synechococcus lineages during the evolution of this genus (Six et al., 2007; Everroad and Wood, 2012). This makes it impossible to identify a Synechococcus taxonomic lineage and pigment type at the same time based on a single gene marker. For examples, the phylogenetic tree based on the cpeBA operon sequences clearly grouped several Synechococcus lineages (see Figure 3 in Everroad and Wood, 2012). On the other hand, some lineages are composed by different pigment types. For example, clade II Synechococcus have at least 4 pigment types: type 2, 3a, 3c, and 3d<sup>1</sup> (Roscoff Syenchococcus database). Hence, different from identification of Synechococcus pigment types which is based on cpeBA and cpcBA operon, taxonomic lineage of a Synechococcus is classified via housekeeping genes, such asITS (Haverkamp et al., 2008), 16s rRNA gene (Fuller et al., 2003), rpoC1 (Mühling et al., 2006), and petB (Farrant et al., 2016).

Synechococcus community composition in estuaries or river plumes is often distinct from that in saline waters. A study conducted in Hong Kong water has shown that the water influenced by freshwater discharge from the Pearl River is dominated by PC-only (type 1) S5.2 Synechococcus, freshwater Synechococcus, and Cyanobium, while the coastal water not directly impacted by the river plume is dominated by various clades of marine Synechococcus S5.1 (Xia et al., 2015). The study also suggested that Synechococcus imported by the freshwater discharge are an important component of the cyanobacterial phytoplankton in the estuarine ecosystems. Similar observation was also reported by the studies carried out in the Zuari estuary and Changjiang estuary (Rajaneesh and Mitbavkar, 2013; Chung et al., 2015).

Due to high nutrient inputs, estuaries often sustain high levels of productivity. Salt wedge estuaries with strong vertical salinity gradient harbor different microbial communities in the surface and deep water (Korlevic et al., 2016 ´ ). The Pearl River is one of the largest rivers in China with a typical salt wedge estuary in the wet season (Harrison et al., 2008). In contrast to the increasing salinity along the river-estuary-coastal water transition, nutrient concentrations gradually decrease (Harrison et al., 2008). The strong gradient of environmental conditions makes the Pearl River estuary an ideal place to evaluate factors affecting the spatial distribution of Synechococcus lineages. However, till now, no study of Synechococcus phylogenetic diversity and pigment diversity along the salinity gradient with different depths was conducted in this strongly stratified estuary.

In order to study Synechococcus abundance, community taxonomic composition and pigment diversity in the salt wedge estuary, we conducted a cruise in July 2014 to collect samples along a salinity gradient in the Pearl River-estuary-coast system.

<sup>1</sup>http://roscoff-culture-collection.org/strains/shortlists/taxonomic-groups/marinesynechococcus

Abundance of Synechococcus was evaluated by flow cytometric analysis. Synechococcus taxonomic composition and pigment diversity were assessed through pyrosequencing of the rpoC1 gene and cpcBA operon, respectively. The relationship between environmental factors and Synechococcus diversity was also analyzed.

## MATERIALS AND METHODS

#### Sample Collection

fmicb-08-01254 July 4, 2017 Time: 16:4 # 3

Samples were collected from the Pearl River estuary on a cruise conducted from 13 to 20 July 2014 (**Figure 1** and **Table 1**). Salinity, temperature and depth were measured by a conductivity-temperature-depth rosette system (CTD, Sea Bird Electronics). At each station, 0.5–1 L of water was collected from surface and bottom (1 m above the bottom) layers, pre-filtered through a 3.0 µm (47 mm) polycarbonate membrane (PALL Corporation) and then filtered onto a 0.22 µm (47 mm) polycarbonate membrane for DNA extraction. Membranes were frozen at –80◦C immediately after filtration. For counting Synechococcus abundance, 1.8 mL water from each station was fixed with seawater buffered paraformaldehyde (0.5%, final concentration), flash frozen in liquid nitrogen and stored at –80◦C. Water samples for nutrient measurement were filtered with 0.45µm cellulose acetate membranes and were stored at -20◦C until analysis. Analytical protocols for nutrients followed Dai et al. (Dai et al., 2008). The method detection limits are 0.5 µM for ammonia, 0.02 µM for nitrite, 0.07 µM for nitrate, and 0.17 µM for phosphate.

## Analysis of Synechococcus Abundance

Synechococcus cells were enumerated using a Becton-Dickson FACSCalibur flow cytometer equipped with dual lasers of 488 and 635 nm with a high flow rate (Liu et al., 2014). Ten microliter yellow–green fluorescent beads (1 µm, Polysciences, Warrington, PA, United States) were added to each sample as an internal standard. Flow cytometric data were analyzed using WinMDI software 2.9 (Joseph Trotter, Scripps Research Institute, LaJolla, CA, United States). PC-only and PE-containing type Synechococcus were counted following the method described by Liu et al. (2014). However, samples from F303 were lost.

#### DNA Extraction, PCR, and Sequencing

Genomic DNA was extracted using the PureLink Genomic DNA mini kit (Invitrogen, CA, United States) and was eluted in TE buffer (Tris-EDTA buffer: 10 mM Tris,1 mM EDTA,pH8.0). For amplification of the rpoC1 gene, the PCR followed the protocol of Mühling et al. (2006). The first round of PCR used the primer rpoC1-N5 and the C-terminal primer rpoC1-C, and the PCR products were used as templates for a second round of PCR with modified primer rpoC1-39F (50 -adaptor+barcode+GGNATNGTNTGYGAGCGYTG) and rpoC1-462R (5<sup>0</sup> -adaptor+CGYAGRCGCTTGRTCAGCTT) (Xia et al., 2015). The PCR products were gel-purified using the Qiaquick gel purification kit (Qiagen, Hilgen, Germany) as described by the manufacturer. Purified amplicons were sequenced using the GS Junior pyrosequencing system according to manufacturer instructions (Roche, 454 Life Sciences, Branford, CT, United States).

For amplification of the cpcBA operon sequences, we used the primer pair SyncpcB-Fw (5<sup>0</sup> -adaptor+barcode+ATGG CTGCTTGCCTGCG-3<sup>0</sup> ) and SyncpcA-Rev (5<sup>0</sup> -adaptor +ATC TGGGTGGTGTAGGG-3<sup>0</sup> ) designed by Haverkamp et al. (2008). The PCR reaction mixture was composed of 1 µL of template DNA, 2.5 µL of 10× PCR buffer, 0.5 µL of 10 mM dNTP mixture, 0.75 µL of 50 mM MgCl2, 1 unit of Platinum taq DNA polymerase (Invitrogen, CA, United States), and 1 µL of each forward and reverse primer (10 nM). Sterile MilliQ-grade water was added to a final reaction volume of 25 µL. The PCR reactions were run on a Bio-Rad PCR machine. The program was 5 min at 94◦C, followed by 40 cycles of 30 s at 94◦C, 30 s at 55◦C and 1 min at 72◦C. The final elongation step was 10 min at 72◦C. The PCR products were gel purified and sequenced in the GS Junior 454 sequencing system.

#### 454 Post-run Sequence Analyses

Analysis of the rpoC1 and cpcBA sequence was conducted using the microbial ecology community software program Mothur<sup>2</sup> (Schloss et al., 2009). Raw sequences were first processed by removing barcodes and primers, then only reads with an average quality score above 25 and length longer than 300 nt were taken into account. Sequences were then denoised using the command shhh.seqs with sigma value of 0.01. Sequences containing ambiguous bases and homopolymer longer than 8 bp were also screened. Chimeras were identified using the command chimera.uchime and were then removed. After the above quality control, sequences were identified by local Blast using BioEdit with an expectation value 0.01 (Hall, 1999). For the analysis of rpoC1 gene, sequences classified as Prochlorococcus and Synechocystis were removed, and the remaining sequences that were less than 90% identical to the S5.1 clades and 85% identical to S5.2, S5.3, Cyanobium, and FS reference sequences (Supplementary Table S1) were assigned as unclassified (Xia et al., 2015). Similarly, for the cpcBA operon, sequences were identified by the local blast with the expectation value 0.01. The reference sequences of the rpoC1 (Xia et al., 2015) and cpcBA operon were listed in Supplementary Tables S1, S2. The cpcBA operon reference sequences were from the NCBI GenBank database<sup>3</sup> and the pigment type of representative strains was determined according to the Roscoff Synechococcus database<sup>4</sup> and Everroad and Wood's work (Everroad and Wood, 2012). As there were three copies of the cpcBA operon in the genomic sequence of type 1 Synechococcus (Six et al., 2007), the number of resulting type 1 sequences was divided by three in calculating the relative abundance of each Synechococcus pigment type. Coverge and operational taxonomic units (OTUs) numbers were calculated at the cutoff level of 3% for the rpoC1 gene and 5% for the cpcBA operon using Mothur's command summary.single.

<sup>2</sup>http://www.mothur.org/wiki/Download\_mothur

<sup>3</sup>https://www.ncbi.nlm.nih.gov/

<sup>4</sup>http://roscoff-culture-collection.org/strains/shortlists/taxonomic-groups/marine -synechococcus


<sup>∗</sup>Lower than the limit of detection.

OTUs which contain only 1 sequence were removed. The relative abundance of each OTU in a sample was calculated using the command get.relabund. The Margalef's species richness (d = (S–1)/ln(N), where S is total OTU number and N is total reads of each sample) and diversity (Shannon index H<sup>0</sup> ) were calculated. Similarity percentage (SIMPER) analysis of the dissimilarity between Synechococcus communities was carried out using Primer 5 (Primer-E Ltd., Plymouth, United Kingdom). The Spearman correlation between Synechococcus groups was calculated using R package Corrplot (Wei, 2016). Only the

fmicb-08-01254 July 4, 2017 Time: 16:4 # 4

correlations with P-value less than 0.05 were considered as significant and were thus visualized.

#### Phylogenetic Analysis of the rpoC1 and cpcBA Sequences

The representative sequences of the 40 most abundant OTUs for the rpoC1 gene (covered 73.1% of total reads) and cpcBA operon (covered 65.7% of total reads) were extracted and aligned with the reference sequences using ClustalW (Thompson et al., 2002) according to their codon structures. Modeltest and maximum likelihood phylogenetic tree construction were done by using Mega 6 (Tamura et al., 2013), in which the model used for the rpoC1 was GTR+G+I and that for the cpcBA operon was TN92+G+I. Bootstrap confidence analysis was carried out with 200 replications for evaluating the robustness of the tree topologies. A heatmap showing the relative abundance of each OTU was generated using iTol (Letunic and Bork, 2007).

#### Sequence Submission

All sequences obtained from this study have been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) under accession numbers: SRS2048774–SRS2048789 and SRS2048826–SRS2048834 (Supplementary Table S3).

#### RESULTS

#### Environmental Conditions of the Sampling Stations

As shown in **Figure 1**, strong salinity gradients between the surface and bottom waters were recorded in all sampling stations, except the well-mixed stations A2 and F303. The surface water salinity ranged from 0 to 33 ppt. Along the Pearl River-estuary-coast transect, temperature of the surface waters gradually increased whereas the bottom waters had an opposite pattern. The surface waters had a generally higher temperature and nutrient concentration than the bottom waters. Concentrations of phosphate, NO<sup>3</sup> <sup>−</sup>, NO<sup>2</sup> <sup>−</sup>, and NH<sup>4</sup> <sup>+</sup> were higher at the stations A2 and A6, which were strongly influenced by the freshwater discharge (**Table 1**). Higher salinity and lower nutrient concentration were recorded in station F303, due to the strong influence of offshore oceanic water.

#### Synechococcus Abundance

Synechococcus distributed ubiquitously in the Pearl River estuary and the adjacent coastal waters (**Figure 2**) with abundance ranging from 1.3 × 10<sup>4</sup> to 2.5 × 10<sup>5</sup> cells mL−<sup>1</sup> in the surface waters and from 5.9 × 10<sup>3</sup> to 2.0 × 10<sup>4</sup> cells mL−<sup>1</sup> in the bottom waters. The abundance of Synechococcus in the medium and high salinity stations were higher than that of the low salinity stations (A2 and A6). PE-containing Synechococcus were found in all samples and its abundance gradually increased with increasing salinity. The highest PE-containing Synechococcus abundance was detected in the surface water of A14, F404, and F504 (around 1.4 × 10<sup>5</sup> cells mL−<sup>1</sup> ). PC-only Synechococcus were also found

in all surface samples, however, they were only detected in the bottom water of stations A02, A06, A10, and A12. In the surface water of stations A10, A12, A14, and F404, PC-only Synechococcus abundance could reach 8.0 × 10<sup>4</sup> cells mL−<sup>1</sup> (**Figure 2**). In general, PC-only and PE-containing Synechococcus were more abundant in the surface waters than the bottom waters at all stations except A6.

#### Diversity of Synechococcus in the Pearl River Estuary

The number of the rpoC1 sequences obtained by pyrosequencing was listed in Supplementary Table S3. The diversities of Synechococcus assemblages were estimated by the Margalef's species richness index and Shannon diversity index (Supplementary Figure S1). The surface water of stations A2 and A6, which had low salinities, displayed the lowest Synechococcus richness and diversity. The richness and diversity of Synechococcus in the bottom waters did not show a large variation, and they were higher in the bottom than the surface waters at all stations, except A10 and A14.

In the phylogenetic tree, all euryhaline (clade VIII, S5.2 and Cyanobium) and freshwater Synechococcus formed a cluster that is separated from the strictly marine Synechococcus clades (Supplementary Figure S2). In the Pearl River estuary, most S5.2 Synechococcus were affiliated with WH8007. Freshwater

Synechococcus were identified into two lineages, FS\_I and FS\_II, and FS\_I had a narrower distribution than FS\_II. In the phylogenetic tree, FS\_I sequences were affiliated with the uncultured clone sequences from the Tucuru hydroelectric power station reservoir in Brazilian Amazonia, while FS\_II sequences were clustered with PS675 and PS676 isolated from Lake Teganuma (Japan) (Supplementary Figure S2). In A2S, almost all of freshwater Synechococcus were contributed by FS\_II, while that in A6S were mainly from FS\_I. OTUs which were belonged to freshwater Synechococcus, Cyanobium and S5.2 had high relative abundance in the medium salinity waters, while those belonged to clade III, such as OTU1 (contain the most reads), mainly occurred in the medium and high salinity waters. Moreover, S5.3, one of the major group Synechococcus in the studied region, had higher relative abundance in the bottom waters. It could be further classified into three subgroups, one was formed by previously reported strains RCC307 and Minos 01, the second by OTU12 and OTU18, and the third by OTU11 and OTU34. However, clade II, which was reported as the dominant Synechococcus in tropical/subtropical warm waters by previous studies (Zwirglmaier et al., 2008; Huang et al., 2012; Xia et al., 2017), was not abundant in the studied area. It is surprising that OTU25, which was widely distributed in the bottom water of the Pearl River estuary (A6B, A12B, A14B, F504B, and F404B with relative abundance from 0.01 to 5.5% of sample's reads), was grouped with clade I Synechococcus - a typical cold water lineage.

## Composition of Synechococcus Assemblages in the Pearl River Estuary

Altogether, 21 Synechococcus lineages were identified from 16 samples based on rpoC1 gene (**Figure 3**). Freshwater Synechococcus could be detected in all samples, with relative abundance ranging from 0.25 to 99.97% of each samples' reads (**Figure 3**). More than 98% of the detected cells were freshwater Synechococcus in A2S and A6S, where the salinity was lower than 6 ppt. It was found that the dominant Synechococcus in the surface waters had shifted with the increase of salinity, from freshwater Synechococcus to a combination of freshwater Synechococcus and S5.2, and then to S5.1. High relative abundance of clade III was mainly recorded in the A14S, F504S, F404S and F303S, where the salinity is intermediate to high. Clade V, which was also a major S5.1 Synechococcus in the studied area, only had high relative abundance in stations F404S and F303S (**Figure 3**).

In general, Synechococcus assemblage compositions in the bottom layer were markedly different from that in the surface water layer (**Figure 3**). Freshwater Synechococcus largely dominated the bottom water of A2 while Cyanobium, S5.2 and 10 clades of S5.1 Synechococcus were also detected. Compared with sample A2B, A6B were found with a higher relative abundance of S5.2 and Cyanobium instead of the freshwater Synechococcus. Moreover, Clade III and S5.3 had a high relative abundance in the bottom water of high salinity stations. Clades I and II were detected in all bottom samples (except A2B and F303B which had no clade I) with relatively low abundance. The highest relative abundance of clade I was detected in F404B, reached 7.89%. Phylogenetic analysis of clade I rpoC1 sequences showed that

OTU71 and OTU85 were affiliated with subclades IC and IA, respectively (Supplementary Figure S3). However, OTU25, which was the most abundant clade I OTU, did not group with reference sequences of reported subclades (Xia et al., 2017) (82%–86% nt identity to the subclades' representative sequences and 99% to uncultured Synechococcus RFLP-type S14 (AJ584725.1)) and may belong to a novel subclade (Supplementary Figure S3).

The dissimilarity between surface and bottom Synechococcus communities was analyzed using SIMPER analysis (**Figure 4**). The lowest dissimilarity (20.96%) was detected at station F303, where water was well mixed. The dissimilarity in the stratified stations ranged from 40.32 to 95.97%. The highest dissimilarity occurred at station A6, which was mainly contributed by FS\_I, FS\_II and S5.2. FS\_II and clade III were the major contributors of the dissimilarity at stations A10 and A14, where FS\_II had higher relative abundance in the surface waters, while clade III were relatively more abundant in the bottom. S5.3, which was mainly distributed in the bottom waters, was also a major contributor to the dissimilarity at stations A12, A14, and F504.

Spearman's correlation coefficients were calculated between the Synechococcus lineages and environmental factors (**Figure 5**). Being significantly correlated to each other positively, clades I, II, XVI, CRD1, and S5.3 were inversely correlated with temperature and were mainly distributed in the bottom layer. Besides, freshwater Synechococcus FS\_I was strongly negatively associated with salinity and positively related with nutrient concentrations, which was contrasting to clades III, IX, WPC1, and S5.3 which preferred high salinity and low NO<sup>3</sup> <sup>−</sup> environment. It was noted that the Synechococcus lineages with the highest relative abundance in the Pearl

River estuary, clade III and freshwater Synechococcus (FS\_I and FS\_II), were negatively correlated to each other, which indicates an opposite distribution pattern. On the other hand, euryhaline Synechococcus S5.2 was highly positively correlated with Cyanobium, which suggests that they shared similar niches.

#### Synechococcus Assemblage Harboring in the Surface and Bottom Waters had Different Pigment Compositions

Based on the successful amplification and sequencing of the cpcBA operon sequences from eight samples (the other samples did not amplify) (Supplementary Table S3), 4 well-separated clusters were formed in the phylogenetic tree (Supplementary Figure S4). Although type 1, 2, and 3b Synechococcus could be easily classified by the sequencing of cpcBA sequence, PUB containing Synechococcus type 3a, 3c, 3d, and 3f (recently defined by Mahmoud et al., 2017) could not be distinguished from each other (Supplementary Figure S4). Type 3 sequences from S5.3 formed a clade (hereafter named S5.3-Type 3) and were separated from the clade formed by those from S5.1 (hereafter named S5.1-Type 3). The phylogenetic tree also shows that most of the type 1 OTUs were affiliated with PS673 and PS676. Only 1 of the 40 most abundant OTUs was identified as S5.3-Type 3, which was mainly distributed in the bottom waters.

Distributed widely in the surface samples (**Figure 6**), proportion of type 1 decreased gradually while type 2 increased with increasing salinity. Only a small portion of Synechococcus detected was identified as type 3 at the stations of lowest salinity (A6S and A10S), comparing to more than 44.8% in the oceanic

water (F303S). Besides, Synechococcus pigment compositions in the surface and bottom waters at the two stratified stations (A10 and F504) were remarkably different. While station A10B was dominated by S5.1-type 3 Synechococcus, A10S were mainly dominated by type 1 and type 2. Moreover, the relative abundance of type 3 Synechococcus was also greatly higher in the bottom than in the surface at station F504. S5.3-Type 3, which was not abundant in the surface waters, had higher relative abundance in the bottom water of stations A10 and F504. However, in well mixed station F303, similar Synechococcus pigment composition in the surface and bottom layers were detected, which were composed of more type 2 and 3 cells than type 1.

## DISCUSSION

The abundance and diversity of Synechococcus were extensively studied in various marine environments, from oligotrophic open ocean to subtropical coastal and estuarine waters. However, none of the studies systematically reported the Synechococcus diversities in the salt wedge estuaries. Here, we used flow cytometric analysis and pyrosequencing method to assess the abundance, pigment diversity (based on the cpcBA operon) and taxonomic diversity (based on the rpoC1 gene) of Synechococcus in the Pearl River estuary, a typical salt wedge estuary in summer. Our results revealed that Synechococcus were highly abundant in this subtropical estuary, with a clear spatial variation in phylogenetic composition and pigment diversity along the surface salinity gradient, as well as between the surface and bottom waters.

Previous study has suggested that next generation sequencing methods with high sensitivity could yield more insights into the Synechococcus community composition than the traditional clone library method and flow cytometry approach (Xia et al., 2015). Consistently, in the present study, PC only Synechococcus were detected in all samples by using the pyrosequencing method while they could not be detected in some bottom samples by applying the flow cytometry approach. Moreover, using sequencing method, different pigment types and phylogenetic groups can be identified, providing more information about the composition of Synechococcus community.

## Abundance of Synechococcus along the Salinity Gradient of River Plume

High abundance of Synechococcus (up to 2.5 × 10<sup>5</sup> cells mL−<sup>1</sup> in surface waters) was observed in the Pearl River estuary in July, which was higher than most other marine environments (Flombaum et al., 2013), suggesting that Synechococcus were important primary producers in the subtropical river-impacted coastal water (Qiu et al., 2010). Spatial variations in Synechococcus abundance and the distribution of Synechococcus groups observed in the Pearl River estuary (**Figure 2**) was consistent with the studies carried out in other estuaries, such as Chesapeake Bay (Wang et al., 2011) and Zuari estuary (Rajaneesh and Mitbavkar, 2013), which have also displayed increasing Synechococcus abundance along the salinity gradient. Low salinity (Wang et al., 2011; Rajaneesh and Mitbavkar, 2013; Xia et al., 2015) and light limitation (Harrison et al., 2008) could be the reasons of low Synechococcus abundance in the freshwater-dominated estuarine water.

#### Shifts in Phylogenetic Composition and Pigment Diversity of Synechococcus Assemblages Along the Salinity Gradient in Subtropical River-Estuary-Shelf

The phylogenetic compositions of Synechococcus assemblage (assessed using the rpoC1 gene) varied along the salinity gradient. It is not surprising that freshwater Synechococcus were dominant in the inner field of the estuary (A2S and A6S), where turbid river water reigns. However, it was shown in the phylogenetic analysis that most of the Synechococcus detected in these two samples belonged to two distinct OTUs, OTU2 and OTU3, which suggests the niche differentiation among subgroups of freshwater Synechococcus. Freshwater Synechococcus were also abundant in A10S, A12S, and A14S, of which the salinity ranged from 13.1 to 19.7 ppt. This observation contrasted with the study in the Chesapeake Bay, the largest estuary in the United States, where freshwater Synechococcus are rare (Chen et al., 2006). Apart from the freshwater Synechococcus, euryhaline Synechococcus S5.2, and Cyanobium were also abundant in the intermediate salinity water. Their preferences of higher salinity environments compared with the freshwater Synechococcus agreed with the finding of a previous study that S5.2 Synechococcus has a high ability to deal with low salinity stress but requires elevated salinity for growth (Wang et al., 2011). Co-occurrence of S5.2 and Cyanobium was also reported by a study conducted in Hong Kong water (Xia et al., 2015) and Baltic Sea blackish water (Celepli et al., 2017), suggesting the two Synechococcus lineages have similar physiological and

ecological characteristics. However, the Spearman analysis did not show any strong correlation between the distribution of these two lineages and any measured environmental factors (**Figure 5**).

The proportion of S5.1 lineages increased with salinity (**Figure 3**). Celepli et al. (2017) reported that in the southern Baltic Sea, Synechococcus community transitioned from being dominated by euryhaline Synechococcus and Cyanobium to a mix of euryhaline and marine Synechococcus strains of S5.1 taking place at a salinity of 13–16 ppt. Similarly, our study showed that the transition occurred at salinity around 15 ppt in the Pearl River estuary (**Figure 3**). In the Baltic Sea coastal water, Synechococcus community is dominated by cold water clades I and IV, while the brackish and saline waters in the Pearl River estuary was widely dominated by the clade III. High relative abundance of clade III found in both the brackish and saline waters is consistent with the observation in the ECS (Choi et al., 2011; Xia et al., 2017). However, this is in contrast with the report in the Mediterranean Sea where clade III Synechococcus was mainly found in high salinity, oligotrophic, and phosphate-depleted water (Mella-Flores et al., 2011). The contrasting results observed by different studies were accounted by the fact that clade III contains several ecologically significant taxonomic units (ESTUs) with distinct niche preferences (Farrant et al., 2016). Furthermore, a strongly positive correlation of clade III and WPC1 (first found in the East China Sea and the Japan Sea (Choi and Noh, 2009)) was shown in the correlation analysis (**Figure 5**), which coincides with the finding of co-occurrence of clade III and WPC1 reported by previous studies (Choi et al., 2015; Xia et al., 2017). Besides that, clades V and VI, which overall distribution is not well understood, also co-occurred with clade III. Clades III, V and, VI and III were negatively related to nutrient concentrations, suggesting they have preferences of oceanic environment. Clade XV, which mainly occur between 30◦ and 35◦N/S (Huang et al., 2012; Sudek et al., 2015) and in upwelling regions (Sohm et al., 2016), was also distributed in the surface of F504 with relatively high relative abundance. Although previous studies reported that clade II is the dominant clade in the tropical/subtropical warm water (Zwirglmaier et al., 2008), we found this clade not abundant in the Pearl River estuary and its adjacent coastal water. Low abundance of clade II in this area may be due to the fact that clade II has fewer regulators (Palenik et al., 2006) to adapt to such dynamic and highly variable estuary-shelf environment. As a single Synechococcus clade can possess different pigment types, it is impossible to identify pigment types based on housekeeping genes, such as 16S rRNA and rpoC1 (Haverkamp et al., 2009; Everroad and Wood, 2012; Xia et al., 2017). Instead, the analysis of cpcBA operon (encoding phycocyanin) and cpeBA operon (encoding phycoerythrin) were applied to study Synechococcus pigment diversity in marine environments. Using the cpeBA sequence, a recent study found four groups of Synechococcus pigment types: 2, 3a, 3dA and the combination of 3c and 3dB can be identified (Xia et al., 2017). However, this gene marker cannot be applied to identify PC-only Synechococcus because they do not have the cpeBA operon. Hence, in this study, we used the cpcBA operon for studying pigment diversity in the Pearl River estuary. Haverkamp et al. (2009) suggested that the high phylogenetic resolution provided by the cpcBA operon is useful to assess the microdiversity of Synechococcus strains. Phylogenetically, this gene marker is capable of differentiating type 1, 2 and type 3 Synechococcus, while subtypes of type 3 (3a, 3c, and 3d) cannot be distinguished (Supplementary Figure S4). Yet, we found that this gene marker allows us to assign type 3 to S5.1 or S5.3 (Supplementary Figure S4). Studies have reported that different Synechococcus pigment types often co-occur in a marine environment, while one phenotype generally predominates (Haverkamp et al., 2009). Consistently, we found co-occurrence of Synechococcus pigment types in our samples. Dominant pigment type shifted from type 1 to type 3 along the high turbid freshwater-dominated estuary to the shelf water, on top of the relatively abundant of the widely occurring type 2 Synechococcus across the whole study area. Such a distribution pattern supports the point that underwater light spectral properties have a strong selective pressure on Synechococcus populations (Vörös et al., 1998; Six et al., 2007; Stomp et al., 2007; Xia et al., 2017).

#### Markedly Different Synechococcus Assemblages Harboring in the Surface and Bottom Waters of the Salt Wedge Estuary

The partition of Synechococcus lineages along depth is not as strong as the horizontal scale in marine water (Zwirglmaier et al., 2008). Therefore, Synechococcus assemblage composition in the surface water is generally representing the community at lower depth (Sohm et al., 2016). Indeed, Synechococcus assemblage had similar compositions in the surface and deep layers of the oceanic station F303, where strong mixing occurred. However, the assemblage displayed vertical differentiation in the stratified water. The surface water, which was a mixture of freshwater and marine water, was characterized with low salinity and high nutrient (Harrison et al., 2008). This environment would favor the selection of euryhaline strains which have a higher requirement of nutrients. On the other hand, the deep layer features high salinity but relatively low nutrient marine water (Harrison et al., 2008) which is suitable for the growth of strictly marine Synechococcus. For example, in the surface water of A10 and A12 euryhaline S5.2 Synechococcus had high relative abundance, while S5.1 Synechococcus had high proportion in the bottom waters.

Interestingly, S5.3, a minor group in marine environments, was widely detected from the bottom layer of stratified stations. S5.3 has at least six clades and shows depth partitioning (Huang et al., 2012). S5.3-I, represented by RCC307, is mainly present in surface water layer, while S5.3-II, -IV, -V, and –VI prevail in the medium to low light layer (Huang et al., 2012). Based on the rpoC1 gene sequence, we found that S5.3 in the Pearl River estuary was not as diverse as in the open ocean and was abundant in the bottom layer (Supplementary Figure S2). Their distribution was significantly positively related to salinity while negatively correlated with temperature, NH<sup>4</sup> <sup>+</sup> and NO<sup>3</sup> − (**Figure 5**). This is in agreement with Hashimoto et al.'s (2012) observation that S5.3 mainly occurs in deep waters. Apart from S5.3, we observed that clade I also widely occurred in the bottom layer where temperature could exceed 23◦C. This is in contrast with the conclusion of previous studies that clade I is restricted in high latitude cold water (Zwirglmaier et al., 2008; Huang et al., 2012; Sohm et al., 2016). A recent study reported that clade I contains at least six subclades with different thermal preferences (Xia et al., 2017). Consistently, only warm water subclades, IA and IC (see Figure 8 in Xia et al., 2017), were detected in the Pearl River estuary (Supplementary Figure S3). Besides these two subclades, OTU25, the most abundant clade I OTU, did not cluster with all reported subclades (Xia et al., 2017), but formed another novel subclade (subclade IG) (Supplementary Figure S3). The fact that subclade IG, defined by this study, was mainly distributed in deep water may be the reason why this subclade has not previously been detected. Huang et al. also detected clade I in the South China Sea at relatively deep layers of 75 and 100 m depth with relatively high abundance by sequencing 16S-23S rRNA internal transcribed spacer (ITS) (Huang et al., 2012). This suggests that clade I may be globally distributed and some subclades are specifically distributed in the deep water of tropical/subtropical region.

#### CONCLUSION

The river-estuary-shelf continuum is a highly complex system, which provides a wide array of niches for a highly diverse Synechococcus assemblage ranging from freshwater Synechococcus to euryhaline and strictly marine Synechococcus. Our data suggest that Synechococcus lineages have markedly different abilities to deal with environmental variations. In the estuary, salinity is an important factor influencing the distribution of Synechococcus groups. More studies are needed to reveal the mechanisms involved in salinity tolerance. The fact that high abundance of clade III occurs in the brackish coastal water may revise our previous understanding that clade III prefers oligotrophic oceanic water. Our results further reveal that clade I and S5.3 contain subgroups that have different niches. Further studies should focus on isolation of Synechococcus strains from the studied area and the physiological traits of clades I, III, and S5.3 strains. Moreover, to uncover more details about the distribution of Synechococcus in the salt wedge estuary, high resolution sampling (both vertical and horizontal) need to be conducted in future studies.

#### AUTHOR CONTRIBUTIONS

HL designed the experiment. XX and WG performed the experiments. Data were analyzed by XX in collaboration with WG and HL. XX and HL wrote the manuscript. ST attended the cruise and collected FM and DNA samples. All authors reviewed and approved the final version of the manuscript.

## FUNDING

This study was funded by National Natural Science Foundation of China (NSFC) (41361164001), and partially supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No. T21/602/16). HL also acknowledge the support of RGC GRF 661813.

## ACKNOWLEDGMENTS

We are grateful to Ms. Candy Lee for analyzing flow cytometric data. Yanping Xu is greatly acknowledged for measuring the nutrients. We also thank Prof. Minhan Dai from Xiamen University for providing us opportunities to collect samples.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.01254/full#supplementary-material

## REFERENCES

fmicb-08-01254 July 4, 2017 Time: 16:4 # 11


cyanobacterium Synechococcus sp. WH8102. ISME J. 3, 835–849. doi: 10.1038/ ismej.2009.31


**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 Xia, Guo, Tan and Liu. 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.

fmicb-08-01254 July 4, 2017 Time: 16:4 # 12

# Patterns and Processes in Marine Microeukaryotic Community Biogeography from Xiamen Coastal Waters and Intertidal Sediments, Southeast China

#### Weidong Chen1,2, Yongbo Pan<sup>1</sup> , Lingyu Yu1,2, Jun Yang<sup>2</sup> \* and Wenjing Zhang<sup>1</sup> \*

<sup>1</sup> State Key Laboratory of Marine Environmental Science, Marine Biodiversity and Global Change Research Center, College of Ocean and Earth Sciences, Xiamen University, Xiamen, China, <sup>2</sup> Aquatic EcoHealth Group, Key Laboratory of Urban Environment and Health, Institute of Urban Environment, Chinese Academy of Sciences, Xiamen, China

#### Edited by:

Martin G. Klotz, Washington State University Tri-Cities, United States

#### Reviewed by:

Adélaïde Roguet, University of Wisconsin–Milwaukee, United States Stefan Bertilsson, Uppsala University, Sweden

#### \*Correspondence:

Wenjing Zhang zhangwenjing@xmu.edu.cn Jun Yang jyang@iue.ac.cn

#### Specialty section:

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

Received: 05 May 2017 Accepted: 20 September 2017 Published: 12 October 2017

#### Citation:

Chen W, Pan Y, Yu L, Yang J and Zhang W (2017) Patterns and Processes in Marine Microeukaryotic Community Biogeography from Xiamen Coastal Waters and Intertidal Sediments, Southeast China. Front. Microbiol. 8:1912. doi: 10.3389/fmicb.2017.01912 Microeukaryotes play key roles in the structure and functioning of marine ecosystems. Little is known about the relative importance of the processes that drive planktonic and benthic microeukaryotic biogeography in subtropical offshore areas. This study compares the microeukaryotic community compositions (MCCs) from offshore waters (n = 12) and intertidal sediments (n = 12) around Xiamen Island, southern China, using high-throughput sequencing of 18S rDNA. This work further quantifies the relative contributions of spatial and environmental variables on the distribution of marine MCCs (including total, dominant, rare and conditionally rare taxa). Our results showed that planktonic and benthic MCCs were significantly different, and the benthic richness (6627 OTUs) was much higher than that for plankton (4044 OTUs) with the same sequencing effort. Further, we found that benthic MCCs exhibited a significant distancedecay relationship, whereas the planktonic communities did not. After removing two unique sites (N2 and N3), however, 72% variation in planktonic community was explained well by stochastic processes. More importantly, both the environmental and spatial factors played significant roles in influencing the biogeography of total and dominant planktonic and benthic microeukaryotic communities, although their relative effects on these community variations were different. However, a high proportion of unexplained variation in the rare taxa (78.1–97.4%) and conditionally rare taxa (49.0– 81.0%) indicated that more complex mechanisms may influence the assembly of the rare subcommunity. These results demonstrate that patterns and processes in marine microeukaryotic community assembly differ among the different habitats (coastal water vs. intertidal sediment) and different communities (total, dominant, rare and conditionally rare microeukaryotes), and provide novel insight on the microeukaryotic biogeography and ecological mechanisms in coastal waters and intertidal sediments at local scale.

Keywords: biogeography, community assembly, plankton, benthos, variation partitioning analysis, neutral model, deterministic processes, stochastic processes

## INTRODUCTION

fmicb-08-01912 October 12, 2017 Time: 17:34 # 2

Microeukaryotes are found in almost all environments on Earth and cover a wide spectrum of cell sizes, shapes and taxonomic affiliations (Schaechter, 2012; de Vargas et al., 2015). Microorganisms, such as algae, protozoa and fungi play a variety of crucial roles in marine ecosystems from primary producers, predators, decomposers to parasites (Sherr et al., 2007). Microeukaryotic community compositions (MCCs) can vary among marine ecosystems with different environmental conditions (Wang et al., 2014; Massana et al., 2015). Sediment and water column form the two most different and important components of marine ecosystems, each with unique environmental conditions and microbial community structure (Feng et al., 2009; Yang et al., 2016). Such differences may lead to different patterns of microbial biogeography. To date, most studies have separately investigated either planktonic or benthic microbial communities (Gong et al., 2015; Yu et al., 2015). Unfortunately, studies comparing planktonic and benthic MCCs, and the relative influence of environmental/spatial factors on the distributions of planktonic and benthic MCCs in marine ecosystems are still extremely limited (Massana et al., 2015).

Revealing the assembly mechanisms that determine the microbial community composition, and how microbes respond to environmental and spatial change, are major challenges in microbial ecology (Hanson et al., 2012; Liu et al., 2015). The growing efforts to understand these mechanisms mainly focus on two major categories (Sloan et al., 2006; Logares et al., 2013; Liu et al., 2015; Wang et al., 2015a; Yang et al., 2016; Liao et al., 2017). The first one is the deterministic mechanism, which considers that environmental filtering determines the biogeographical pattern of microbes (Gilbert et al., 2012). In general, deterministic mechanism predicts that environmental factors will influence microbial community composition (Dumbrell et al., 2010). A large number of studies have confirmed that microbial community composition might be influenced by a variety of environmental variables including nutrients, salinity, pH, temperature and biotic interactions (e.g., predators) (Fernandez-Leborans et al., 2007; Wang et al., 2014, 2015b; Yu et al., 2015). To fully understand the importance of deterministic processes, the sampling scale should be large enough to incorporate multiscale environmental variables (Yeh et al., 2015). The second category is the stochastic mechanism, which suggests that microbial community assembly is influenced by stochastic processes (e.g., drift). Some studies have quantified the importance of stochastic processes using the neutral model (NM) published by Sloan et al. (2006) since it can correctly explain MCC in diverse environments (Logares et al., 2013; Roguet et al., 2015; Burns et al., 2016). Further, the spatial factors (e.g., distance decay), which are part of the stochastic processes, also cause widespread concern in the study of microbial community assembly. However, some literature suggests that both environmental filtering and stochastic processes simultaneously shape microbial biogeography (Liu et al., 2015; Wang et al., 2015c; Yang et al., 2016), therefore both of these two processes need to be considered when investigating the community assembly of plankton or benthos.

Recently, high-throughput sequencing (HTS) has revealed the biogeography of marine rare biosphere community for bacteria or archaea (Galand et al., 2009; Hugoni et al., 2013), which is essential for understanding overall microbial diversity (Telford et al., 2006). Some HTS studies of microbial diversity have provided insight into distribution patterns, and influencing factors, of both abundant and rare microbes in marine and freshwater systems (Logares et al., 2014; Gong et al., 2015; Liu et al., 2015). For example, Liu et al. (2015) revealed that local environmental factors (e.g., water temperature) significantly affected the distribution of rare bacterial taxa in lakes and reservoirs, whereas spatial factors predominately influenced the distribution of abundant taxa (AT). However, it is still unclear how different environmental and spatial factors affect the distribution of planktonic and benthic abundant/rare microeukaryotic taxa in marine ecosystems (Logares et al., 2014; Lynch and Neufeld, 2015). The abundant taxa (AT) and rare taxa (RT) have significantly different richness and community composition, play different ecological functions, and may have discrepant ecological niches (Liu et al., 2015). Thus, they are likely to have distinct responses to environmental and spatial changes (Pedrós-Alió, 2012; Logares et al., 2013), and we hypothesize that abundant and rare microeukaryotes have different assembly mechanisms between water and sediment habitats.

In this study, we compared the community composition of marine planktonic and benthic microeukaryotes, and their interactions with the environmental and spatial factors. Further, we assessed the relative importance of environmental and spatial factors that structure the distribution of microeukaryotic communities (including total, dominant, rare and conditionally rare taxa) around Xiamen Island, southeast China. We aimed to address the following main questions: (1) How different are the MCCs found in the Xiamen coastal waters and intertidal sandy sediments? (2) How are the distribution of planktonic and benthic MCCs (including total, dominant, rare and conditionally rare taxa) influenced by the environmental and spatial variables? (3) Do stochastic processes explain the community variation of plankton and benthos?

## MATERIALS AND METHODS

#### Study Area, Sampling and Environmental Factors

Sampling was conducted during two field cruises around Xiamen offshore area (118◦ 010–118◦ 14<sup>0</sup> E, 24◦ 250–24◦ 34<sup>0</sup> N, **Figure 1**). Three biological replicates were selected in water sample stations in the north (N), east (E), south (S), west (W) and sediment stations A, B, C, D to represent of local ecosystem conditions around Xiamen Island (**Figure 1**). The intertidal sandy sediment and coastal water sampling sites were not at the exact same locations, which were ascribed to the intertidal sandy sediment are only distributed in East and South of Xiamen Island. For water samples, the surface (<0.5 m) water samples were transported to the laboratory and processed immediately. For planktonic microeukaryotic community analyses, 800 ml of

surface seawater samples were pre-filtered by a 200 µm poresize sieve to remove debris, large metazoans and grains, and next the water samples with microeukaryotes (smaller than 200 µm) were filtered through 0.22 µm pore polycarbonate membrane (Millipore, Billerica, MA, United States). For sediment samples, the surface (0–5 cm) intertidal sandy sediment samples were transported to the laboratory and processed within 4 h of sampling. The microeukaryotes were mechanically shaken and separated from the sand with sterile seawater over five times, and mixture waters with microeukaryotes were consecutively pre-filtered by a 200 µm pore-size sieve (to remove debris, large metazoans and grains) and filtered/concentrated through 0.22 µm pore polycarbonate membrane (Millipore, Billerica, MA, United States). The membranes were stored at −80◦C until DNA extraction. To limit DNA contamination, the sterile seawater was pre-filtered through a nominal 0.22 µm pore-size membrane (Millipore, Billerica, MA, United States). Filtration equipment was rinsed with sterile water before each sample filtering.

At each sampling site, we measured environmental variables (temperature, pH, salinity, turbidity, chlorophyll-a, and dissolved oxygen) using a Hydrolab DS5 multiparameter water quality meter (Hach Company, Loveland, CO, United States). Total carbon (TC) and total nitrogen (TN) were analyzed with a TOC/TN-VCPH analyzer (Shimadzu, Tokyo, Japan) and total phosphorus (TP) was determined using spectrophotometry according to established standard methods (Wang et al., 2015b). In addition, a Lachat QC8500 Flow Injection Autoanalyzer (Lachat Instruments, Hach Company, Loveland, CO, United States) was used to determine nitrite and nitrate nitrogen (NOX-N), and phosphate phosphorus (PO4-P) concentrations.

#### DNA Extraction, PCR and Illumina Sequencing

Planktonic and benthic DNAs were extracted by E.Z.N.A. DNA Kit (Omega Bio-Tek Inc., Norcross, GA, United States) and FastDNA SPIN kit (MP Biomedicals, Santa Ana, CA, United States) following the manufacturers' instructions,

respectively. The hyper-variable V9 region of eukaryotic 18S rDNA was amplified using the primers 1380F and 1510R (Amaral-Zettler et al., 2009). Each DNA sample was individually PCR-amplified. The 30 µl PCR mixture contained 15 µL of Phusion Master Mix (New England Biolabs, Beverly, MA, United States), 0.2 µM of forward and reverse primers and 10 ng of the sample DNA. PCR reactions including 1 min initial denaturation at 98◦C, followed by 30 cycles of denaturation at 98◦C for 10 s, annealing at 50◦C for 30 s, extension at 72◦C for 60 s. Finally, the amplicons were subjected to 10 min extension at 72◦C. Triplicate PCR products for each sample were conducted and purified using GeneJET Gel Extraction Kit (Thermo Scientific, Hudson, NH, United States). Sequencing libraries were generated using NEB Next Ultra DNA Library Prep Kit for Illumina (New England Biolabs, Beverly, MA, United States) according to manufacturer's instructions, and index codes were added. The library quality was evaluated using the Agilent Bioanalyzer 2100 system (Agilent Technologies, Palo Alto, CA, United States) and Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, United States). The barcoded amplicons from each sample were mixed in equimolar amounts and were then sequenced using an Illumina Miseq platform (Illumina Inc., San Diego, CA, United States) following a paired-end (2 × 250 bp) approach (Caporaso et al., 2012).

#### Bioinformatics

Sequenced paired-end reads were merged with FLASH (Magocˇ and Salzberg, 2011). Raw data were processed and analyzed using the QIIME v.1.8.0 to remove low-quality reads (Caporaso et al., 2010). Sequences were quality controlled using the following settings: maximum number of consecutive low-quality base = 3; minimum of continuous high-quality base = 75% of total read length; ambiguous bases >0 were removed, last quality score = 3 (Liu et al., 2017). Chimeras were identified and removed using UCHIME before the downstream analyses (Edgar et al., 2011). After that, sequences were clustered into OTUs using UPARSE (Edgar, 2013) with the 97% sequence similarity cut-off. Representative sequences in each OTU were selected and blasted against the SILVA 115 reference database (Quast et al., 2013). Unassigned OTUs (sequence similarity to a reference sequence is <80%) and singletons were discarded prior to further analyses, resulting in a mean of 72048 sequences recovered per sample (range from 33996 to 116717). To minimize biases associated with sequencing coverage and allow for comparison between samples, we randomly selected 33996 sequences from each sample.

#### Definition of Abundant and Rare Taxa

We artificially defined thresholds as 0.01% for rare taxa (RT) and 1% for abundant taxa (AT), and classified all OTUs into six categories based on their relative abundance as previously described (Liu et al., 2015; Dai et al., 2016). (1) AT were defined as the OTUs with abundance ≥1% in all samples; (2) RT were defined as the OTUs with abundance <0.01% in all samples; (3) moderate taxa (MT), OTUs with abundance between 0.01 and 1% in all samples; (4) conditionally rare taxa (CRT), with abundance below 1% in all samples and <0.01% in some samples; (5) conditionally abundant taxa (CAT), taxa with abundance ≥0.01% in all samples and ≥1% in some samples but never rare (<0.01%); (6) conditionally rare and abundant taxa (CRAT), OTUs with abundance varying from rare (<0.01%) to abundant (≥1%) (Dai et al., 2016). In this study, we combined AT, CAT, and CRAT as AT to perform further analyses with these three pooled categories taxa (AT, CAT, and CRAT) referred to as 'dominant taxa' to avoid confusion. The ecological studies usually consider 'rarity' to be a continuous variable, therefore, there is always an effect of arbitrariness when defining a cutoff point for rarity (Gaston, 1994). To reduce the effect of arbitrary definition of dominant and rare OTUs, we performed the Multivariate Cutoff Level Analysis (MultiCoLA) to assess the impact of various abundance or rarity cutoff levels on our resulting data set structure and on the consistency of the further ecological interpretation (Gobet et al., 2010). MultiCoLA is an effective method to systematically explore how large community data sets are affected by different definitions of rarity.

## Analyses of Community Diversity

Rarefaction curve, Venn diagram and alpha diversity indices including OTU richness, ACE (abundance based coverage estimator), Chao 1, Shannon-Wiener, Pielou's evenness indices and Simpson index of diversity (1-D) were calculated for each sample or entire samples in vegan 2.4-1 with R software (version 3.2.3) (R Core Team, 2015). Good's coverage was performed in MOTHUR v.1.33.3 software (Schloss et al., 2009). We compared alpha-diversity using one-way ANOVA and Student's t-test.

Bray–Curtis similarity matrix is considered to be one of the most robust similarity coefficients for ecological studies (Kent, 2012) and was applied to our microeukaryotic community dataset. The non-metric multidimensional scaling analysis (NMDS) was employed for detecting possible differences in microeukaryotic planktonic and benthic communities using PRIMER v.7.0 package (PRIMER-E, Plymouth, United Kingdom) (Clarke and Gorley, 2015). Significant difference (P < 0.01) between groups was assessed by analysis of similarities (ANOSIM). The analysis of similarities statistic global R represents separation degree of between-group and within-group mean rank similarities. R = 0 indicates no separation, whereas R = 1 indicates complete separation (Clarke and Gorley, 2015).

#### Relationships between Community Composition, Environmental Variables, and Geographical Distance

To explore the potential controlling factors for the composition of the microeukaryotic community, we used standard and partial Mantel tests to reveal the correlations between the community similarity and environmental factors. Environmental factors, except pH, were square-root transformed and Euclidean distances between samples were calculated. A geographical distance matrix was calculated based on the longitude and latitude coordinates of each sampling site.

Spearman's rank correlation coefficients were calculated to explore the relationships between the Bray–Curtis similarity of microeukaryotic communities and the geographical distance/environmental factors, and the correlation between the geographical distance and the Euclidean distance of all environmental factors.

We quantified the relative effects of environmental and spatial factors in shaping MCC with variation partitioning analysis (VPA) based on redundancy analysis (RDA), as previously described (Wang et al., 2015c). First, a set of spatial variables were generated through the method of principal coordinates of neighbor matrices (PCNM) analysis (Borcard and Legendre, 2002), based on the longitude and latitude coordinates of sampling sites. Subsequently, variance inflation factors (VIFs) were calculated to check the presence of collinearities among environmental and PCNM variables using the "vegan" package and variables with VIF >10 were removed to avoid the impact of collinearity. In addition, to provide unbiased estimates of the variation partitioning based on RDA, microeukaryotic data were Hellinger-transformed prior to the analyses (Legendre and Gallagher, 2001). Forward selection procedure was used to select environmental and spatial variables (Blanchet et al., 2008). Finally, VPA was performed using the "varpart" function of the package vegan. For the seawater habitat, we artificially removed sites N2 and N3 due to their distinct microeukaryotic plankton communities compositions with 10 other sites based on the results of NMDS analyses, and also performed both Mantel tests and VPA using the 10 other water samples.

#### Neutral Community Model for Microeukaryotes

We evaluated the fit of the Sloan neutral community model for MCCs to determine the potential importance of stochastic processes (Sloan et al., 2006, 2007), which considers OTU occurrence frequency in a set of local communities, and their regional relative abundance across the wider metacommunity. The model used here is an adaptation of the neutral theory (Hubbell, 2001) adjusted to be appropriate for large microbial populations. In this model, Nm is an estimate of dispersal between communities. The parameter Nm determines the correlation between occurrence frequency and regional relative abundance, with N describing the metacommunity size and m being immigration rate (Sloan et al., 2006). In this study, we separately used the following data sets – seawater, sediment, and all 24 samples. Further, for seawater habitat, we assessed the fit of the NM for all 12 water samples and 10 water samples (without sites N2 and N3), respectively. All computations were performed in R (version 3.2.3) (R Core Team, 2015).

#### Accession Number

All raw sequences from this study have been submitted to the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) database under the BioProject number PRJNA342297 and the accession number SRP089752.

## RESULTS

#### Comparison of Environmental Factors between Water and Sediment Pore Water

The environmental factors of the studied sites were summarized in Supplementary Table S1. In general, the dynamics of environmental factors from surface waters were more complex than those from intertidal sediment pore waters. The temperature in sediment sites (31.5–33.9◦C) was generally higher than water sites (28.1–30.2◦C) except N2 and N3 (31.5 and 31.7◦C, respectively). Mean values of salinity and pH in the water sites (26.4 psu and 9.35) were higher than sediment sites (25.0 psu and 7.97). However, both TN and TP showed higher mean values in sediment sites (3.07 and 0.112 mg l−<sup>1</sup> ) than water sites (0.57 and 0.018 mg l−<sup>1</sup> ).

#### Comparison of Richness and Diversity between Plankton and Benthos

Benthic communities showed a significantly higher species richness compared to planktonic communities (Supplementary Figures S1–S3), although Shannon-Wiener, Simpson index of diversity and Pielou's evenness indices did not exhibited significant differences. Rarefaction curves for each sample was seldom saturated (Supplementary Figure S1A). However, the global rarefaction curve indicated that sequencing of ∼0.81 million V9 rDNA sequences from 24 samples was sufficient to approach saturation of microeukaryotic richness (Supplementary Figure S1B). Further, the total number of microeukaryotic OTUs (9003) was roughly equivalent to the number estimated by abundance based richness estimators such as Chao 1 (9005 ± 1.74) and ACE (9026 ± 47) (Supplementary Table S2). Good's coverage ranged from 96.62 to 98.92% in each sample and the index of all 24 samples combined was 99.99%, respectively (Supplementary Table S2). The rarefaction curves, extrapolated species richness indices (Chao 1 and ACE) and Good's coverage indices indicated that the majority of the microeukaryotic taxa had been recovered from the studied metacommunity (Supplementary Figure S1 and Table S2).

A total of 9003 microeukaryotic OTUs were identified from 815 904 high-quality reads at 97% identity level for the 24 samples. Further, Supplementary Table S3 summarized the contribution of each microbial taxa category to microeukaryotic community. MultiCoLA showed that when the structure patterns of community data were compared between the truncated and the original matrices, little variation in data structure was observed after removing up to 60% of the rare types, indicating consistent ecological patterns between the two different matrices. In addition, when the increasing amount of rare types was >5%, consistent ecological patterns were also maintained between the truncated and the original matrices (Supplementary Figure S4). These results indicate that our definitions of dominant taxa (47.54%) and RT (1.88%) were reasonable and objective in this study.

The planktonic microeukaryotic metacommunity generated 4044 OTUs, while the benthic microeukaryotic metacommunity generated 6627 OTUs with the same sequencing depth

(Supplementary Figure S3). It appeared that the absolute number of OTUs was higher in benthic than planktonic metacommunities. The most abundant OTUs that had mean relative abundance >1% represented 29.88% and 21.39% sequences in 12 water and 12 sediment samples, respectively. All of these OTUs were classified as either Stramenopiles or Animalia. In addition, one Arthropoda OTU (OTU\_3) and one Diatomea OTU (OTU\_7452) were present in both water

and sediment habitats with mean relative abundance ≥1%

## Comparison of Microeukaryotic Community Compositions

(Supplementary Table S4).

Overall, the most diverse and abundant OTUs in both water and sediment habitats were assigned to the groups of Opisthokonta (mean relative abundance: 15.64% OTUs and 31.81% sequences for plankton vs. 19.63% OTUs and 40.98% sequences for benthos, respectively) and Stramenopiles (21.13% OTUs and 33.94% sequences for plankton vs. 19.69% OTUs and 27.17% sequences for benthos, respectively) (**Figure 2A**). Some groups such as Alveolata (21.84% OTUs and 12.82% sequences for plankton vs. 14.77% OTUs and 7.11% sequences for benthos, respectively) and Diatomea (14.13% OTUs and 31.88% sequences for plankton vs. 8.37% OTUs and 21.06% sequences for benthos, respectively) exhibited significantly higher abundance and diversity in the water habitat compared with the sediment habitat. However, some groups such as Excavata (1.19% OTUs and 0.33% sequences for plankton vs. 4.28% OTUs and 1.97% sequences for benthos, respectively) exhibited an inverse trend, which were more diverse and abundant in benthic habitat (**Figure 2A**). Further, Venn diagrams indicated that benthic microeukaryotes were more diverse than plankton (including total, dominant taxa, RT, and CRT), since more OTUs were only presented in sediment than in water habitats (**Figure 2B**).

There was a clear difference between planktonic and benthic communities with high global R-values in the ANOSIM tests (**Figure 3**). Interestingly, planktonic communities were further separated into two subgroups, the first subgroup comprised sampling sites N2 and N3, while the second contained 10 other sample sites. Further, the total, dominant, rare and conditionally rare communities showed similar patterns in all NMDS plots.

#### Relationships between Environmental, Spatial Factors and Microeukaryotic Community

The Mantel tests showed that three physicochemical variables (salinity, TP, and TN) were significantly related to the changes of all of the four categories (total, dominant taxa, RT, and CRT) of benthic communities (**Table 1**). Further, both PO4-P and NO<sup>X</sup> were found to be significant determinants of the total, dominant and CRT subcommunities (P <0.05). Temperature was significantly related to the variation of dominant and CRT for the benthic community. However, the Mantel tests for the planktonic communities (including 12 water sites and 10 water sites) showed that no environmental factor was significantly related to the variation of planktonic community composition.

The Bray–Curtis similarity in the total, dominant, rare and conditionally rare benthic communities between any pair of samples significantly decreased with the increasing of geographical distance (P < 0.01, **Table 1**), indicating that sites in sandy sediments which were geographically distant had less similar communities – presumably because of dispersal limitation. However, no significant distance-decay relationship was found for planktonic communities (including both of 12 sites and 10 sites, **Table 1**). Further, both planktonic and benthic samples that were closer to each other showed more similar environmental conditions (P < 0.01, Supplementary Figure S5). However, benthic sites exhibited a stronger environment-distance relationship than planktonic sites (Spearman's correlation coefficient 0.378 for planktonic sites and 0.839 for benthic sites, respectively).

The VPA indicated that both environmental and spatial factors play significant roles in structuring microeukaryotic metacommunity in water and sediment habitats (**Table 2**). For total and dominant taxa, the spatial factors exhibited a slightly higher contribution to the community variation than environmental factors, however, a high proportion of unexplained variation was observed in the RT (78.1–97.4%) and CRT (49.0–81.0%) (**Table 2**). For total MCC, the variance explained by purely spatial factors was slightly higher (27.1% vs. 24.2% for plankton and 30.2% vs. 19.1% for benthos, respectively) than that of purely environmental factors in both habitats. For the dominant subcommunity, spatial factors alone also explained higher community variations (28.5% vs. 23.2% for plankton and 33.0% vs. 18.5% for benthos, respectively) than the environmental factors. Interestingly, for the planktonic samples after removing sites N2 and N3, spatial factors exhibited a much higher contribution to the community variations than environmental factors among all of total, dominant taxa, RT, and CRT. Moreover, the contribution of environmental factors to the planktonic MCC variations exhibited a rapid decline among all of total (3.4%), dominant (11.0%), rare (0%), and conditionally rare (4.2%) planktonic taxa. For the benthic microeukaryotic community, the unexplained variation of the RT (78.1%) and CRT (49.0%) was higher than that of the total (46.3%) and dominant (40.4%) communities. Similarity, in the water sites (12 sites and 10 sites, respectively), the unexplained variation of the RT (87.1 and 97.4%, respectively) and CRT (53.3 and 81.0%, respectively) was also higher than that of the total (44.6 and 76.2%, respectively) and dominant (43.2 and 60.3%, respectively) communities.

#### Fit to the Neutral Model of Community Assembly

The NM estimated a low fraction of the variation in the frequency of occurrence of different OTUs in planktonic communities (**Figure 4A**, R <sup>2</sup> = 0.25). The benthic microeukaryotic community showed a larger fraction of the variation in the frequency of occurrence than planktonic microeukaryotic community (**Figure 4B**, R <sup>2</sup> = 0.43), indicating a medium fit to the NM, although the Nm-value was higher for planktonic communities (Nm = 10103) than for benthic communities (Nm = 9338). We

and abundant taxa; RT, rare taxa; CRT, conditionally rare taxa.

also tested the fit to the NM for 10 planktonic communities excluding sites N2 and N3, and found the highest Nmvalue (Nm = 34403). Interestingly, this model explained the largest proportion of the variability in occurrence frequency of planktonic community without sites N2 and N3 among the all different data-sets (**Figure 4C**, R <sup>2</sup> = 0.72). Further, we evaluated the fit to neutral community model on all 24 planktonic and benthic communities, and found that the observed data exhibited no fit to the neutral curve (Supplementary Figure S6, R <sup>2</sup> = −0.09, note that negative R 2 -values can occur when there is no fit to the model), indicating a strong role for habitat filtering relative to stochastic processes.

## DISCUSSION

## Microeukaryotic Composition and Diversity Patterns in Water and Sediment

Our DNA metabarcoding results validated a well-known observation that planktonic and benthic microeukaryotic communities composition and structure are significantly different (Zinger et al., 2011; Massana et al., 2015; Forster et al., 2016), which was evident in the alpha-diversity (e.g., OTU richness and ACE index), beta-diversity and taxonomic composition (e.g., Diatomea) (**Figures 2**, **3** and Supplementary Figures S2, S3). The comparison of MCCs between water and sediment habitats provided additional insight into the diversity and distribution of subtropical marine microeukaryotes.

The microeukaryotic community differences between planktonic and benthic taxa held at both supergroup and OTU levels, although 18.53% OTUs were common to both habitats (**Figure 2**). First, the dominant overlap between benthic and planktonic taxa in our data was related to Alveolata, Stramenopiles (represented mainly by Diatomea) and Opisthokonta (represented mainly by Arthropoda). Although a 200 µm pore-size sieve was used to remove large metazoans, we found that the Opisthokonta was an abundant group, and they were dominated by Arthropoda, which accounted for 23.62% of planktonic sequences and 11.76% of benthic sequences. One explanation for this is that some large-sized organisms can get through the 200-µm pores and/or presence of eggs, spores, or larvae of the large-sized organisms (Liu et al., 2017). Another alternative explanation for this is that our DNA-based molecular approach cannot exclude the potential effect by "free DNA in the water" or "animal debris" passing the 200 micron filter on the perceived microeukaryote communities (Thomsen and Willerslev, 2015; Liu et al., 2017). Both Alveolata and Stramenopiles are dominant benthic microeukaryotic groups, as is also shown by previous studies (Gong et al., 2015; Forster et al., 2016). However, our finding showed that Alveolata was more dominant in water habitat. Further, another study (Mann and Evans, 2007) showed that Diatomea were more often detected in the benthos compared with plankton among phototrophic protists, which is in contrast to our data. This difference might be ascribed to different sampling areas in these various studies. Excavata is a major eukaryotic supergroup, however, only a few studies have focused on it (Hampl et al., 2009) compared to other groups (e.g., Alveolata and Diatomea). Second, our study showed that benthic microbial communities exhibited higher richness than planktonic communities although there was a little overlap between the two communities (**Figure 2B**). Studies in the Gulf of Maine and Long Island Sound in North America also revealed little overlap between genetic signatures of the water and sediment microeukaryote assemblages (Doherty et al., 2010). One explanation for this observation might be that benthic habitats exhibited stronger horizontal and vertical gradients

TABLE 1 | Mantel tests for the correlations between geographical distance, environmental distance, individual environmental factors, and microeukaryotic community similarity in water and sediment using Spearman's coefficients.


Geo\_distance, pairwise geographical distances between sampling sites; Env\_distance, Euclidean distance of all environmental variables between sampling sites; Community similarity was based on the Bray–Curtis similarity. The significances are tested based on 999 permutations, and bold values indicate statistical significance (P < 0.05). Water samples with 10 sites indicated removing sites N2 and N3. All, all taxa; AT, abundant taxa; CAT, conditionally abundant taxa; CRAT, conditionally rare and abundant taxa; RT, rare taxa; CRT, conditionally rare taxa. TN, total nitrogen; TP, total phosphorus; DO, dissolved oxygen; TC, total carbon; NOX-N, nitrate and nitrite nitrogen; PO4-P, phosphate phosphorus.

in both physical and chemical characteristics. Environmental heterogeneity is much more pronounced in the benthic than in the planktonic sites, and higher environmental heterogeneity likely promotes the existence of highly specialized organisms, thereby probably driving species-richness patterns (Hortal et al., 2009).

Although further investigations are necessary to characterize the effects of DNA extraction protocols and PCR methods on MCCs, the robust patterns in microeukaryotic community were distinct between coastal waters and intertidal sediments. Recently, a study of amplicon-based characterization of microbial community structure revealed that using different DNA extraction kits did not significantly affect alpha diversity/richness, beta diversity and dominant members of the community of samples (Staley et al., 2015). Therefore, we consider that different DNA extraction kits did not significantly affect the overall biological conclusions drawn in this study. Our NMDS analysis showed that communities from the same habitat cluster together (**Figure 3**), which indicated a distinct composition among planktonic and benthic communities. Further, in the water habitat, the differences between sites N2 and N3 compared to the 10 other water sites have been showed by a previous study (Yu et al., 2015).

#### Controlling Factors Shaping the Microeukaryotic Community Compositions

The distinct distribution patterns of MCCs between water and sediment could be attributed to two types of mechanisms: environmental filtering and stochastic processes (Liu et al., 2015; Yang et al., 2016; Liao et al., 2017).


TABLE 2 | Variation partitioning of the microeukaryotic communities among environmental and spatial variables in water and sediment.

Environment, community variation explained by pure environmental factors; Spatial, community variation explained by pure spatial factors; Shared, community variation explained by joint effect of environmental and spatial factors. Bold values indicate statistical significance (P < 0.05). Negative value of adjusted coefficients of determination (Adjusted R<sup>2</sup> ) was converted to zero. Water (10 sites) indicated removing sites N2 and N3.

First, environmental factors such as nutrient concentrations can influence MCCs because they are essential for the growth and development of microorganisms, and different microorganisms are adapted to their optimal growth concentrations (Kneip et al., 2007; Wang et al., 2015b). MCCs can be both directly and indirectly affected by nutrient concentrations, which can influence the photosynthesis of autotrophs, and heterotrophic microeukaryotes which can prey on autotrophs (Hecky and Kilham, 1988). In this study, the Mantel tests revealed that benthic MCCs (total, dominant taxa, RT, and CRT) were significantly correlated with salinity, TP and TN in sediments (**Table 1**). These results are consistent with studies of bacterioplanktonic or microeukaryotic communities responses to TP or TN concentrations (Liu et al., 2011; Wang et al., 2015a). Further, salinity is also an important environmental factor affecting the survival and growth of marine microeukaryotes. Changes in salinity directly affect the osmoregulation and metabolism of microorganism, leading to effects on the life activities of microorganism (Decamp et al., 2003). However, our results indicate that there is a clear difference in correlation between plankton and benthos with environmental factors (**Table 1**). Interestingly, no measured environmental variable was significantly associated with planktonic MCC (including 12 water sites and 10 water sites). There are at least two reasons that might possibly explain the observed non-significant relationship between plankton and environmental variables. The first possibility is that plankton communities are extremely dynamic, and snapshot or even regular but low resolution sampling includes a large fraction of noise, which may mask main ecological patterns (Özkan et al., 2014). Second, variation in planktonic communities not explained by environmental (e.g., abiotic) controls might indicate strong biotic interactions (Wang et al., 2015b). In addition, there are some potential limitations that merit further discussion. Some unmeasured environmental

variables, such as irradiance and ocean currents, are important in shaping the distributions of MCCs (Yeh et al., 2015). The low statistical power for detecting patterns was perhaps largely due to small sample size in field metacommunity studies, and the sampling scale should be considered when designing sampling schemes (Martiny et al., 2011). That is, the sampling size should be large enough to include most important environmental gradients and dispersal factors.

Second, an increasing body of literature in microbial community ecology indicates that spatial factors are another major process shaping biogeographical patterns besides environmental selection (Vyverman et al., 2007; Cermeno and Falkowski, 2009; Hanson et al., 2012). Our results showed that spatial effects were significant in shaping benthic MCCs (total, dominant taxa, RT, and CRT), whereas marine planktonic MCCs (including 12 water sites and 10 water sites) do not adhere to significant distance-decay relationship (**Table 1**). Recently, Gong et al. (2015) surveyed microeukaryotes of coastal sediment sites in the Yellow Sea and revealed limited dispersal of benthic microeukaryotes, which showed a similar distribution pattern to our results. Comparable and similar results have been noted for benthic foraminifera in deep sea and benthic diatoms in lakes (Telford et al., 2006). However, marine planktonic picoeukaryote such as MAST-4 did not associate with significant distance-decay pattern, but the environmental selection (e.g., temperature) drove MAST-4 distribution (Rodríguez-Martínez et al., 2013). Another study on marine planktonic diatom assemblages also showed that these eukaryotic microbes were not limited by dispersal (Cermeno and Falkowski, 2009). Overall, these results suggest that benthic and planktonic microeukaryotes might have fundamentally different biogeographical patterns or ecological mechanisms. There are large differences in benthos compared to plankton including life environment, food-web and body structure (Schaechter, 2012). In sediments, movements of microeukaryotes are obviously more limited relative to water columns, where planktonic microbes can disperse passively to distant locations because of ocean currents (Yeh et al., 2015). These differences can lead to distinct biogeographical patterns between plankton and benthos at a regional scale.

More importantly, our results show that the rare and conditionally rare benthic subcommunities follow the similar and general biogeographical distribution patterns with dominant taxa, because all of the three subcommunities exhibited significant distance-decay patterns (**Table 1**). This could be because of dispersal limitation and/or the fact that benthic samples probably have similar environmental conditions, if they are close to each other (Supplementary Figure S5; Hanson et al., 2012). Recent study of bacteria using HTS in the lakes and reservoirs indicated that rare bacteria showed a similar spatial pattern with the abundant bacterial subcommunity (Liu et al., 2015). These results indicate that similar structuring processes (e.g., dispersal) can influence benthic dominant, rare and conditionally rare subcommunity compositions (Hanson et al., 2012). However, although there were similar distribution patterns of dominant taxa, RT, and CRT, we found that dominant taxa have a weaker distance-decay relationship (r = −0.400) than RT and CRT (r = −0.579 and r = −0.798, respectively, **Table 1**). This suggests that the benthic dominant microeukaryotic community has a higher capacity to dispersal than RT, hence leading to a more cosmopolitan distribution (Liu et al., 2015).

#### Relative Importance of Environmental and Spatial Factors Influencing Microeukaryotic Distribution

Our results indicate that both the environmental and spatial factors play significant roles in structuring the microeukaryotic metacommunity, although the relative effects of the two types of influencing factors on community variations were different (**Table 2**). The mechanisms underlying the assembly processes are important question in the field of microbial ecology (Hanson et al., 2012). The different relative effects of environmental and spatial factors influencing MCCs may be ascribed to the dominant taxa, RT, and CRT which had different community compositions and properties (Liu et al., 2017). In aquatic ecosystems, the responses of microbial communities to environmental and spatial changes are mediated by their properties; such as physiological tolerance, dispersal capacity, taxonomic and functional diversity (Gong et al., 2015; Liu et al., 2015). Our result was consistent with previous study on aquatic and sedimentary bacteria communities in 16 lakes from western China (Yang et al., 2016). However, our study was not in agreement with another study in Antarctica coastal lakes (Logares et al., 2013), which suggested that MCC was strongly influenced by environmental factors (e.g., salinity) and weakly correlated with geographical distance. This inconsistency maybe attributed to the spatial distance differences among the studied area and different environmental gradients (e.g., salinity ranged from 0 to 100 psu in coastal lakes in Antarctica vs. 12.6 to 30.9 psu in water habitat and 23.2 to 30.0 psu in sediment habitat in Xiamen) (Logares et al., 2013). In addition, our VPA confirmed that a large proportion of the observed MCCs variations could not be explained, especially for the RT (78.1–97.4% of variance unexplained) and CRT (49.0–81.0%), indicating that more complex mechanisms may generate and maintain the rare biosphere diversity in the coastal waters and intertidal sandy sediments. The unexplained variation may be due to unmeasured environmental and ecological factors, or interactions between taxa and species' own vital rate. For example, Barberán and Casamayor (2010) found that species sorting solely dominated microbial community composition. Evolutionary drift (stochastic genetic diversification), and ecological drift (stochastic processes of birth, death, colonization, and extinction) could also contribute to spatial effects and unexplained variation (Hanson et al., 2012). Further, another studies indicated that the co-occurrence correlations among microbes should also be responsible for the community structure (Lima-Mendez et al., 2015; Wei et al., 2016). These findings suggested that extending the number of sampling sites, and integrating more environmental factors, along with conducting manipulative experiments across space and time, are necessary for better understanding of the influence and mechanisms of environmental selection and stochastic processes on the biogeography of planktonic and benthic microeukaryotes

(Yeh et al., 2015). Therefore, the relative importance of spatial and environmental factors on microeukaryotic distribution awaits further investigation given the unexplained variation.

After removing sites N2 and N3, interestingly, the contribution of environmental factors to the planktonic MCC variations exhibited a rapid decline among all of total, dominant, rare, and conditionally rare planktonic taxa, whereas the contribution of spatial factors did not show significant variation. This result confirmed that environmental factors had a stronger influence on microbial biogeography than spatial factors in these 12 water sites than for the 10 water sites in this study (**Table 2**). More details of sites N2 and N3 were reported in the previous study (Yu et al., 2015).

Finally, we fitted the NM of community assembly for determining whether stochastic processes could explain variation of MCCs (**Figure 4** and Supplementary Figure S6). The NM was very powerful in explaining microbial community structure (R <sup>2</sup> = 0.76) within aquatic environments in a previous study (Roguet et al., 2015). In this study, the fit to the NM showed that stochastic processes appeared to have more influence in total benthic MCC (R <sup>2</sup> = 0.43, 12 sediment sites) compared with total planktonic MCC (R <sup>2</sup> = 0.25, 12 water sites). After removing sites N2 and N3, however, this model explained 72% of the variation in planktonic MCC (R <sup>2</sup> = 0.72, 10 water sites, **Figure 4**). These results indicate that stochastic processes contributed the most impact in planktonic MCC without sites N2 and N3 (Sloan et al., 2006), and this finding may be ascribed to high and random dispersal rate of microeukaryotes in surface water. Their movements are more restricted in sediment than in water habitat, where they can disperse randomly or passively with ocean currents at regional or local scales. The NM findings (the R 2 -value of 12 water sites was 0.25 and 12 sediment sites was 0.43, respectively) were almost fully consistent with results revealed by VPA (the community variation explained by spatial factors for plankton and benthos), although the contributions of spatial factors to the MCC variations differed between the two statistical methods. The difference may be attributed to the MCC variation explained by VPA only involve the influence caused by geographical distance of sampling sites (dispersal). However, various stochastic-related processes (drift and other mechanisms) could contribute major influence to the MCC variation in NM (Sloan et al., 2006, 2007).

#### CONCLUSION

Our data demonstrated that MCCs were significantly different between water and sediment habitats, and benthos exhibited significantly higher species richness than plankton largely due to high proportions of RT and CRT. The proportions of OTUs and sequences of Excavata were significantly higher in the benthos than in the plankton, whereas Alveolata and Diatomea had significantly higher richness and abundance in planktonic than benthic communities. Mantel tests revealed that environmental factors (e.g., salinity, TP, and TN) were significantly related to the variation of benthic MCCs. However, the lack of significant relationship between environmental factors and planktonic MCCs might indicate plankton communities were affected by more complex mechanisms (such as biotic interactions, stochastic processes and other unmeasured environmental variables). Further, benthic MCCs (including total, dominant, rare, and conditionally rare taxa) exhibited a significant distancedecay relationship, whereas the planktonic communities did not. However, more than 70% variation in planktonic community was well explained by stochastic processes when removing two unique sites (N2 and N3). Both the environmental and spatial factors play significant roles in structuring total and dominant microeukaryotic metacommunities although their relative effects on the planktonic and benthic community variations were different. However, unlike the higher explained variation of the total and dominant taxa, the high proportion of unexplained variation in the RT and CRT indicated that more complex mechanisms may influence the rare subcommunity assembly in the coastal waters and intertidal sediments. Altogether, our results indicate that patterns and processes in marine microeukaryotic community assembly may differ among the different habitats (coastal water and intertidal sediment) and different MCCs (total, dominant, rare and conditionally rare microeukaryotes).

## AUTHOR CONTRIBUTIONS

WZ conceived the idea and designed the comparison experiments. YP and LY collected the samples and performed the experiments. WC, JY, and WZ analyzed the data. JY and WZ contributed reagents/materials/analysis tools. WC, JY, and WZ wrote the paper.

#### FUNDING

This work was supported by the National Natural Science Foundation of China (41276133), the Natural Science Foundation of Fujian Province (2014J01163) and the Fundamental Research Funds for the Central Universities (20720160115).

#### ACKNOWLEDGMENTS

The authors thank Lemian Liu, Yuanyuan Xue, and Min Liu for data analysis assistance, and thank Prof. David M. Wilkinson and Dr. Alain Isabwe for constructive comments on the manuscript. The authors also thank two reviewers for insightful comments that helped in improving the clarity of this paper.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2017.01912/full#supplementary-material

## REFERENCES

fmicb-08-01912 October 12, 2017 Time: 17:34 # 13



**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 Chen, Pan, Yu, Yang and Zhang. 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.

# Distinct Seasonal Patterns of Bacterioplankton Abundance and Dominance of Phyla α-Proteobacteria and Cyanobacteria in Qinhuangdao Coastal Waters Off the Bohai Sea

#### Yaodong He<sup>1</sup>† , Biswarup Sen<sup>1</sup>† , Shuangyan Zhou<sup>1</sup> , Ningdong Xie<sup>1</sup> , Yongfeng Zhang<sup>2</sup> , Jianle Zhang<sup>2</sup> and Guangyi Wang1,3 \*

<sup>1</sup> Center for Marine Environmental Ecology, School of Environmental Science and Engineering, Tianjin University, Tianjin, China, <sup>2</sup> Qinhuangdao Marine Environmental Monitoring Central Station, State Oceanic Administration, Qinhuangdao, China, <sup>3</sup> Key Laboratory of Systems Bioengineering (Ministry of Education), Tianjin University, Tianjin, China

#### Edited by:

Hongyue Dang, Xiamen University, China

#### Reviewed by:

Mohd Ikram Ansari, King Abdullah University of Science and Technology, Saudi Arabia Jérôme Hamelin, Institut National de la Recherche Agronomique (INRA), France

\*Correspondence:

Guangyi Wang gywang@tju.edu.cn †These authors have contributed equally to this work.

#### Specialty section:

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

Received: 17 February 2017 Accepted: 03 August 2017 Published: 18 August 2017

#### Citation:

He Y, Sen B, Zhou S, Xie N, Zhang Y, Zhang J and Wang G (2017) Distinct Seasonal Patterns of Bacterioplankton Abundance and Dominance of Phyla α-Proteobacteria and Cyanobacteria in Qinhuangdao Coastal Waters Off the Bohai Sea. Front. Microbiol. 8:1579. doi: 10.3389/fmicb.2017.01579 Qinhuangdao coastal waters in northern China are heavily impacted by anthropogenic and natural activities, and we anticipate a direct influence of the impact on the bacterioplankton abundance and diversity inhabiting the adjacent coastal areas. To ascertain the anthropogenic influences, we first evaluated the seasonal abundance patterns and diversity of bacterioplankton in the coastal areas with varied levels of natural and anthropogenic activities and then analyzed the environmental factors which influenced the abundance patterns. Results indicated distinct patterns in bacterioplankton abundance across the warm and cold seasons in all stations. Total bacterial abundance in the stations ranged from 8.67 × 10<sup>4</sup> to 2.08 × 10<sup>6</sup> cells/mL and had significant (p < 0.01) positive correlation with total phosphorus (TP), which indicated TP as the key monitoring parameter for anthropogenic impact on nutrients cycling. Proteobacteria and Cyanobacteria were the most abundant phyla in the Qinhuangdao coastal waters. Redundancy analysis revealed significant (p < 0.01) influence of temperature, dissolved oxygen and chlorophyll a on the spatiotemporal abundance pattern of α-Proteobacteria and Cyanobacteria groups. Among the 19 identified bacterioplankton subgroups, α-Proteobacteria (phylum Proteobacteria) was the dominant one followed by Family II (phylum Cyanobacteria), representing 19.1– 55.2% and 2.3–54.2% of total sequences, respectively. An inverse relationship (r = −0.82) was observed between the two dominant subgroups, α-Proteobacteria and Family II. A wide range of inverse Simpson index (10.2 to 105) revealed spatial heterogeneity of bacterioplankton diversity likely resulting from the varied anthropogenic and natural influences. Overall, our results suggested that seasonal variations impose substantial influence on shaping bacterioplankton abundance patterns. In addition, the predominance of only a few cosmopolitan species in the Qinhuangdao coastal wasters was probably an indication of their competitive advantage over other bacterioplankton groups in the degradation of anthropogenic inputs. The results provided an evidence of

**89**

their ecological significance in coastal waters impacted by seasonal inputs of the natural and anthropogenic matter. In conclusion, the findings anticipate future development of effective indicators of coastal health monitoring and subsequent management strategies to control the anthropogenic inputs in the Qinhuangdao coastal waters.

Keywords: anthropogenic impacts, environmental variables, phylogenetic diversity, bacterioplankton abundance, seasonal variations, redundancy analysis

#### INTRODUCTION

fmicb-08-01579 August 16, 2017 Time: 15:58 # 2

Marine and coastal environments are of major concern nowadays on account of the rapid developments and growth in anthropogenic activities. Most commonly the anthropogenic activities include terrestrial pollution, aquaculture, urban development, tourism, maritime transport, agricultural and industrial activities, oil refineries, and mining, etc. (Islam and Tanaka, 2004). The nutrient inputs from these activities are the major drivers of eutrophication in most of the coastal areas, ultimately affecting the marine food web components (Vaquer-Sunyer et al., 2015). In addition, rivers and allochthonous biological processes also enrich coastal water with dissolved compounds (Hiriart-Baer et al., 2008; Barrera-Alba et al., 2009; Vargas et al., 2011; Gustafsson et al., 2014; Lafon et al., 2014). Thus, organic matter mineralization and nutrient cycling are crucial for maintenance of coastal ecosystem health. In this context, the assemblage and abundance of bacterioplankton involved in the marine microbial loop (Farooq and Francesca, 2007) are canonical indicators of ecosystem status and the extent of anthropogenic impact on the coastal waters (Zhang et al., 2014). The spatial and temporal dynamics of bacterioplankton community is driven by multiple environmental factors such as latitudinal gradient (Fuhrman et al., 2008; Milici et al., 2016), temperature (Kim et al., 2016), predation (Oguz et al., 2013; Karus et al., 2014), nutrients (Kim et al., 2016), succession (Chen H. et al., 2016), seasons (Gilbert et al., 2012), and timescales (Fuhrman et al., 2015). Apart from natural environmental factors, anthropogenic activities also shape the bacterioplankton assemblage and abundance pattern (Zhang et al., 2009; Thiyagarajan et al., 2010; Zhou et al., 2011; Sauret et al., 2012; Jeffries et al., 2016; Meziti et al., 2016). For instance, the abundances of the bacterioplankton groups that increased and decreased in the impacted sites were significantly correlated with nutrients enrichment (Fodelianakis et al., 2014). Also, there is a direct evidence of specific metal (Cd)-induced patterns in bacterioplankton communities in coastal systems (Wang et al., 2015). In oligotrophic coastal water microcosm with nitrate perturbation, the bacterioplankton community composition was greatly influenced by nitrate loading mode, indicative of nitrate loading impact on marine environments (Dong et al., 2017). Thus, nutrients and toxic pollutants enrichment seem to be the important drivers of the bacterioplankton composition in sites receiving direct human impact (Aguilo-Ferretjans et al., 2008).

A closer inspection of the bacterioplankton community dynamics reveals many interesting relationships and patterns in the taxonomic diversity and abundance. A high concentration of Synechococcus and low concentration of Prochlorococcus has been observed in sites impacted by human activities with inputs from the land (Bouvy et al., 2012). In a microcosm study, it was found that anthropogenic impacts do not necessarily influence the abundance pattern of rare species (<0.1% of relative abundance), and in fact, nutrients enrichment increased the relative abundance of only abundant species (>1% of relative abundance) Polaribacter, Tenacibaculum, and Rhodobacteraceae (Baltar et al., 2015). By the use of 16S rRNA gene cloning and sequencing, the bacterioplankton diversity of oligotrophic marine environments is relatively well studied than coastal environments (Zhou et al., 2011; Ortega-Retuerta et al., 2014; Hartmann et al., 2016; Milici et al., 2016; Dong et al., 2017). Bacterioplankton diversity and abundance patterns on a regional-scale and also for local sites under anthropogenic impact remains a hot topic and several reports are available (Simonato et al., 2010; Niu et al., 2011; Peierls and Paerl, 2011; Zhou et al., 2011; Fodelianakis et al., 2014; Laas et al., 2014; Almutairi, 2015; Baltar et al., 2015; Meziti et al., 2015; Xiong et al., 2015; Jeffries et al., 2016). Overall, the available reports on bacterioplankton community composition at regionalscale or in local sites indicate diverse patterns and complex relationships between species composition and environmental factors. In addition, phylogenetic analysis and abundance patterns of bacterioplankton assemblages provide significant insights into the bacterioplankton dynamics in the impacted coastal areas which further leads us to the development of coastal environmental monitoring and management strategies (Demarcq et al., 2012). This calls for more research to identify the bacterioplankton species that show strong linkages with the implicit environmental factors for developing effective descriptors of the anthropogenic impacts.

Qinhuangdao is a famous port city, an important energy export port, and coastal tourist spot on the northwest coast of the Bohai Sea in North China with an area of approximately 7812 km<sup>2</sup> and a population of 2.9 × 10<sup>6</sup> (Gu et al., 2016). The Bohai Sea coastline experiences the strong influence of riverine systems, aquaculture, municipal sewage, industrial wastewaters, and agricultural runoffs (Zhu et al., 2014). Land-use surrounding the Qinhuangdao coastal area shifted drastically from agriculture to industrialization in the last 30 years (Xu et al., 2010). The area has undergone development of several petrochemical, steel, and other projects, besides land reclamation and construction of embankments (Zhu et al., 2014). The resulting land-use change has overall influenced the coastal seawater nutrients cycling and associated bacterioplankton, phytoplankton, and other biotic factors (Xu et al., 2010; Liu et al., 2011). Results from

other and our previous studies have shown a distinct pattern of bacterioplankton abundance and diversity as a consequence of environmental influence such as estuaries, mariculture, and rainfall (Li J. et al., 2011; Li et al., 2013; Shang et al., 2016; Zhou et al., 2016). These reports further encourage us to investigate closely the relationship between the bacterioplankton community and nutrient levels across different seasons in the near-shore and off-shore stations near Qinhuangdao coast and address the following questions. What are the patterns in abundance of major taxonomic groups and their spatial diversity? What drives the spatiotemporal variations of bacterioplankton in the Qinhuangdao coastal waters?

The objectives of the present study were to: (1) analyze the abundance of total bacteria and major bacterioplankton groups across different seasons, (2) identify the key environmental factors that influence the bacterioplankton abundance, and (3) assess the bacterioplankton distribution and diversity along the near-shore and off-shore coastal areas of Qinhuangdao.

## MATERIALS AND METHODS

#### Study Area and Water Collection

Qinhuangdao, one of the highly urbanized regions, is a port city on the coast of China located in the northeastern Hebei province, and it is situated about 280 km east of Beijing on the northwest Coast of Bohai Sea, the innermost gulf of Yellow Sea. The Qinhuangdao coastal area has a humid continental climate with four distinct seasons. It has a monsoon-influenced humid continental climate with the highest precipitation (152–189 mm) during July–August period. From October month the temperature starts to fall (<5 ◦C) until April, and a very low (<10 mm) precipitation occurs during November to March period (Gu et al., 2016).

Six stations (W1–W6) were selected that had the distinct influence of the local environment (**Figure 1**). The W1 station is located outside the port of Qinhuangdao mainly polluted by the shipping industry and municipal sewage, the W3 station is located in the Xinkai river estuary near the Beidaihe Forest Wetland and Qinhuangdao Wildlife Park, and the W5 station is located between the Yang River Estuary and Dai River Estuary. The Yang River and Dai River are severely polluted by microorganisms and the total number of bacteria exceeds the standard levels. W3 and W5 stations are also located in the tourist area and near the bathing beach. W2, W4, and W6 are offshore stations with aquaculture facility near W6 station. Water samples were collected in the month of March, July, October, and December (2014) at the stations W1, W2, W3, W4, W5, W6 (**Figure 1**). For each month one water sample (1 L) at 1 m depth from the surface was collected using a sample hydrophore. A plexiglass hydrophore (JC-800, Juchuang, China) of capacity 5 L, diameter 15 cm, and height 37.5 cm was used as the sample hydrophore. All samples were stored in sterile glass bottles at 4 ◦C and then transported to the laboratory. A 500 mL volume of water sample was filtered using 0.22 µm membrane filter (AmeriTech Inc., United States). The membrane filters coated with microorganisms were stored at −80◦C.

## Water Quality and Nutrient Analysis

The pH, temperature, dissolved oxygen (DO), and salinity of the water samples at six stations were measured using a portable YSI Pro Plus Multiparameter instrument (YSI Inc., United States). The contents of total nitrogen (TN), total phosphorus (TP), ammonia nitrogen (NH4-N), nitrate nitrogen (NO3-N), nitrite nitrogen (NO2-N) and Chlorophyll a (Chl a) were determined in water samples in accordance with the "Specification for oceanographic survey" (GB/T 12763.4-2007). Briefly, NH4-N was measured by sodium bromate oxidation method which involves oxidization of ammonium salt to nitrite in alkaline medium, followed by determination of the TN content using diazo spectrophotometry. After deducting the concentration of NO2-N, the concentration of NH4-N was estimated. NO3-N was measured by zinc cadmium reduction method where nitrate is quantitatively reduced to nitrite in water by zinc cadmium reduction method following determination of total nitrite by the diazo and azo method, the original nitrite was then corrected and the nitrate content was calculated. NO2-N was measured by a diazo azo method which involves the reaction of nitrite with sulfanilamide in acidic medium following product reaction with hydrochloric acid and synthesis of a red azo dye whose absorbance was measured at 543 nm. Chl a measurement was done by an extraction-fluorimetric method which involves excitation by blue light of the acetone extract of Chl a to produce red fluorescence (Holm-Hansen and Bo, 1978). The fluorescence value of the extract was determined at 685 nm before and after acidification.

The abundance of total bacteria at the 6 stations near Qinhuangdao coastal area over warm and cold seasons were monitored using fluorescence microscopy to assess the influence of anthropogenic and natural activities in the near-shore and off-shore waters. The total bacterial count in water samples (formaldehyde-fixed) at each station was estimated by fluorescence microscopy (Eclipse Ni-U, Nikon Instruments Inc., United States). Sample (2 mL water) filtration was done at a pressure of 150 mm Hg through a polycarbonate filter (0.22 µm aperture, 25 mm diameter) (AmeriTech Inc., United States). The membrane filter was stained with 2X SYBR-Gold solution (300 µl) and was visualized at 480–485 nm wavelength (blue light excitation). Data reported are the means of counts in 20 randomly selected fields per sample. For each station, triplicate samples were processed and for each replicate sample, bacterial numbers were counted in 20 randomly selected fields. The data reported are the means of counts in triplicate samples.

#### DNA Isolation and Sequence Analysis of Bacterioplankton Assemblage

Five hundred milliliter of surface water samples collected in July 2014 from 6 stations were filtered onto 0.2 µm polycarbonate membrane filters (AmeriTech Inc., United States). The resulting filters were stored in ultra-low temperature refrigerator (−80◦C) until DNA isolation. The total DNA was extracted using E.Z.N.A.TM Water DNA Kit (Omega Bio-Tek Inc., United States) following the manufacturer's instructions. The isolated DNA was used as template for bacterial 16S rRNA gene amplification using

Frontiers in Microbiology | www.frontiersin.org

universal primers 27F (5<sup>0</sup> -AGAGTTTGATCCTGGCTCAG-3<sup>0</sup> ) and 1492R (5<sup>0</sup> -GGTTACCTTGTTACGACTT-3<sup>0</sup> ) on Bio-Rad S1000 thermal cycler (Bio-Rad Laboratories Inc., United States) using the following protocol: 95◦C 5 min; 94◦C 45 s, 55◦C 45 s, 72◦C 1 min; 35 cycles; 72◦C 10 min. PCR amplification for each sample was performed in triplicates and the triplicate PCR products were mixed to obtain the final sample for clone library construction. All PCR products were gel-purified, cloned into

fmicb-08-01579 August 16, 2017 Time: 15:58 # 4

**92**

a pMD18-T Simple Vector (Sino Biological Inc., China), and then transformed into Escherichia coli DH5α competent cells by traditional heat shock treatment. 16S rRNA clone libraries of a total of six water samples (one sample per station) for the month of July were constructed and 90 positive clones were randomly picked from each sample plate for sequence analysis (Beijing Genomics Institution, China). The coverage of the clone libraries ranged from 31.1 to 56.7% and the impact of low coverage on diversity analysis was minimized by calculating Simpson diversity estimate (Haegeman et al., 2013) in Mothur software package (Schloss et al., 2009). The sequences have been submitted to NCBI Genebank database with accession number MF498066 – MF498485.

All sequence analyses were performed within versions 1.35.1 of the Mothur software package (Schloss et al., 2009). The FASTA format sequences were aligned with SILVA alignments and the distance was calculated at cutoff 0.03 (97% sequence homology) using align.seqs and dist.seqs commands, respectively. Chimeras were identified using chimera.uchime command with default parameters. Sequences were classified using classify.seqs command with reference to Ribosomal Database Project (RDP) as the taxonomy file (silva.bacteria.rdp.tax) and the remove.lineage command was used to remove the sequences that belong to mitochondria, chloroplast, eukaryote, archaea, and unknown lineages. Cluster command was used to assign sequences to OTUs with opticlust clustering method. α-diversity was calculated by counting the number of OTUs and using the reciprocal of Simpson Index (invsimpson). The invsimpson calculator is preferred to other measures of α-diversity as it indicates the richness in a community with a uniform evenness that would have the same level of diversity. Moreover, the inverse of the Simpson index has some biological interpretation and do not tend to be affected by sampling effort because it is independent of abundance distributions (Haegeman et al., 2013).

## Quantitative PCR Analysis of Bacterioplankton Groups

To assess the phylum/class level abundance distribution of total bacteria in warm and cold seasons, qPCR experiments with phylum and class specific primers (Supplementary Table S1) were conducted for all 6 stations. qPCR method was used to detect and quantitate the abundance of the major phylogenetic groups of bacterioplankton in the water samples from Qinhuangdao coastal area. The advantage of qPCR method in our study was its ability to quantitate specific bacterioplankton groups by using phylum/class specific primers. In addition, qPCR quantitation is fast, reliable, and accurate. Based on the results of clone library analysis and identification, the following dominant classes were selected to carry out quantitative PCR experiments: α-Proteobacteria, β-Proteobacteria, Cyanobacteria, Actinobacteria, Firmicutes, and Bacteroidetes. PCR was carried out in a 20 µL reaction mixture containing 1 µL DNA template (2.8-10.5 ng/µL), 10 µL SYBR Select Master Mix (Thermo Fisher Scientific Inc.), 1 µL forward primer (10 nmol/mL), 1 µL reverse primer (10 nmol/mL), and 7 µL nuclease-free molecular-grade water. The qPCR cycling condition was: 95◦C 55 s; 95◦C 20 s, 60◦C 35 s, 72◦C 30 s; 40 cycles; 95◦C 1 min, 60◦C 1 min, heating-up from 75 to 95◦C, the increment of 0.5◦C for 8 s. The annealing temperature of each primer is shown in Supplementary Table S1. The standard curves were constructed with known amounts of plasmid DNA containing the sequence insert. For the positive plasmid preparation, the PCR products were ligated to pMD18-T Vector by pMD18-T Vector Cloning Kit (Takara Biotechnology (Dalian) Co., Ltd.) and then transformed into E. coli competent cells (DH5α) followed by a screening of positive clones. Plasmid DNA was extracted using E.Z.N.A <sup>R</sup> Plasmid Midi Kit (Omega Bio-Tek Inc., United States) and the concentration of the extracted DNA was measured using NanoDrop ND-1000 spectrophotometer (Nanodrop Technology). The gene copies were calculated using the mean mass of pMD18-T Vector ligated with the target DNA sequence. DNA of each sample and standard plasmid DNA dilution (10<sup>1</sup> and 10<sup>5</sup> ) were used as templates for qPCR. Each sample and standard were analyzed in triplicates. The standard curve was constructed to estimate the copy number of each sample according to the dilution ratio and copy number of the standard plasmid template.

## Statistical Analysis

Principal component analysis (PCA) was done to reveal the relationships between the environmental parameters and the distribution of the sampling stations in an ordination plot. Pearson's correlation analysis was performed to assess the correlation between the abundance of total bacteria, major bacterioplankton groups, and environmental parameters. Redundancy analysis (RDA) was carried out to identify the explanatory environmental parameters that significantly (p < 0.01) constrain the bacterioplankton abundance data by forward selection in Canoco version 5.02. RDA was performed using abundance (qPCR data) and environmental data. The PCA of environmental parameters was done using normalized (z-score) data in Canoco version 5.02. For RDA, the qPCR data were log(x+1) transformed and the environmental data was normalized by z-score.

## RESULTS

## High Seasonal Variations of Environmental Parameters

The environmental parameters across 6 stations near Qinhuangdao coastal area appeared highly heterogeneous spatiotemporally. In the warm season (July–October), the temperature of coastal waters in all stations was within 21–25◦C, whereas in the cold season (December–March) the temperature was significantly lower ranging from −0.3 to 6.8◦C. pH was notably higher (11.6 ± 0.09) in October at all stations. Salinity levels were similar (∼29 ppt) in all stations and also during warm and cold seasons. The DO levels were relatively higher in cold period than warm as a result of temperature effect on oxygen solubility in water. The average values of environmental parameters (nutrients) with their variation over different periods at the 6 stations near Qinhuangdao coastal area are shown in **Table 1**. The NH4-N level was high in W5 station than


TABLE 1 | The data of environmental parameters at near-shore (W1, W3, and W5) and off-shore (W2, W4, and W6) stations near Qinhuangdao coastal area.

TN, TP, and Chl a indicate total nitrogen, total phosphate, and Chlorophyll a.

Mean (SD) of values over 4 seasons (July, October, December, and March) are provided.

that of the others. In October the NH4-N was relatively high at all stations (0.05–0.17 mg/L). No significant variation in NO2-N and NO3-N content was observed at the 6 stations. TN was detected in all stations (0.35–0.6 mg/L) with relatively large fluctuations (0.6–1.8mg/L) in the warm season (July) and was lowest in the cold season (December). In July and March, the TP was found to be comparatively higher than in October and December. Furthermore, the average TP content was higher in W1, W2, W3, W5 (0.12–0.13 mg/L) than that in W4 and W6 stations (0.07–0.08 mg/L). W1 and W5 stations showed highest average Chl a level (11.6–16.0 mg/m<sup>3</sup> ) followed by W3 and W6 (5.6–5.9 mg/m<sup>3</sup> ) and lowest levels at W2 and W4 (2.6-3.2 mg/m<sup>3</sup> ). The variation in Chl a level was substantially less in W2 and W4 stations suggesting a low abundance of phytoplankton at these stations than at the polluted stations (W1, W3, W5, and W6). In the Bohai Sea, silicates and Si/dissolved inorganic N ratio have been associated with phytoplankton abundance (Chen Y.H. et al., 2016). In addition, climate and mariculture activity have a significant correlation with the Chl a trends (Fu et al., 2016). River discharge and suspended sediment also influence Chl a in Bohai Sea coast. These factors might have played a role in the spatial variation of Chl a observed in the present study. Also, Chl a mostly was higher in July in all stations probably due to more growth of phytoplankton resulting from optimal light and temperature conditions.

By analyzing all the measured environmental variables in combination, the resultant ordination plot showed distinct partitioning of coastal water samples across warm and cold seasons rather than across near-shore and off-shore stations (**Figure 2**). The ordination plot could explain 59.6% of total variation in the environmental data and revealed a linear positive correlation between TN, temperature, and Chl a, and a negative correlation of DO with these variables. These parameters were most crucial in the partitioning of samples from the warm and cold seasons. Similarly, NH4-N and NO3- N had a positive correlation with each other, and salinity, NO2-N, and TP did not show any notable correlations to any of the other variables. Overall, the PCA biplot clearly showed that seasonal variation of the measured environmental parameters was more than spatial variation. Thus, seasonal loads of anthropogenic inputs seem to have a significant impact on the nutrients conditions in the coastal areas of Qinhuangdao.

#### Bacterioplankton Abundance Patterns and Influencing Factors

The results of fluorescence microscopy based detection of total bacterial abundance in the near-shore and off-shore stations are shown in **Figure 3**. In July, all the stations mostly showed highest abundance (5.32 × 10<sup>5</sup> – 1.71 × 10<sup>6</sup> cells/mL) of bacteria compared to other months. Overall, the total bacterial abundance in each of the 6 stations over the 4 months was in between 8.67 × 10<sup>4</sup> and 2.08 × 10<sup>6</sup> cells/mL. Within the other 3 months, March recorded higher abundance than October and December in all the stations. The stations W1, W3, W5, and W6 typically had relatively more abundance than the W2 and W4 perhaps due to nutrients overload in these stations as a consequence of higher anthropogenic and natural activities as evident from their higher level of Chl a. A significant (p < 0.01) positive correlation of total bacterial abundance with TP was seen (**Figure 4**).

The qPCR based abundance estimates of 6 major phylum/class i.e., α-Proteobacteria, β-Proteobacteria, Cyanobacteria, Actinobacteria, Firmicutes, and Bacteroidetes are shown in **Figure 5**. α-Proteobacteria, Actinobacteria, Firmicutes, and Bacteroidetes showed a fairly similar pattern of abundance distribution in the 6 stations. Notably, in the warm season, these 4 groups exhibited relatively higher abundances than in cold season. However, β-Proteobacteria and Cyanobacteria showed distinct trends, wherein the former one showed a similar pattern in all stations over the 4 months, i.e., higher abundance in cold than warm season. Bacteroidetes abundance at W1 and W6 stations remained high in October, and in W5 its abundance increased only in July and December. Near-shore and off-shore differences in the abundance of the major phylotypes were also apparent. In the warm season, near-shore stations possessed more Cyanobacteria and Bacteroidetes than off-shore stations. Alternatively, in cold months, near-shore stations possessed a relatively higher abundance of α-Proteobacteria, Firmicutes, Actinobacteria, and β-Proteobacteria than off-shore stations.

**Table 2** shows the Pearson's correlation coefficients for the relationship between bacterioplankton abundance and some environmental parameters. α-Proteobacteria showed significant (p < 0.05) positive correlation with TN, Chl a, DO, and temperature. Furthermore, from the ordination plot (**Figure 6**) based on redundancy analysis, temperature and Chl a were the major factors that positively correlated with the first principal component and were the significant explanatory parameters (p < 0.01) that constrained the abundance of α-Proteobacteria

and Cyanobacteria phylotypes. On the contrary, DO showed a negative influence (p < 0.01) and salinity was non-significant in explaining any variation of the abundance of bacterioplankton groups. Overall, samples collected in warm July or October had a much higher abundance of α-Proteobacteria, Firmicutes, and Actinobacteria than that of cold March or December. In contrast, β-Proteobacteria, Cyanobacteria, and Bacteroidetes did not show any distinct seasonal influence which likely suggests their diverse metabolic potential and ability to utilize organic matter even at low concentrations that allow them to attain invariable growth throughout the year.

## Spatial Distribution and Diversity of Bacterioplankton Assemblages

The assemblage of bacterioplankton communities featuring diversity and distribution at the 6 stations in the warm season was assessed and the results are reported in **Table 3** and **Figure 7**. Based on the environmental data, July exhibited greater levels and fluctuations in the measured parameters perhaps due to more nutrient inputs that resulted from anthropogenic and natural activities in warm season. Thus, further characterization of bacterioplankton phylogeny, composition, and diversity was conducted on water samples collected in July. Out of the selected clones analyzed, 62, 62, 68, 44, 47, and 71 OTUs were identified in W1, W2, W3, W4, W5, and W6 stations, respectively (**Table 3**). A broad phylogenetic distribution of the bacterioplankton community at the 6 stations in warm season is shown in **Figure 7A**. The analysis of 16S rRNA gene sequences from these stations revealed 7 distinct phyla (**Figure 7A**). The majority of the sequences generated from the 6 bacterioplankton clone libraries were affiliated to phyla Proteobacteria (21.7–75.9%), Cyanobacteria (4.6–54.2%), Bacteroidetes (9–21.2%), Actinobacteria (1.1–5.6%), Firmicutes

FIGURE 3 | Fluorescence microscopy based quantification of total bacteria at the near-shore and off-shore stations near the Qinhuangdao coastal area of Bohai Sea. 'W1' to 'W6' indicate the different sampling stations.

(0–4.4%), Verrucomicrobia (0–3.5%), Planctomycetes (0–3.3%). The unclassified group represented sequences in the range of 9.2–26.7%. At the phylum level Proteobacteria (49% of total sequences), Cyanobacteria (24% of total sequences), and Bacteroidetes (17% of total sequences) were the dominant groups in all the stations. There were 19 bacterioplankton classes in the total sequences acquired from the coastal waters of Qinhuangdao (**Figure 7B**). The predominant phylum Proteobacteria comprised of five classes namely α-Proteobacteria, γ-Proteobacteria, ε-Proteobacteria, β-Proteobacteria, and δ-Proteobacteria accounting to ca. 82, 9, 3.6, 2, and 1.6% of Proteobacteria, respectively. Further taxonomic resolution of the assemblages revealed nearly 30 different families that were assigned to the identified OTUs from the bacterioplankton assemblages (Supplementary Figure S1). The OTUs within Proteobacteria phylum were assigned to 15 families: Campylobacteraceae, Comamonadaceae, Methylophilaceae, Pseudobacteriovoracaceae, Burkholderiaceae, Bacteriovoracaceae, Moraxellaceae, Halomonadaceae, Oceanospirillaceae, Enterobacteriaceae, Rhodospirillaceae, SAR11, Hyphomicrobiaceae, Rhodobiaceae, and Rhodobacteraceae. The second major phylum Cyanobacteria comprised OTUs which were all assigned to a single family — Family II. The OTUs representing phylum Bacteroidetes belonged to 4 families: Flavobacteriaceae, Cryomorphaceae, Saprospiraceae, and Chitinophagaceae. Rhodobacteraceae represented major proportion (>10% of total 16S r RNA sequences) in all the stations followed by Family II especially in W1 and W4 stations. Only three OTUs assigned to Rhodobacteraceae, Family II, and Cryomorphaceae were cosmopolitans, i.e., they were found in all the stations. Overall, α-Proteobacteria (Rhodobacteraceae) (19.1– 55.2%) and Cyanobacteria (Family II) (2.3–54.2%) were the most dominant cosmopolitan groups. Interestingly, it was noted that these two dominant groups had an inverse relationship (r = −0.82).

Based on inverse Simpson index (1/D), the ascending order of diversity at the 6 stations was W5 < W4 < W1 < W6 < W2 < W3 ranging from 10.2 to 105 (**Table 3**). The results suggested a wide difference in the bacterioplankton diversity, which likely resulted from the presumed shifts in the environmental conditions of the coastal waters of Qinhuangdao. The inverse Simpson index (1/D) may serve as a sensitive indicator of the anthropogenic impact in the Bohai Sea and can accurately monitor the extent of human activities in Qinhuangdao coastal region.

#### DISCUSSION

#### Spatiotemporal Variations and Factors Influencing Bacterioplankton Abundance

Most global-scale studies on the biogeography of bacterioplankton communities have revealed intriguing findings and have largely shown that diversity and abundance pattern are closely linked to latitudinal gradient (Fuhrman et al., 2008), seasons (Cuevas et al., 2004; Fuhrman et al., 2015), temperature and nutrient levels (Pommier et al., 2007; Vichi et al., 2007). In addition, previous studies have well established that bacterioplankton abundance is potentially linked to elevated anthropogenic nutrient loads in aquatic ecosystems. Overall, nutrient contents (TP, TN, Chl a, and DO) and gradient of the natural parameter (temperature) displayed significant differences across seasons and sampling stations in our study. In the warm season, we observed a greater concentration of nutrients (TN, TP, and Chl a) and decreased DO which is consistent with other studies (Lee et al., 2011; Shang et al.,

2016). The heavy rainfall in the warm season seems one of the main causes which overload organic matter and pollutants through the estuaries and run-offs. Freshwater inflows and aquaculture activities also deliver nutrients and contaminants besides nutrients from fertilizers and anthropogenic activities, which may also cause the nutrient concentration variations, and shape the bacterioplankton abundance pattern (Jeffries et al., 2016). Generally, low salinity in coastal water indicates that the

area has received river discharge, however, in the present study any significant lowering of salinity in the coastal waters was not observed. This perhaps is an indication that Qinhuangdao coastal area does not experience the heavy impact of riverine system round the year. In addition to nutrient inputs, the direct input of pathogenic and fastidious microorganisms from wastewater treatment, wildlife sanctuary, and sewage sources, which are known to influence the bacterioplankton abundance


∗∗P < 0.01, <sup>∗</sup>P < 0.05; NS, non-significant.

The values represents correlation coefficient (significant bilateral) for sample number N = 24. Total bacteria and bacterioplankton group abundance data were based on fluorescence microscopy and quantitative PCR, respectively.

phylotypes, gray arrows represent environmental parameters, stars represent near-shore samples and circles represent off-shore samples. Different colors of the star or circle symbols represent different sampling time, with March samples in green, July samples in red, October samples in yellow and December samples in purple.

(Vandewalle et al., 2012; Liu et al., 2015) also might affect bacterioplankton abundance and diversity in Qinhuangdao coastal waters.

Our results demonstrated that bacterioplankton abundance in the Qinhuangdao coastal waters exhibit substantial shifts over time and relatively less in space. Furthermore, the

TABLE 3 | The number of 16S rRNA sequences, operational taxonomic units (OTUs), and inverse Simpson diversity (1/D) estimates for the bacterioplankton clone libraries.


W1–W6 represents the 6 stations near Qinhuangdao coastal area for which bacterioplankton clone libraries were constructed and analyzed.

strong correlation in total phosphorous and abundance agree well with previous findings that describe inorganic phosphorous as an essential nutrient and that it serves a vital role in cellular energy storage and transformation and is also a growth-limiting nutrient (Brandsma, 2016). Although bacterioplankton and phytoplankton abundance (Chl a) in many studies are closely linked, our study show no relationship between them, an uncoupling phenomenon often observed when inorganic phosphorus is available (Robarts and Carr, 2009). The considerable reduction in bacterioplankton abundance in October perhaps is an indication of anthropogenic inputs or short-term physical forcing events that bring in nutrients which offer phytoplankton an opportunity to outgrow bacterioplankton. The higher abundance of bacterioplankton in warm July and cold March is probably associated with dissolved organic carbon of algal exudates from algal bloom and allochthonous carbon, respectively. Overall, this study extends upon the mechanism of bacterioplankton dynamics and explains the consistency of abundance pattern with the utilization of specific substrates. Nevertheless, the tight connection of abundance with TP bespeaks high implication in coastal pollution monitoring.

Bacterioplankton abundance in the near-shore area is larger than that in the open ocean, especially in the area of the Gulf and the estuarine area. In our study, generally the nearshore abundance was higher than off-shore, perhaps caused by the uneven distribution of organic matter in the coastal stations that are affected directly by run-offs, sewage, human activities, etc. than off-shore stations. As the seawater gets polluted by industrial wastewater and/or domestic sewage, the bacterioplankton may go in a state of unstable nutrition which subsequently affects the total bacterioplankton abundance, transiently. In our study, the abundance was relatively higher in the stations adjacent to river estuaries than that of other stations, while within the near-shore stations, the difference was not significant. Interestingly, in station W5 the bacterioplankton abundance fluctuated dramatically that possibly indicated a strong influence of nearby river estuaries heavily polluted with invasive microorganisms. The results showed that the total abundance in the stations near Qinhuangdao coastal area fluctuated between 8.67 × 10<sup>4</sup> (W1 in March) to 2.08 × 10<sup>6</sup> (W4 in October) cells/ml which is in good agreement with the estimate projected by Whitman et al. (Whitman et al., 1998) and empirical data (Al-Rifaie et al., 2008; Robarts and Carr, 2009; Li H.-Y. et al., 2011).

Although total abundance pattern provides an indication of anthropogenic impacts on coastal waters to a certain extent, knowing the pattern of variation in the composition of bacterioplankton allows further understanding of how the growth of different groups of bacterioplankton is controlled by sporadic nutrient inputs. In addition, different phylogenetic groups of bacterioplankton play a specialized role in their consumption of nutrients, and thus dissecting their abundance fluctuations would provide greater insight into the extent of nutrient loading and their recycling by functional groups. The distinct pattern of seasonal variations in the abundance of α-Proteobacteria, Actinobacteria, Firmicutes, and Bacteroidetes groups most likely explains their role in organic matter transformation as chemoheterotrophs. These groups exhibited higher abundance mostly in warm season when there were intense anthropogenic activities and subsequent enrichment of seawater with pollutants. In fact, it was shown that warming promotes bacterioplankton to benefit from the phytoplankton bloom and increase their abundance ultimately shifting toward a more heterotrophic system (Scheibner et al., 2014). These groups perhaps are directly involved in the decomposition of anthropogenic chemicals and algal exudates via a microbial loop as evident from their strong relationship with TN and Chl a. In contrast, β-Proteobacteria was found to be tightly coupled with Chl a, which probably indicated that this less known bacterioplankton group feeds mostly on the algal exudates and may not be directly involved in the decomposition of complex polymers. Bacteroidetes also take part in organic matter degradation and have been associated with the occurrence of algal bloom. Some members of this group use proteorhodopsin to capture light energy for growth making them photoheterotrophic (Gomez-Consarnau et al., 2007). The strong correlation between Bacteroidetes group and temperature could likely be a direct upshot of more light intensity in the warm season.

In our study, temperature, TN, Chl a, and DO were identified as key drivers of spatiotemporal dissimilarity between bacterioplankton abundance patterns in the coastal waters of Qinhuangdao area. Salinity, temperature, and pH are naturally varying parameters, while nutrient concentrations and DO levels indicate the extent of anthropogenic activities besides allochthonous inputs. Temperature and salinity are strong drivers of variations in microbial assemblages of estuaries (Jeffries et al., 2016), but in coastal water salinity changes are marginal, and therefore non-significant in explaining community variations. Previous studies have reported strong association of TN, Chl a, temperature, and TP with the microbial community composition

variations and perhaps are the major drivers of community dynamics in aquatic ecosystems (Fodelianakis et al., 2014; Wang et al., 2015; Jeffries et al., 2016; Dong et al., 2017). However, in the present study, these key parameters were tenuous in partitioning the near-shore and off-shore waters near Qinhuangdao area, perhaps indicating the influence of other factors such as Si/dissolved inorganic N ratio, river discharge, short-term physical forcing events, and suspended sediments. In conclusion, this study shows temporal variations in the environmental parameters were more pronounced than spatial in shaping the abundance patterns of bacterioplankton inhabiting the coastal waters of Qinhuangdao. This further connotes the consequence of sporadic anthropogenic activities and their chemical inputs in influencing the bacterioplankton assemblages in the harbor area. Our study provided the first insight into the bacterioplankton ecology of an anthropogenically impacted regional-scale coastal water of Qinhuangdao, by combining bacterioplankton abundance patterns with physicochemical characteristics.

#### Spatial Patterns of Bacterioplankton Composition and Diversity

The regional and global ocean sampling expeditions have highlighted the importance of bacterioplankton community dynamics and their assembly into functional communities by elucidating their community structure inhabiting different marine locations (Yutin et al., 2007; Lauro et al., 2014). Yet, the information on bacterioplankton diversity within littoral areas adjoining harbors is quite limited perhaps due to complex interactions of multitudinous factors that underline their structure and distribution. As far as the littoral area of Bohai Sea is concerned, this is one of the first studies to our knowledge which reports the diversity of the bacterioplankton assemblage nearby harbor and the factors that shape their structure and composition. Proteobacteria, Cyanobacteria, Actinobacteria, and Bacteroidetes represented the major phylotypes at the 6 stations, and is in agreement with other reports which have found these bacterioplankton groups to be the major inhabitant of marine habitats (Li J. et al., 2011; Li et al., 2013; Smedile et al., 2014; Xiong et al., 2015; Jeffries et al., 2016). Proteobacteria and Cyanobacteria exhibited higher dynamics in their composition among others. Although it is quite challenging to relate phylogenetic diversity and ecological function in marine microbial diversity, yet it is hypothesized that these major groups efficiently metabolize exudates and dissolved organic matter composed of carbohydrates and/or amino acids derived from both growing and senescent phytoplankton (Teira et al., 2008). In addition, these groups are also known to degrade hydrocarbons and inorganic compounds from allochthonous sources (Robarts and Carr, 2009).

Within phylum Proteobacteria, the class α-Proteobacteria represented the majority of the 16S rRNA gene sequences and was the predominant group in all stations. α-proteobacteria are usually abundant in coastal waters and they can form the predominant surface- and particle-colonizing group (Dang and Lovell, 2002a,b). Rhodobacteraceae (also called the marine Roseobacter clade) is usually the largest subgroup in the α-proteobacteria and they are the pioneering and dominant microorganisms on submerged surfaces and organic particles (Dang and Lovell, 2000; Dang and Lovell, 2016). The class γ-Proteobacteria constituted only a minor fraction in our molecular-based study, although members of this class are best known for their rapid growth and also represented as a major group in global surveys (Pommier et al., 2007). OTUs assigned to α-Proteobacteria and Cyanobacteria phyla were cosmopolitans in our study, and this finding is consistent with previous reports (Pommier et al., 2007; Smedile et al., 2014). Cyanobacteria dominated at W1 and W4 stations and most OTUs belonged to the genus Synechococcus. They are mainly free-living and are the main participants and contributors to primary productivity of the global carbon cycle and are the common bacteria inhabiting coastal, temperate, and cold environments (Brown et al., 2014; Xia et al., 2015). From genome analysis, Synechococcus sp. WH8102 was found to be able to use some new organic compounds as sources of nitrogen and phosphorus (Palenik et al., 2003). In addition, Cyanobacteria (Synechococcus) are typically more abundant in higher nutrients or dynamic conditions and are often described as generalist better suited to exploit the fluctuating environments (Palenik et al., 2003). In contrast, Rhodobacteraceae comprise mainly aerobic photo- and chemo-heterotrophs besides purple non-sulfur bacteria which perform photosynthesis in anaerobic environments. Members of this subgroup are metabolically heterogeneous and are capable of anoxygenic phototrophy, aromatic and organosulfur degradation (Lenk et al., 2012). They are the main groups involved in the demethylation of dimethylsulfoniopropionate in the water column (Curson et al., 2011). They are largely involved in sulfur and carbon biogeochemical cycling and symbiosis with other micro- and macro-organisms (Pujalte et al., 2014). The dominance of Rhodobacteraceae in the coastal waters of Qinhuangdao port area epitomizes the degradation of organic and inorganic compounds containing sulfur. The inverse relationship between the two most predominant taxonomic groups (Rhodobacteraceae and Family II) may be either due to competition, niche partitioning, or another mechanism. Further investigation is needed to ascertain the exact cause of the inverse relationship. The low proportion of γ- and δ-Proteobacteria at the 6 stations perhaps indicate the absence of chronic polyaromatic hydrocarbons (PAHs) influence in the coastal waters near Qinhuangdao port, because these groups are often seen associated with high levels of chronic PAHs (Jeanbille et al., 2016). Moreover, the SAR11 considered as highly diverse and abundant at different depths and across habitat types within euphotic and mesopelagic zones was detected only in few stations (W1, W2, and W5). The absence of SAR11 clade at other stations perhaps indicates their rarity and thus this clade was unable to represent in our clone libraries. Nevertheless, the use of inverse Simpson Index which does not tend to be affected by rare species could predict diversity more accurately than other estimators of diversity, and thus diversity analysis in our study was relatively robust and could give a reliable and meaningful interpretation. The wide range and high variation in diversity estimate (1/D) indicate environmental heterogeneity

of the Qinhuangdao coastal waters due to varied nutrient types and their concentrations that directly influence the relative abundance of bacterioplankton functional groups.

In the present study, the richness (species number) of the bacterioplankton community was not estimated taking into account the high rate of estimation error with low coverage libraries (Haegeman et al., 2013). Instead, we implemented the estimation of Simpson diversity index which is a robust estimator and provides an accurate estimation of the species diversity. It is independent of any species abundance distributions (SADs) unlike the richness estimators (Haegeman et al., 2013). The use of Simpson index in our study assured accurate estimation of bacterioplankton diversity even with low coverage clone libraries. Also, as explained in an earlier study, coverage does not correlate with library size and the authors clearly mentioned that "a large library size did not guarantee a stable estimate of phylotype richness, nor was a small library necessarily too small" (Kemp and Aller, 2004). The authors also showed that the average size of 220 bacterial libraries was 81 ± 10, ranging in size from 5 to 417 clones. Therefore, our study with 90 clones per library is in good agreement with the average size 81 ± 10 (Kemp and Aller, 2004).

Overall, our study on the bacterioplankton diversity at stations near Qinhuangdao port suggests prominence of only two cosmopolitan OTUs with several endemic OTUs. The predominance of few species is probably an indication of selective advantage of these species to overcome a load of chemical inputs despite scatting grazing and predation by virus and macro-organisms. Also, the effect of anthropogenic and allochthonous inputs was evident from the wide range of inverse Simpson diversity estimates for bacterioplankton assemblages at the stations near Qinhuangdao port. This is consistent with the notion that disturbed marine and coastal environments possess

#### REFERENCES


high variations of diversity as a result of complex and unstable nutrient resources, in contrary to oligotrophic marine habitats and deep oceans. The overall findings of this study are expected to serve the basis for Qinhuangdao coastal water monitoring and development of strategies for anthropogenically impacted coastal areas.

#### AUTHOR CONTRIBUTIONS

YH has contributed to data acquisition, analysis and drafting the work. BS has contributed to data interpretation, analysis, drafting the work and critical revision. SZ, YZ, and JZ have contributed to data acquisition. NX helped in data analysis. GW contributed to conception of the work, data interpretation and manuscript revision. All authors have final approval to the published version and accountable for all aspects of the work.

#### ACKNOWLEDGMENTS

The work was partially supported by NSFC (grant # 31670044 and 31602185) and National Marine Public Welfare Industry Special Scientific Research Project (201305022). The views expressed herein are those of the authors and do not represent the views of the funding agencies or any of its subagencies.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.01579/full#supplementary-material



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

Copyright © 2017 He, Sen, Zhou, Xie, Zhang, Zhang and Wang. 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.

# Long-Term Survey Is Necessary to Reveal Various Shifts of Microbial Composition in Corals

Shan-Hua Yang<sup>1</sup>† , Ching-Hung Tseng<sup>2</sup>† , Chang-Rung Huang<sup>1</sup>‡ , Chung-Pin Chen<sup>3</sup> , Kshitij Tandon1,4,5, Sonny T. M. Lee<sup>6</sup> , Pei-Wen Chiang<sup>1</sup> , Jia-Ho Shiu1,7,8, Chaolun A. Chen<sup>1</sup> and Sen-Lin Tang<sup>1</sup> \*

<sup>1</sup> Biodiversity Research Center, Academia Sinica, Taipei, Taiwan, <sup>2</sup> Germark Biotechnology Co., Ltd., Taichung, Taiwan, <sup>3</sup> Yourgene Bioscience, New Taipei City, Taiwan, <sup>4</sup> Bioinformatics Program, Institute of Information Science, Taiwan International Graduate Program, Academia Sinica, Taipei, Taiwan, <sup>5</sup> Institute of Bioinformatics and Structural Biology, National Tsing Hua University, Hsinchu, Taiwan, <sup>6</sup> Section of Gastroenterology, Hepatology and Nutrition, Department of Medicine, University of Chicago Medicine, Chicago, IL, United States, <sup>7</sup> Molecular and Biological Agricultural Sciences Program, Taiwan International Graduate Program, Academia Sinica, Taipei, Taiwan, <sup>8</sup> Graduate Institute of Biotechnology, National Chung Hsing University, Taichung, Taiwan

#### Edited by:

Hongyue Dang, Xiamen University, China

#### Reviewed by:

Luisa I. Falcon, National Autonomous University of Mexico, Mexico Fabiano Thompson, Federal University of Rio de Janeiro, Brazil

#### \*Correspondence:

Sen-Lin Tang sltang@gate.sinica.edu.tw †These authors have contributed equally to this work and have shared co-first authorship.

‡Deceased 17 June, 2016

#### Specialty section:

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

Received: 06 March 2017 Accepted: 30 May 2017 Published: 13 June 2017

#### Citation:

Yang S-H, Tseng C-H, Huang C-R, Chen C-P, Tandon K, Lee STM, Chiang P-W, Shiu J-H, Chen CA and Tang S-L (2017) Long-Term Survey Is Necessary to Reveal Various Shifts of Microbial Composition in Corals. Front. Microbiol. 8:1094. doi: 10.3389/fmicb.2017.01094 The coral holobiont is the assemblage of coral host and its microbial symbionts, which functions as a unit and is responsive to host species and environmental factors. Although monitoring surveys have been done to determine bacteria associated with coral, none have persisted for >1 year. Therefore, potential variations in minor or dominant community members that occur over extended intervals have not been characterized. In this study, 16S rRNA gene amplicon pyrosequencing was used to investigate the relationship between bacterial communities in healthy Stylophora pistillata in tropical and subtropical Taiwan over 2 years, apparently one of the longest surveys of coral-associated microbes. Dominant bacterial genera in S. pistillata had disparate changes in different geographical setups, whereas the constitution of minor bacteria fluctuated in abundance over time. We concluded that dominant bacteria (Acinetobacter, Propionibacterium, and Pseudomonas) were stable in composition, regardless of seasonal and geographical variations, whereas Endozoicomonas had a geographical preference. In addition, by combining current data with previous studies, we concluded that a minor bacteria symbiont, Ralstonia, was a keystone species in coral. Finally, we concluded that long-term surveys for coral microbial communities were necessary to detect compositional shifts, especially for minor bacterial members in corals.

Keywords: long-term survey, microbial composition, corals, Stylophora pistillata, shifts

## INTRODUCTION

Corals harbor a variety of microorganisms, including algae, fungi, bacteria, archaea, and viruses (Knowlton and Rohwer, 2003; Wegley et al., 2007; Marhaver et al., 2008). Coral-associated microbes, which live in coral mucus, tissue, and skeleton, have various interactions with their host and maintain coral holobiont function, including antimicrobial defense and nutrient acquisition (Bourne et al., 2016). With advances in high-throughput sequencing technologies, there is growing

**104**

research interest in microbial ecology in rare biospheres (Lynch and Neufeld, 2015), characterizing genetic and functional diversity supporting global ecosystem health (Yachi and Loreau, 1999; Elshahed et al., 2008).

Composition of coral-associated microorganisms is affected by coral species (Littman et al., 2009; Reis et al., 2009; Kvennefors et al., 2010; Lee et al., 2012; Lema et al., 2012; McKew et al., 2012; Morrow et al., 2012; Carlos et al., 2013), environmental factors (Hong et al., 2009; Ceh et al., 2011; Chen et al., 2011), geographical differences (Hong et al., 2009; Littman et al., 2009; Kvennefors et al., 2010; Lee et al., 2012; Morrow et al., 2012; Lema et al., 2014), and human activities (Klaus et al., 2007). According to previous studies, some microbes had host specificity, regardless of geographical difference (Rohwer et al., 2001, 2002), whereas others had regional specificity rather than host types (Littman et al., 2009).

Various coral-associated microbial species reportedly shared similar functional roles in coral ecology (Li et al., 2013). However, bacterial succession in corals has not been well characterized, as the longest monitoring survey lasted only 1 year (Rohwer et al., 2002; Hong et al., 2009; Ceh et al., 2011; Carlos et al., 2013; Lema et al., 2014; Hernandez-Agreda et al., 2016). Furthermore, although some microbes were consistently detected in coral (Rohwer et al., 2002; Carlos et al., 2013), it is still unknown whether these common microbes change their abundance over longer intervals. A 2-years survey characterizing dynamics of Symbiodinium rare biosphere provided evidence for symbiont switching in reef-building corals after environmental changes (Boulotte et al., 2016), but dynamics of the bacterial rare biosphere in coral, and bacterial shifting/switching within corals, are not established.

The objective was to determine whether a long-term survey was necessary to detect compositional dynamics of coral-associated bacteria in various locations. Based on previous short-term studies, we hypothesized that coral-associated bacteria require varying time scales to manifest detectable variations in different environments. Cumulative stressors, e.g., nutrient availability and elevated sea surface temperature, have been postulated to change coral reef ecosystems; therefore, coral-associated microbes might either attenuate or exacerbate effects from stressors through their roles in coral holobiont fitness (Bourne et al., 2008, 2016). Since cumulative stressors vary from place to place, duration of detectable changes in coral-associated microbes is also likely to vary, which should be apparent in a long-term survey (Mouchka et al., 2010; Ceh et al., 2011; Chen et al., 2011; Lee et al., 2012).

To determine dynamics of coral-associated microbial communities among locations and to study temporal variations of dominant and minor bacterial groups, 16S rRNA gene amplicon pyrosequencing was used in a 2-years observational study investigating bacterial communities in coral Stylophora pistillata, which is widely distributed in Taiwan and around the globe. That Taiwan lies across the Tropic of Cancer, S. pistillata in tropical (Kenting and Lyudao) and subtropical (Yehliu) Taiwan were collected to provide a comparative basis for latitude influence on bacterial composition in corals. This 2-years survey was to assess coral-associated microbial differences among various climate zones, and determine importance of long-term observation to detect bacterial dynamics in coral hosts.

## MATERIALS AND METHODS

## Coral Samples

Coral and seawater samples were collected from three locations, Kenting, Yehliu, and Lyudao, in Taiwan (**Figure 1**) from February 2008 to January 2010. In each location and at each sampling time, three colonies (replicates) of S. pistillata at depths of 3–7 m were selected and tagged for sampling. For each colony, a coral fragment approximately 2 cm long and 2 cm in diameter was collected. Samples were washed twice with filtered seawater and stored at 4◦C until coral tissue was processed. In addition, 1 L of seawater surrounding the coral colony was simultaneously collected to analyze the microbial community in reef water. A total of 88 S. pistillata and seawater samples were collected (Supplementary Table S1). Environmental data (precipitation, daylight duration and air and water temperatures) were obtained from the Central Weather Bureau in Taiwan (Supplementary Table S1).

Coral tissues were sprayed (using an airbrush) with 2–5 mL 10× TE buffer (10 mM Tris–HCl, pH 8.5, 1 mM EDTA, pH 8.0) and collected in a sterilized bag. Coral tissues were transferred to a 15 mL sterilized tube and stored at −20◦C pending DNA extraction.

## DNA Extraction, Primers, PCR, and Sequencing

Coral tissue was recovered for genomic DNA extraction by centrifugation (10,000 × g, 15 min). After discarding the supernatant, tissue pellets were put into 600 µL of 10× TE buffer and liquid nitrogen, homogenized (mortar and pestle), and the resulting solution transferred to a clean tube. Seawater samples were filtered through membrane filters (0.2 µm pore size; Adventec, Tokyo, Japan) to collect microbial particulate. Total genomic DNA of coral tissue and seawater were extracted as outlined in the UltraClean Soil DNA Kit (MioBio, Solana Beach, CA, United States).

The PCR was performed using two bacterial universal primers: 27F (5<sup>0</sup> -AGAGTTTGATCMTGGCTCAG-3<sup>0</sup> ) and 341R (50 -CTGCTGCCTCCCGTAGG-3<sup>0</sup> ), which were designed for the V1–V2 hypervariable region of the bacterial 16S rRNA gene (Hamp et al., 2009). The reaction mixture contained 1 µL of 5 U TaKaRa Ex Taq HS (Takara Bio, Otsu, Japan), 5 µL of 10× Ex Taq buffer, 4 µL of 2.5 mM deoxynucleotide triphosphate mixture, 1 µL of each primer (10 µM), and 1–5 µL (10–20 ng) template DNA in a volume of 50 µL. The temperature program for 30 PCR cycles was the initial step of 94◦C for 3 min, 94◦C for 20 s, 52◦C for 20 s, 72◦C for 20 s, and 72◦C for 2 min (final extension after the last cycle).

The PCR amplicons of the bacterial 16S rRNA gene V1–V2 region were verified by DNA agarose gel electrophoresis with 1.5% agarose gel and 10× TE buffer. The expected DNA band (∼320 bp) was cut from the gel and DNA was recovered with a QIAEX II Agarose Gel Extraction kit (QIAGEN,

Hilden Germany) and quality verified with a Scandrop spectrophotometer (Thermo Scientific, Vantaa, Finland).

A DNA tagging PCR was used to tag each PCR product of the bacterial 16S rRNA gene V1–V2 region (Chen et al., 2011). According to Chen et al. (2011), the tag primer was designed with four overhanging nucleotides; this arrangement ensured 256 distinct tags, at the 5<sup>0</sup> end of the 27F and 341R primers for bacterial DNA. The tagging PCR conditions consisted of an initial step of 94◦C for 3 min, 5 cycles of 94◦C for 20 s, 60◦C for 15 s, 72◦C for 20 s, and a final step at 72◦C for 2 min.

Pooled 40 ng lots of each tagged bacterial 16S rRNA gene V1–V2 region DNA samples (88 samples in total) were used for massively parallel pyrosequencing using a Roche 454 Genome Sequencer FLX System (Mission Biotech, Taipei, Taiwan). Raw sequencing reads were sorted according to unique tags using an in-house script, and deposited in NCBI Sequence Read Archive database, under accession number SRP098878, after the primer was removed. Reads from replicate colonies, sampled at the same time and location, were combined into one sample.

## Data Analysis

On a per-sample basis, bacterial reads were quality-filtered using Mothur v1.36.1 (Schloss et al., 2009) to retain reads of lengths from 290 to 340 base pairs (bp), and an average quality score >27. Reads containing any ambiguous base or homopolymer >8 bp were excluded. Chimeric reads were detected and removed using USEARCH v8.1.1861 (Edgar, 2010) with reference mode (3% minimum divergence). Quality-filtered and non-chimeric reads were analyzed using UPARSE pipeline (Edgar, 2013) to generate OTUs per sample (97% identity level), which were classified with RDP classifier (v2.12) (Wang et al., 2007) with a pseudo-bootstrap threshold of 80%.

The rarefied diversity measures (e.g., Shannon, Simpson, Chao 1, ACE, and Good's coverage) were estimated using Mothur with 1000 iterations. The sample size (N) varied from 567 to 29819; therefore, 567 was chosen for rarefaction analysis.

#### Cross-Sample OTU Identification of Acinetobacter and Endozoicomonas

Bacterial OTU representative sequences of all samples affiliated with a given taxonomy were collectively clustered using USEARCH, with options –cluster\_smallmem and –id 0.97. Abundance of each centroid (or cross-sample OTU) was represented by average relative abundance of its cluster members. Phylogenetic analysis was conducted in MEGA7 (Kumar et al., 2016), with maximum-likelihood trees generated with 1000 bootstraps for aligned representative sequences.

#### Resident Types of Bacterial Lineages

The present study defined six resident types of bacteria in S. pistillata and seawater, based on relative abundance and detection frequency. Type I (prevalent abundant group), type II (moderately prevalent abundant group), and type III (non-prevalent abundant group) each had ≥1% average relative abundance and were detected in ≥80, 80–60, and <60% samples, respectively. Type IV (prevalent minor group), type V (moderately prevalent minor group), and type VI (non-prevalent minor group) had <1% average relative abundance and were detected in ≥80, 80–60, and <60% samples, respectively. Coral and seawater samples were separately considered (in terms of detection frequency).

#### Clustering Analysis of Bacterial Taxa

Relative abundance of each classified bacterial genera in individual samples was incorporated into a matrix to estimate the Bray–Curtis distance. Results were presented by non-metric multidimensional scaling (nMDS) analysis using Primer 6 (PRIMER-E, Lutton, Lvybridge, United Kingdom; Clarke and Warwick, 2001) to determine relationships of bacterial communities among times and locations. Analysis of similarity (ANOSIM) via Primer 6 was used to test bacterial diversity indices and compositions in spatiotemporal variations.

For analysis of abundant and minor bacterial members, abundant groups (type I–III) and minor groups (type IV–VI) were separately incorporated into two matrixes. These matrices were used to perform ANOSIM in a Primer 6 based Bray–Curtis distance derived from relative abundance transformed by three methods, including no transformation, square root transformation, and binary transformation (i.e., presence and absence analysis).

Bacterial compositions between sampling times in each location were compared with Pearson correlations using R package Hmisc<sup>1</sup> . In addition, Pearson correlations were used to compare fluctuations of Endozoicomonas and Acinetobacter.

Hierarchical clustering analysis was performed using the average linkage via R package pheatmap (Kolde, 2015) and visualized on a heat map. The distance matrix used for clustering was calculated using the R package vegan (Dixon, 2003) based on a base-10 logarithmic transformation of relative abundance (in percentage), plus a pseudo-count of 1 × 10−<sup>8</sup> (Costea et al., 2014). Genera of unknown taxonomy or <0.2% relative abundance were excluded.

## RESULTS

#### Diversity of Microbial Community

Seawater sample K1001S yielded the most operational taxonomic units (OTUs), 406.15 (in rarefaction), whereas coral sample K0808C had the fewest (58.68 OTUs; Supplementary Table S2). According to ANOSIM, bacterial diversity indices in corals and seawater samples differed (richness: R = 0.634, p = 0.001; Shannon: R = 0.298, p = 0.001) and bacterial community in seawater samples had greater richness and Shannon diversity than those in corals (Supplementary Table S2).

#### Composition of Bacterial Community

Overall, bacterial compositions in S. pistillata from Kenting, Yehliu, and Lyudao were similar at a class level,

<sup>1</sup>https://github.com/harrelfe/Hmisc

similar to seawater samples (Supplementary Figure S1). However, bacterial communities in S. pistillata and seawater differed from each other. Dominant bacterial classes in seawater were Flavobacteria, unclassified Marinimicrobia, Gammaproteobacteria, and Alphaproteobacteria. Among these, most dominant genera were unclassified, such as unclassified Marinimicrobia, unclassified Gammaproteobacteria, and unclassified Flavobacteriaceae. In addition, the genus Vibrio, which belongs to Gammaproteobacteria, was higher than those in corals.

In S. pistillata, Actinobacteria, Betaproteobacteria, and Gammaproteobacteria were dominant, and their relative abundances higher than those in seawater. In Gammaproteobacteria, dominant genera were Endozoicomonas, Acinetobacter, Pseudomonas, and Serratia (Supplementary Figure S1), among which Endozoicomonas was particularly more dominant in Kenting and Lyudao compared to Yehliu. Propionibacterium and Actinophytocola were the two dominant Actinobacteria genera, and Massilia Betaproteobacteria in coral.

Bacterial community structure in S. pistillata and seawater at various sampling times is shown (**Figure 1**). Some genera, such as Ralstonia, Cupriavidus, and Aquabacterium, were only detected in coral samples, whereas Pseudoalteromonas was common only in seawater, and Gramella was exclusively retrieved from seawater samples. In addition, among coral samples, although some bacterial taxa were prevalent at every collection, their relative abundance varied from time to time (**Figure 1**). For example, Endozoicimonas and Acinetobacter fluctuated in abundance throughout this survey.

## Variation of Bacterial Composition in Different Time

There were differences among sampling years in bacterial community structure (ANOSIM: R = 0.301, p = 0.002; Supplementary Figure S2). Among all sites, based on nMDS, the bacterial community changed along the sampling time (**Figure 2**). Within each location, bacterial composition also differed among sampling years or times (Supplementary Figure S3), especially in samples from Kenting (ANOSIM: R = 0.370, p = 0.028). In Lyudao and Yehliu, bacterial compositions were clustered according to sampling years (difference not significant), but the Pearson correlation value between each sampling time in Lyudao and Yehliu decreased significantly (in Lyudao, from 1.00 to 0.24; in Yehliu, from 1.00 to 0.16) with increasing time span (Supplementary Figure S4), which indicated changes in bacterial composition during sampling times. In addition, there was no difference between seasons (ANOSIM: R = −0.072, p = 0.791).

To determine differential effects of short- versus long-term observations, consecutive patterns of correlations between bacterial communities were analyzed (Supplementary Figure S5). Correlations fluctuated over five consecutive samplings (Supplementary Figures S5A–F). However, a longer time

abundance (percentage). Genus names are prefixed with "g\_," and phylum names with "p\_" in parenthesis. Columns are clustered by Bray–Curtis distance and rows by Euclidean distance of their abundance profile.

span revealed a more holistic view of variation in bacterial composition (Supplementary Figure S5G), not apparent in short-term observations. The longer time span also revealed that changes followed a pattern. For example, two peaks (times 4–6 and time 9) had different duration, indicating different resilience status (Supplementary Figure S5G). Both abundant and minor (<1% relative abundance) genera had a significant differences between sampling years, and only minor genera did so by considering presence/absence data (binary transformation; Supplementary Table S3). In Yehliu, these abundant genera (except Endozoicomonas) retained similar abundance and moderately increased in 2009, although they had more variable fluctuations in abundance in Lyudao and Kenting, especially Endozoicomonas and Acinetobacter (**Figure 3**). In Lyudao, Acinetobacter increased during July to November in 2008, and Endozoicomonas increased during May and July in 2008, but decreased thereafter. In Kenting, abundance of Acinetobacter was low during 2008, increased during May and July in 2009, and decreased from November 2009 to January 2010. However, Endozoicomonas bloomed in most sampling times during 2008 and decreased during May and July in 2009, followed by an increase in November 2009. Intriguingly, both in Lyudao and Kenting, the times of Endozoicomonas and Acinetobacter blooms were staggered (**Figure 3**). Furthermore, in Kenting, there was a negative correlation (Pearson correlation: R = −0.899, p = 0.038) between Endozoicomonas and Acinetobacter abundance fluctuations from November 2008 to November 2009.

## Variation of Bacterial Composition among Locations

Based on nMDS, there was distinct clustering of coral-associated bacterial composition in Yehliu, Lyudao, and Kenting (ANOSIM: R = 0.224, p = 0.014; **Figure 2**). Kenting and Yehliu had differences not only in bacterial composition (ANOSIM: R = 0.437, p = 0.004) but also in community richness (ANOSIM: R = 0.218, p = 0.045). There was a compositional difference among sampling sites in dominant genera (ANOSIM: R = 0.240, p = 0.008), but not in minor genera (Supplementary Table S3).

Endozoicomonas and Acinetobacter were the two frequently detected genera of high abundance (**Figure 3**), of which both had disparate inhabiting preferences on S. pistillata in different locations (**Figure 4**). Endozoicomonas were mostly present in corals from Lyudao and Kenting, and composed of distinct genotypes in each locations (i.e., OTU02, 03, 04, 07, 08, and 10 in Kenting; OTU05, 06, and 09 in Lyudao). These Endozoicomonas OTUs were merely detected in both Kenting and Lyudao, indicating that geographical setups determined their inhabiting specificity (**Figure 4A**). Based on phylogenetic analysis, these OTUs were highly associated with E. elysicola and E. atrinae (**Figure 4C**). In contrast to exclusive detection of Endozoicomonas in coral, Acinetobacter were present in both S. pistillata and seawater (**Figure 4B**). Furthermore, most Acinetobacter OTUs were shared by three sampling sites, with less geographical preference than Endozoicomonas (**Figure 4B**).

These Acinetobacter OTUs were also more diverse compared to Endozoicomonas OTUs in terms of phylogenetic affiliation (**Figure 4D**).

## The Resident Types of Coral-Associated and Seawater-Associated Bacteria

According to relative abundance and detection frequency, bacteria in S. pistillata and seawater were categorized into six resident types (Section Resident Types of Bacterial Lineages in Materials and Methods and **Table 1**). In S. pistillata, type I group (i.e., with >1% relative abundance and detected in ≥80% samples) had eight genera, including Acinetobacter, Endozoicomonas, Propionibacterium, Corynebacterium, Staphylococcus, Arthrobacter, Actinophytocola, and Pseudomonas. However, there were only Vibrio and unclassified Marinimicrobia as type I in seawater. Among these abundant genera (resident type I–III), coral and seawater shared no genus.

## DISCUSSION

The present study demonstrated that long-term observation of coral-associated microbial communities provided a baseline to elucidate coral–microbe and microbe–microbe interactions. Abundant and minor bacteria in S. pistillata had different inclinations to geographical and temporal changes. Based on relative abundance and detection frequency, prevalent and nonprevalent S. pistillata-associated bacterial genera were allotted into six putative resident types.

#### Long-Term Survey Revealed Various Shifts in Coral-Associated Bacterial Community

Shifts in bacterial communities in healthy coral tissues were only apparent with a sufficient duration of survey (∼2 years). Although changes in bacterial composition in Lyudao and Yehliu were not confirmed (samples were lost due to typhoons and human activities), based on the correlation between sampling times, community resilience may be present over a longer sampling interval. In the present study, the long and complete observation in Kenting successfully captured compositional resilience that was not apparent in short-term surveys. A small group of bacteria (that comprised the core microbiome) were ubiquitously associated with coral, regardless of abiotic environmental parameters (Hernandez-Agreda et al., 2016). However, based on our results, prolonged observation enabled characterizing their succession in coral (Lynch and Neufeld, 2015). In this study, changes in prevalent abundant genera Acinetobacter and Endozoicomonas had disparate fluctuation patterns among locations, with abundances constant in Yehliu but variable in Kenting in 2009. Negative abundance correlations between these two genera were not detected before November

Yehlui; LD, Lyudao; C, colony; SW, seawater.

2008 but afterward until November 2009; therefore, this successional pattern was far from conclusive. Apart from abundant ones, minor bacterial genera in S. pistillata also changed in abundance throughout the study, which emphasized the need for longer surveys of coral-associated bacteria. Although Pantos et al. (2003) proposed some explanations for observed fluctuations in coral-associated bacterial communities, their hypotheses were only based on short-term observations. It has been suggested that coral-associated microbes can attenuate or intensify cumulative stressors that may disrupt coral reef ecosystems (Bourne et al., 2016). Therefore, we believe that these stressors, which varied in times and degrees from place to place, also contributed to spatial and temporal dynamics in coral-associated bacterial composition and abundance.

#### Abundant Bacterial Genera Showed Differential Geographical Inclinations

Different coral reproduction strategies (i.e., broadcast spawning and brooding) affect symbiotic bacterial composition differently, and S. pistillata, a brooder coral, harbored distinguishable bacterial communities strongly clustered by geographical location (Neave et al., 2017b), implying, in S. pistillata, vertical transfer of microbes from parents to offspring, thereby restricting the microbial structure and development and leading to the observed geographical grouping. As some bacteria have inhabiting specificity to coral species (Hong et al., 2009; Reis et al., 2009), discrepant microbial compositions among corals reported by different studies were plausible.

Among abundant genera, Acinetobacter, Propionibacterium, and Pseudomonas were more constantly detected in S. pistillata regardless of seasonal and geographical differences, whereas Endozoicomonas had geographical variations. Endozoicomonas spp. have diverse roles in different hosts as being symbiotic (La Rivière et al., 2013) and parasitic (Mendoza et al., 2013). The documented high proportion of repeats and insertion sequences in E. montiporae (Ding et al., 2016) and pathogenic strain Candidatus Endozoicomonas cretensis (Katharios et al., 2015) corresponded to their high genome plasticity for wide adaptation to various hosts (reviewed by Neave et al., 2017a). Since Endozoicomonas was abundant in Lyudao and Kenting but not in Yehliu, their adaptation appeared more favorable for tropical versus subtropical oceans.


<sup>∗</sup>Genera detected in both coral and seawater.

#### Minor Bacterial Genera Might be Opportunistic or Keystone Players in S. pistillata

Based on the presence/absence transformation of abundance, minor genera had significant differences among sampling years, reflecting their fluctuating nature in S. pistillata over time. Ralstonia and Propionibacterium were two acknowledged minor symbiotic genera in corals and intimately associated with dinoflagellate endosymbionts (Ainsworth et al., 2015). However, in our study, Propionibacterium was abundantly detected among three locations during sampling, whereas Ralstonia had constantly low abundance among sampling sites. Lynch and Neufeld (2015) suggested that prevalent, minor taxa were just conditionally minor in abundance, and would opportunistically grow with abundance under optimal conditions. This accounted for the discrepant abundance of Propionibacterium in the current versus previous studies. In contrast, consistent detection of Ralstonia's low abundance described its potential of being a keystone taxon permanently holding rare abundance in coral (Giovannoni and Stingl, 2005; Lynch and Neufeld, 2015), although the functional role of Ralstonia in S. pistillata remains unclarified.

#### Various Functional Roles Are Likely Mediated by Type I Genera in S. pistillata

Many bacterial taxa were reported as abundant members in coral (Littman et al., 2009; Liu et al., 2012; McKew et al., 2012; Morrow et al., 2012; Carlos et al., 2013), and mostly on Spongiobacter and Endozoicomonas (Lee et al., 2012; Morrow et al., 2012; Bayer et al., 2013; Carlos et al., 2013; Lema et al., 2014). These abundantly, frequently detected bacteria seemed adapted to live in coral tissues and acting as coral nutrition, pathogens, probiont, or purely commensal bacteria in the coral holobiont (Klaus et al., 2005). In the present study, we categorized S. pistillata-associated bacteria into six resident types, based on their abundance and detection frequency, with Acinetobacter (in type I) being the prevalent genera. Acinetobacter had been detected in stony corals from various regions (Littman et al., 2009; McKew et al., 2012; Morrow et al., 2012; Carlos et al., 2013; Li et al., 2014), and potentially played either beneficial or detrimental roles in corals. For example, Shnit-Orland and Kushmaro (2009) considered Acinetobacter sp. as a first-line defender that assisted the coral holobiont against pathogens resistant to multiple antibiotics. However, in a study of Dark Spot Syndrome in coral Stephanocoenia intersepta, Acinetobacter was regarded as a potential pathogen (Sweet et al., 2013). Given the common presence in stony corals, the uncertain role of Acinetobacter in S. pistillata needs to be clarified.

Among the eight genera in resident type I, four (i.e., Propionibacterium, Corynebacterium, Arthrobacter, and Actinophytocola) belong to the phylum Actinobacteria, consistent with other reports regarding Actinobacteria as a dominant genus associated with corals. Propionibacterium was present in coral Cirrhipathes lutkeni (Santiago-Vázquez et al., 2007), Mussismilia hispida (de Castro et al., 2010) and Acropora digitifera (Nithyanand et al., 2011). Using Actinobacteria-specific primers, Yang et al. (2013) reported 19 actinobacterial genera in two corals, suggesting high diversity of Actinobacteria in both hard and soft corals. Coral-associated actinobacteria reportedly had antimicrobial activity (Sweet et al., 2013; Mahmoud and Kalendar, 2016) attributed to production of bioactive substances (Yang et al., 2013). Therefore, these four type-I actinobacterial genera warrant careful consideration in future studies, considering their broad distribution in corals and antimicrobial properties.

## Metabolic Capability May be Associated with Bacterial Dynamics in S. pistillata

In the ocean, heterotrophic bacterial communities (in terms of species composition, spatiotemporal variations, and community structure) are largely affected by the availability, composition, and spatiotemporal distribution of organic substrates. Similarly, variations in coral-associated bacterial community are also controlled, to a great extent, by organic matter secreted by coral, similar to the situation of algae-associated microbial communities (Dang and Lovell, 2016). For example, the genome of Endozoicomonas has a wide spectrum of genes related to generic transport for uptake of extracellular organic compounds and carbohydrates (Ding et al., 2016; Neave et al., 2017a). Numbers of transport molecules also implied interactions between Endozoicomonas and compounds produced by corals (Neave et al., 2014; Ding et al., 2016). Similarly, Acinetobacter was reportedly able to metabolize dimethyl sulfide (Horinouchi et al., 1997) produced by symbiotic algae in corals; furthermore, its precursor (dimethylsulfoniopropionate) has served as a chemoattractant to heterotrophic bacteria in oceans (Seymour et al., 2010). Therefore, availability of extracellular compounds in coral holobiont is likely associated with fluctuations in the abundance of Endozoicomonas and Acinetobacter.

## CONCLUSION

This long-term survey revealed dynamics of S. pistillataassociated bacterial community (e.g., compositional resilience), with fluctuating patterns of abundant and minor genera stratified by sampling times and geographical locations. Although Endozoicomonas and Acinetobacter were detected with high abundance and frequency, this long-term observation identified differential inhabiting inclination, defined by Endozoicomonas being more acclimated to tropical oceans, whereas Acinetobacter was merely confined by climatic zones. Regarding minor genera, despite the lack of functional evidence clarifying their opportunistic or keystone roles in S. pistillata, a long-term survey could legitimately distinguish them from somewhat transient colonizers, and also identify microbial candidates for future functional studies. Lastly, abundance-shuffling and

#### REFERENCES


species-switching of symbiotic algae improved environmental adaptation of corals (Buddemeier and Fautin, 1993; Fautin and Buddemeier, 2004; Boulotte et al., 2016), although some bacteria had beneficial effects relating to holobiont resilience without abundance fluctuations under short-term heat stress (Ziegler et al., 2017). Therefore, we inferred that large-scale and long-term time series observations are required to characterize fluctuating and steady microbial components in coral holobiont.

## AUTHOR CONTRIBUTIONS

S-HY substantially analyzed data and wrote the manuscript. C-HT performed bioinformatics analysis, provided detailed methods and results for the analysis, and wrote the manuscript. C-RH performed sampling, DNA extraction, PCR, and data analysis. C-PC acquired samples and performed DNA extraction. KT performed statistical test and participated in the manuscript editing. SL participated in the manuscript writing and editing. P-WC performed sampling, DNA extraction, and PCR. J-HS and CC acquired samples. S-LT planned the experimental design and participated in manuscript editing.

## FUNDING

Funding for this study was provided by the Biodiversity Research Center, Academia Sinica, Taiwan.

## ACKNOWLEDGMENTS

We thank CC's team for coral sampling. We also thank our dear lab member, the late C-RH for his efforts devoted to this project; may he rest in peace.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.01094/full#supplementary-material



endemic reef coral Mussismilia braziliensis. J. Appl. Microbiol. 106, 1378–1387. doi: 10.1111/j.1365-2672.2008.04106.x


taxonomy. Appl. Environ. Microbiol. 73, 5261–5267. doi: 10.1128/AEM.0 0062-07


**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 Yang, Tseng, Huang, Chen, Tandon, Lee, Chiang, Shiu, Chen and Tang. 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.

# Community Composition and Transcriptional Activity of Ammonia-Oxidizing Prokaryotes of Seagrass Thalassia hemprichii in Coral Reef Ecosystems

Juan Ling<sup>1</sup> , Xiancheng Lin1,2, Yanying Zhang1,3, Weiguo Zhou1,2, Qingsong Yang1,2 , Liyun Lin1,2, Siquan Zeng1,2, Ying Zhang1,2, Cong Wang1,2, Manzoor Ahmad1,2 , Lijuan Long<sup>1</sup> and Junde Dong1,3 \*

<sup>1</sup> CAS Key Laboratory of Tropical Marine Bio-resources and Ecology, Guangdong Provincial Key Laboratory of Applied Marine Biology, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, China, <sup>2</sup> University of Chinese Academy of Sciences, Beijing, China, <sup>3</sup> Tropical Marine Biological Research Station in Hainan, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Sanya, China

#### Edited by:

Hongyue Dang, Xiamen University, China

#### Reviewed by:

Wei Xie, Tongji University, China Anyi Hu, Institute of Urban Environment (CAS), China Qichao Tu, Zhejiang University, China

> \*Correspondence: Junde Dong dongjd@scsio.ac.cn

#### Specialty section:

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

Received: 31 May 2017 Accepted: 04 January 2018 Published: 25 January 2018

#### Citation:

Ling J, Lin X, Zhang Y, Zhou W, Yang Q, Lin L, Zeng S, Zhang Y, Wang C, Ahmad M, Long L and Dong J (2018) Community Composition and Transcriptional Activity of Ammonia-Oxidizing Prokaryotes of Seagrass Thalassia hemprichii in Coral Reef Ecosystems. Front. Microbiol. 9:7. doi: 10.3389/fmicb.2018.00007 Seagrasses in coral reef ecosystems play important ecological roles by enhancing coral reef resilience under ocean acidification. However, seagrass primary productivity is typically constrained by limited nitrogen availability. Ammonia oxidation is an important process conducted by ammonia-oxidizing archaea (AOA) and bacteria (AOB), yet little information is available concerning the community structure and potential activity of seagrass AOA and AOB. Therefore, this study investigated the variations in the abundance, diversity and transcriptional activity of AOA and AOB at the DNA and transcript level from four sample types: the leaf, root, rhizosphere sediment and bulk sediment of seagrass Thalassia hemprichii in three coral reef ecosystems. DNA and complementary DNA (cDNA) were used to prepare clone libraries and DNA and cDNA quantitative PCR (qPCR) assays, targeting the ammonia monooxygenase-subunit (amoA) genes as biomarkers. Our results indicated that the closest relatives of the obtained archaeal and bacterial amoA gene sequences recovered from DNA and cDNA libraries mainly originated from the marine environment. Moreover, all the obtained AOB sequences belong to the Nitrosomonadales cluster. Nearly all the AOA communities exhibited higher diversity than the AOB communities at the DNA level, but the qPCR data demonstrated that the abundances of AOB communities were higher than that of AOA communities based on both DNA and RNA transcripts. Collectively, most of the samples shared greater community composition similarity with samples from the same location rather than sample type. Furthermore, the abundance of archaeal amoA gene in rhizosphere sediments showed significant relationships with the ammonium concentration of sediments and the nitrogen content of plant tissue (leaf and root) at the DNA level (P < 0.05). Conversely, no such relationships were found for the AOB communities. This work provides new insight into the nitrogen cycle, particularly nitrification of seagrass meadows in coral reef ecosystems.

Keywords: seagrass, ammonia-oxidizing archaea and bacteria, ammonia monooxygenase subunit A (amoA), cDNA, community structure, coral reef ecosystems

## INTRODUCTION

fmicb-09-00007 January 23, 2018 Time: 17:52 # 2

Many investigations into the effect of ocean acidification (OA) on coral reefs have been conducted (Andersson and Gledhill, 2013), and results indicate that that marine organisms which inhabit the carbonate structures of coral reefs are important and sensitive to OA (Larkum et al., 2006). Albright et al. (2016) and Lough (2016) found that changes to pH in the seawater surrounding natural coral reefs in the southern Great Barrier Reef can significantly affect calcification rates, suggesting that OA may already be altering the growth of coral reefs. By decreasing the ocean pH, OA can affect the calcifying rates of calcifying creatures such as coralline algae with carbonate skeletons (Hendriks et al., 2014; Albright et al., 2016). However, seagrasses can increase the pH of the ambient environment via high photosynthesis rates (Lai et al., 2013; Hendriks et al., 2014). Moreover, Lamb et al. (2017) found that seagrass in coral reef ecosystems can reduce disease levels twofold in comparison with the corals located adjacent to seagrass meadows and corals at paired sites without seagrass.

Seagrass is highly productive and of great ecological importance in the marine environment. For instance, it can provide food, nursery and breeding habitats for other marine organisms inhabiting the ecosystem and nutrients for coral reefs. Nonetheless, available nitrogen is usually the main factor limiting the primary productivity of seagrass because coral reef ecosystems are largely oligotrophic. Nitrification is a key process in the nitrogen cycle in the marine environment. The first and rate-limiting step of nitrification is performed by ammonia-oxidizing archaea (AOA) and ammonia-oxidizing bacteria (AOB), both of which are ammonia oxidizers and responsible for converting ammonia to nitrite. However, these microbes have different phylogenetic and physiological features, resulting in significant variations in their abundance, diversity and activity under different environmental conditions. For example, AOA can adapt to a variety of habitats and account for a large proportion of marine and estuary communities (Dang et al., 2008, 2009, 2010a,b, 2013; Boyd et al., 2011; Cao et al., 2012; Rusch and Gaidos, 2013). However, Di et al. (2009), Wu et al. (2011), Zhang et al. (2014), Zheng et al. (2014) reported that AOB might play a more significant role in the ammonia oxidation process under certain conditions. Consequently, the relative contribution of AOA and AOB to ammonia oxidation is still in debate.

Zhang J. et al. (2015) suggested that the presence of certain freshwater plants, such as Iris pseudacorus, Thalia dealbata, and Typha orientalis L., affected the ecological characteristics of AOA and AOB significantly by increasing the abundance of ammonia oxidizers in the rhizosphere sediments. For the marine environment, investigations of Moin et al. (2009) and Chen and Gu (2017) revealed that the presence of Spartina alterniflora or S. patens and mangrove roots had a strong influence on the diversity and abundance of AOA and AOB in the coastal area and mangrove ecosystems, respectively. Furthermore, investigations of the ammonia-oxidizing prokaryotes have been conducted extensively in many different marine environments including the Changjiang Estuary, the Jiaozhou Bay, the tropical West Pacific Continental Margin, the Okhotsk Sea, the Sargasso Sea, the Northern South China Sea (Mincer et al., 2007; Nakagawa et al., 2007; Dang et al., 2013; Li and Gu, 2013; Newell et al., 2013; Lipsewers et al., 2014; Cao et al., 2015; Yu et al., 2016).

Regardless, almost all the above-mentioned investigations focused on the seawater and sediment using only DNA as a proxy to assess AOA and/or AOB. Few investigations have focused on ammonia-oxidizing prokaryotes in the aquatic benthic flora, particularly for the seagrass ecosystem (Zhao et al., 2014; Bernhard et al., 2016; Frame et al., 2016; Huang et al., 2016; Mejia et al., 2016; Ettinger et al., 2017). Not to mention the studies of AOA and AOB of seagrass in coral reef ecosystems aimed at revealing the transcriptional activity of relevant functional groups in their natural physiological state. Accordingly, in this study, we prepared archaeal and bacterial amoA gene DNA and cDNA libraries and performed reverse-transcription polymerase chain reaction (PCR) and real-time quantitative PCR (qPCR) to examine the community abundance, diversity and transcriptional activity of ammonia-oxidizing prokaryotes of the seagrass Thalassia hemprichii in the Luhuitou fringing reef, Sanya Bay and Yongxi Island, Xisha Islands. The aims of our investigation were as follows: (i) to evaluate the abundance and diversity of AOA and AOB communities, (ii) to compare community variations within and among the sample types and locations, and (iii) to analyze the transcriptional activity of AOA and AOB.

#### MATERIALS AND METHODS

#### Study Sites and Sampling

Seagrass T. hemprichii is one of the most widely distributed seagrass species among tropical southern Indo-Pacific flora, and exists in a monospecific or mixed-species status. The sampling locations were distributed in Sanya Bay (SYT) and Yongxing Island (AT and ST), South China Sea (Supplementary Figure S1). Samples at SYT were collected from the Luhuitou fringing reef (18◦ 120 19<sup>00</sup> N, 109◦ 280 27<sup>00</sup> E) located in Sanya Bay, Hainan Province. The average atmospheric temperature at this site is 30.74◦C, with warm summers (34.75◦C) and cold winters (27.20◦C). The Luhuitou fringing reef area is under the effect of the northeast monsoon (cold and dry winter and spring) and southwest monsoon (warm and wet summer and autumn) during the East Asian monsoon climate (Wu et al., 2008; Cao et al., 2017). Two other sampling locations, AT (16◦ 500 32<sup>00</sup> N, 112◦ 200 41<sup>00</sup> E) and ST (16◦ 500 6 00 N, 112◦ 220 10<sup>00</sup> E), are located on Yongxing Island, which is a reef island formed by the accumulation of white coral skeletal material and shell sand on a reef platform. The annual average temperature on Yongxing Island is 26.5◦C. This island is also under the effect of the East Asian monsoon (Shen et al., 2017).

Seagrass meadows in the Luhuitou fringing reef and the Yongxing flat reef are representative of the different styles of seagrass meadows in coral reef ecosystems. T. hemprichii

appeared at the Luhuitou fringing reef after the coral reef has degraded, and it is the only seagrass species present. In the Yongxing Islands, the seagrass is found in a mixedspecies status, with seagrasses Syringodium isoetifolium, Halodule uninervis, and Halophila ovalis at AT, whereas T. hemprichii was dominant over seagrass H. ovalis at ST. In comparison with the Yongxing flat reef, Luhuitou fringing reef areas show higher nutrient concentration, particularly nitrogen, which was attributed to the increasing anthropogenic activity (Cao et al., 2017).

Samples from the Luhuitou fringing reef and Yongxing flat reef were collected on May 28th and June 1st, 2015, respectively. Sampling was carried out according to the methods of Jensen et al. (2007) at low tide. Plants with surrounding sediment were randomly collected using a spade, and immediately transported in sterilized boxes for subsequent subgrouping in triplicate. Sediment from the plant roots and associated invertebrates from leaves and roots were separated by washing with autoclaved seawater. Bulk sediment was also collected at the same area. All sediment samples were collected in triplicate at each location and thoroughly homogenized using a sterilized spoon. Samples collected from one site were divided into four sections: leaves (L); rhizomes and roots (R); rhizosphere sediment (RS) and bulk sediment (S). All samples for DNA/RNA analysis were stored in sample protectors (TaKaRa, Dalian, China), frozen immediately, and stored at −80◦C until further analysis.

Environmental data, samples used for microbial analysis and physiochemical analysis were collected simultaneously. The temperature and salinity of the seawater adjacent to seagrass samples (within 3 cm) was measured using a YSI 6600V2 water quality sonde (YSI, Yellow Springs, OH, United States). Dissolved oxygen (DO) concentrations and pH values were measured using a portable pH/DO Meter (Thermo Fisher Scientific, Inc., Beverly, MA, United States). Inorganic nutrients in seawater, including ammonium, nitrate, nitrite, and phosphate, were measured using standard methods as described previously (Huang et al., 2003). Nitrogen and carbon content of seagrass tissues (L and R) were determined according to Lee et al. (2004), and phosphorus content was analyzed by the colorimetric analysis of phosphate concentration (Fourqurean et al., 1992). Chemical data (Nitrate, ammonium and active phosphorous) of sediments were determined by using standard oceanographic methods (General Administration of Quality Supervision, Inspection and quarantine of the People's Republic of China, 2002).

## DNA and RNA Extraction, cDNA Synthesis, PCR, Cloning, and Sequencing

DNA and RNA from approximately 1 g of sample (sediment or plant tissue; wet weight) were extracted using the E.Z.N.A <sup>R</sup> Soil DNA kit and E.Z.N.A <sup>R</sup> Soil RNA kit (Omega Bio-tek, Norcross, GA, United States) according to the manufacturer's protocols. Synthesis of cDNA from extracted RNA was performed according to Li and Gu (2013). The nucleic acid concentrations were quantified using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Wilmington, DE, United States). All qualified DNA and cDNA were stored at −80◦C until analysis. For clone library analysis, archaeal and bacterial amoA gene sequences were amplified using the primer sets Arch-amoA F (5<sup>0</sup> -GGGGTTTCTACTGGTGGT-3<sup>0</sup> ) and Arch-amoA R (50 -CCCCTCKGSAAAGCCTTCTTC-3<sup>0</sup> ) (Francis et al., 2005) and amoA-1F (5<sup>0</sup> -STAATGGTCTGGCTTAGACG-3<sup>0</sup> ) and amoA-2R (5<sup>0</sup> -GGGGTTTCTACTGGTGGT-3<sup>0</sup> ) (Rotthauwe et al., 1997), respectively.

PCR reaction mixture for amplifying the amoA gene was prepared in accordance with details described by Hu et al. (2011). The PCR amplification conditions for amoA gene in AOA and AOB were in accordance with previously established protocols (Rotthauwe et al., 1997; Francis et al., 2005; Schmid et al., 2005; Li et al., 2011). The amplification was performed as follows: 5 min at 95◦C, followed by 30 cycles of 45 s at 95◦C, 60 s at 53◦C and 60 s at 72◦C, and 10 min at 72◦C. For bacterial amoA gene amplification, the PCR conditions were 5 min at 95◦C, followed by 30 cycles of 45 s at 95◦C, 90 s at 56◦C and 60 s at 72◦C, and 10 min at 72◦C. The PCR products from three reactions were pooled together to minimize PCR amplification bias, purified, and ligated into pMD18-T vector (TaKaRa, Dalian, China) according to the manufacturer's instructions. Recombinant Escherichia coli cells were inoculated in Luria-Bertani broth containing ampicillin and incubated overnight at 37◦C, and the plasmids carrying the target genes were extracted using a MiniBEST Plasmid Purification kit (TaKaRa, Dalian, China). Cloned amoA gene fragments were reamplified using primers M13-F (5<sup>0</sup> -AGGGTTTTCCCAG-TCACGACG-3<sup>0</sup> ) and RV-R (5<sup>0</sup> - AGCGGATAACAATTTCACACAGG-3<sup>0</sup> ). The target fragment sizes of archaeal and bacterial amoA genes were 491 and 635 bp, respectively. The PCR products were cloned into the pMD18-T vectors (TaKaRa, Dalian, China). PCR products were screened for the correct size and purity by 1% agarose gel electrophoresis, and clones showing the correct size were sequenced.

#### Quantification of amoA Gene Copy Number at the DNA and Transcript Levels

Absolute quantification of archaeal and bacterial amoA genes were determined for both DNA and cDNA using qPCR in triplicate reactions with the LightCycler 480 System (Roche Diagnostics, Mannheim, Germany) and the following conditions: 95◦C for 30 s, followed by 40 cycles of 30 s at 95◦C, 60 s at 56◦C, and 60 s at 72◦C for archaeal amoA gene, or 95◦C for 30 s, followed by 40 cycles of 30 s at 95◦C, 60 s at 58◦C, and 35 s at 72◦C for bacterial amoA gene. To construct standard curves, archaeal and bacterial amoA genes were cloned into the pMD18-T vector (TaKaRa, Dalian, China) and then transformed into E. coli DH5a. The methods were similar to those used for the clone library construction. Recombinant plasmids carrying the target genes were extracted using a TaKaRaMini BEST Plasmid Purification Kit and quantified with a NanoDrop

2000 spectrophotometer. The copy numbers of amoA gene from the extracted plasmids were calculated by the concentrations and average base pairs of the plasmid. Standard curves for the archaeal amoA gene were constructed using standard plasmids obtained from the most dominant genotype clone ARSYTL81(KY794979), and for the bacterial amoA gene from the most dominant genotype clone BRNASYTRS53 (KY795002). Standard curves ranging from 10<sup>3</sup> to 10<sup>8</sup> gene copies/µL were obtained using 10-fold serial dilutions of linearized plasmid pMD-18T containing the cloned archaeal and bacterial amoA genes, respectively.

Real-time PCR efficiencies for AOA and AOB for DNA and cDNA were calculated according to E = 10 [−1/slope] (Rasmussen, 2001). The results showed that the amplification efficiencies ranged from 94.5 to 101%, with an R<sup>2</sup> of standards higher than 0.99. The specificity of the amplification products was confirmed by melt-curve analysis, and the amplified fragments were checked by electrophoresis in 1.0% gel to confirm the expected sizes of amplicons. The size of archaeal and bacterial amoA genes were 491 and 635 bp, respectively. As the gene copies in the 1 µL of template DNA were determined, the final amoA gene and cDNA abundance of the seagrass and sediment samples were obtained by calculation. The results were expressed as gene copy abundance per gram of sediment or plant tissue (wet weight).

#### Statistical Analysis

The obtained DNA and cDNA sequences were examined and checked for chimeras using the Check Chimera program of the Ribosomal Database Project (Cole et al., 2007). The operational taxonomic unit (OTU) reads were checked against a local amoA gene database (Ribosomal Database Project FunGene<sup>1</sup> ) and the NCBI database<sup>2</sup> . Diversity indices were also evaluated using the MOTHUR program (Schloss et al., 2009; Li and Gu, 2013). Diversity statistics, including Shannon–Wiener (H<sup>0</sup> ), Simpson (D) and species richness estimator (SChao1), were calculated. Library coverage (C) was calculated as [1-(n/N)] × 100, where n is the number of OTUs represented by one clone (singletons) and N is the total number of sequences (Good, 1953). Diversity indices and richness estimators are useful statistical methods for comparing the relative complexity of AOA and AOB communities and for assessing the completeness of sample analysis. Reference sequences were selected by comparison with the GenBank database using BLAST, and the closest matches were included in the alignment and phylogenetic analysis with MEGA 6 (JCVI, Rockville, MD, United States) through neighbor-joining trees using Kimura 2 parameter distance with 1000 replicates to produce bootstrap values (Tamura et al., 2013).

One-way statistical analysis of variance (ANOVA) (confidence limit of 95%, P < 0.05) was performed to analyze variables among the three locations. In addition, one-way analysis of similarity (ANOSIM) was performed based on Bray-Curtis distances of AOA and AOB communities among the

<sup>1</sup>http://fungene.cme.msu.edu/index.spr

<sup>2</sup>http://blast.ncbi.nlm.nih.gov/Blast.cgi

three sampling locations using the PRIMER v.6 software package (PRIMER-E, Plymouth, WA, United States) (Clark and Gorley, 2006). Moreover, Pearson's correlation analysis of the abundance and diversity of AOA and AOB with the determined physicochemical parameters (water, tissue and sediment) was analyzed by SPSS v19.0 software (IBM, Inc., Chicago, IL, United States). Multi-Variate Statistical Package (MVSP, version 3.2, Kovach Computing Services, Anglesey) software was used to construct the similarity matrix and dendrograms (Kovach, 1999). In addition, genetic similarities among all clones were calculated by the percent similarity coefficient. Principal coordinate analysis (PCoA) was employed to depict the general ordination patterns of all samples at both DNA and transcript levels. In addition, the weighted pair-group method with arithmetic analysis (WPGMA) was used to generate similarity matrices and dendrograms by percent similarity using MVSP (Kovach, 1999).

#### Nucleotide Sequence Accession Numbers

Representative sequences of archaeal and bacterial amoA genes for each OTU reported in this study have been deposited in the GenBank database under accession numbers MF796361– MF796384 and MF796347–MF796360, respectively.

#### RESULTS

#### Environmental Parameters, Plant Tissues, and Sediment Characteristics of the Samples

The features of four different types of samples from the collection locations were analyzed, and the results are listed in Supplementary Table S1. The pH and salinity of all sampling locations ranged from 8.22–8.39 and 24.70–27.87h, respectively. The highest DO concentration was recorded at AT, and the lowest at site SYT. The nitrate concentrations for the three locations ranged from 0.014 to 0.025 µM. The pH, salinity and DO concentration at SYT was significantly lower than at AT and ST, though the concentration of nitrite and ammonium were higher at SYT than at AT and ST (P < 0.05). Nonetheless, there was no significant difference in the nitrate concentrations among the three sampling sites (P > 0.05). The highest carbon content was found in the leaves from AT and the lowest value was recorded in the leaves from ST. The nitrogen content, phosphorus content and carbon percentage of roots and leaves from the three different locations were all significantly different (P < 0.05). The values of the bulk sediment environmental parameters (nitrate, ammonium, and active phosphorous concentrations) were all below the limit of detection. As illustrated in Supplementary Table S1, the ammonium and active phosphorous contents of the three RS samples ranged from 1.45 to 5.97 mg/kg and 14 to 15 mg/kg, respectively. Statistically, these three environmental parameters of AT, ST, and SYT rhizosphere sediments were significantly different (P < 0.05).

## Abundance of Archaeal and Bacterial amoA Genes

The abundances of AOA and AOB amoA genes at the DNA and transcript levels as determined by real-time PCR are shown in **Figure 1**. The abundance of all AOB communities (with the exception of that in sample SYTL) was higher than that of AOA at both the DNA and transcript levels. The abundance of AOA communities at the DNA level ranged from below the detection limit to 5.35 × 10<sup>6</sup> gene copies per gram of sample (L, R, RS, and S); the abundance of AOB communities at the DNA level was between 7.68 × 10<sup>5</sup> and 5.74 × 10<sup>6</sup> gene copies per gram of sample (wet weight). In addition, the abundance of AOA communities at the transcript level was one order of magnitude lower than that at the DNA level. Their abundance at the DNA level ranged from below the detectable levels to 6.45 × 10<sup>6</sup> copies g−<sup>1</sup> per gram of sample, and the abundance of AOB communities had a wide range, from below the limit of detection to 1.15 × 10<sup>6</sup> copies g−<sup>1</sup> per gram of sample. Moreover, the ratio of AOB to AOA at the DNA level was highly variable in all samples. The highest value 42.21 was detected in the SYTRS sample (Supplementary Table S2). The ratio of DNA

to cDNA for amoA gene copies for AOA community in sample SYT was greater than 20 and greater than 200 for sample SYTL. However, the values for AOB communities ranged only from 2.04 to 10.74 (Supplementary Table S2). When combing all the samples, the abundance of AOB communities was higher than that of AOA.

Bacterial amoA gene libraries from 13 samples (10 DNAbased and 3 cDNA-based) were successfully constructed. Overall, DNA sequences from 342 clones and cDNA sequences from 94 clones for archaea were recovered, while for bacterial amoA gene, there were 255 clones at the DNA level and 228 clones at the transcript level (Supplementary Tables S3, S4). Pearson's correlation analysis revealed that the abundance of AOA at the DNA level in all RS showed significant positive relationships with the concentrations of seawater ammonium and the nitrogen contents of seagrass roots and leaves (P < 0.05), respectively. However, there was no such relationship between the environmental parameters and the abundance of AOB communities.

#### Diversity of the Archaeal and Bacterial amoA Genes

The number of sequenced clones differed among samples (ranging from 17 to 47), and they were then used to calculate diversity estimators (**Table 1**). The coverage, diversity, and richness indices of the nine cDNA-based libraries are summarized in **Table 1**. The coverage for AOA and AOB at the two different levels ranged from 72.73 to 100% and 94.59 to 100%, respectively (**Table 1**). Consequently, our results might have reflected the majority of archaeal and bacterial amoA gene clones at the DNA and transcript levels in our samples (**Figures 2A,B**).

The biodiversity and richness indices of AOA and AOB communities at the DNA and cDNA levels are presented in **Table 1**. Rarefaction analyses were performed for all the bacterial or archaeal amoA gene clone libraries (Supplementary Figure S2). The highest diversity indices for AOB at the two levels were found for SYT. Overall, the indices indicated that AOB were less diverse than those for AOA, and the diversity indices at the DNA level were higher than those at the transcript level.

BLAST results indicated that over 80% of the obtained sequences recovered from this study were related to the sequences of marine sources, such as the marine water column, intertidal and marine sediments, mangrove sediments and marine sponges. The 24 archaeal amoA gene sequences share 87.50–99.48% sequence similarity with the closest GenBank matches. For bacterial amoA gene sequences, similarity of the 14 bacterial amoA genes ranged from 98.36 to 99.79%.

For AOA communities, 24 OTUs and 12 OTUs were found at the DNA and transcript levels, respectively, with 11 OTUs shared at both levels (**Figure 2** and **Table 1**). For AOB communities, the number decreased to 14 OTUs and 7 OTUs, with 6 OTUs shared at the two levels (**Figure 2** and **Table 1**). Based on the phylogenetic analysis, all bacterial amoA gene sequences (256 DNA sequences and 194 cDNA sequences) were mainly grouped into the Nitrosomonadaceae cluster (9 OTUs: 9 DNA sequences and 3 cDNA sequences) (**Figure 2B**).


TABLE 1 | Biodiversity and predicted richness of the archaeal and the bacterial amoA gene sequences.

OTUs of amoA sequences were generated by FunGene. C, coverage of the constructed clone libraries; H<sup>0</sup> , Shannon–Weiner index; D, Simpson index; SChao1, richness estimators. SYT: T. hemprichii from Sanya Bay; AT: T. hemprichii from the airport area of Xisha Islands; ST: T. hemprichii from the stone Islands area of Xisha Islands; -L, -R, -S and -RS referred to leaf, root, bulk sediment and rhizosphere sediment, respectively; the total OTUs for the whole AOA and AOB communities were obtained by analyzing all the clones at the DNA level and transcript level, respectively.

The diversity indices showed no significant relationships between the values of H<sup>0</sup> and SChao1 with the concentrations of ammonium and phosphate (P > 0.05) for both AOA and AOB communities at the DNA level, respectively. However, our investigation revealed that there was a significant negative relationship between SChao1 and the concentration of nitrite for AOB communities (P < 0.05).

#### Variations in AOA and AOB Community Composition within and among Locations

For the archaeal amoA gene libraries, 1 to 11 and 5 to 9 OTUs at the DNA and transcript levels, respectively, were found for different samples. For AOB, only 2 to 5 and 1 to 2 OTUs were obtained for different clones at the DNA and transcript levels, respectively. Some OTUs could be found among almost all the samples, such as ADSR 93 for AOA at the DNA level, sharing approximately 99.37% similarity with an uncultured archaeon clone (KY357274) isolated from mangrove sediment. For AOB at the DNA level, the OTU BSYTR23 exhibited 99.79% similarity with the uncultured bacterial clone HaAOB1 (JN177536) retrieved from marine sponges. However, other OTUs were only detected in ATR, such as the OTU BRAZ73 at the transcript level. The BLAST result for OTU BRAZ73 indicated high similarity to a clone (KC893630) retrieved from the marine sponge Spheciospongia vesparium.

As shown in Supplementary Figure S3, samples obtained from the same location tended to group together regardless of the sample type. For instance, AT and ST samples first grouped with samples obtained from the same location and then grouped together, whereas, SYTS, SYTRS, and SYTR samples shared a high similarity of community composition (Supplementary Figure S3). For the DNA-based analysis, the first two principal coordinates (P1 and P2) could explain 70.40 and 76.77% of the total community variability in PCoA in archaeal and bacterial ammonia oxidizers, respectively (**Figures 3A,B**). The percentage of variability explained by the first two principal coordinates was 87.10% at the transcript level for AOB (**Figure 3C**). The WPGMA results were consistent with the PCoA plots (**Figure 3**). Consequently, the AOA and AOB communities shared higher

and their closest matches in GenBank from DNA samples and cDNA samples. Bootstrap values greater than 50% of 100 resamplings are shown near the nodes.

similarity within the same location than within the same type of samples.

## DISCUSSION

## AOA and AOB Abundance in Different Niches and Locations

All samples collected from the three locations were analyzed by DNA-based and transcript-based approaches. Most of the archaeal and bacterial amoA gene sequences at the DNA level in this study were successfully recovered, whereas a few samples at the transcript level were retrieved. This may be due to low gene copy number or expression of amoA gene in the relatively oligotrophic coral reef ecosystems, resulting in an abundance below the limit of detection. Compared with samples from AT and ST, archaeal and bacterial amoA gene sequences in most samples collected at SYT were successfully recovered. Previous investigations demonstrated that environmental factors, such as ammonia, temperature, salinity, dissolved oxygen and pH, had strong influences on the distribution of AOA and AOB (Li et al., 2011; Cao et al., 2015; Wang et al., 2015). For instance, Li and Gu (2013) showed that the diversity, abundance, and transcriptional activity of AOA and AOB shift in response to N conditions, specifically noting that ammonium amendment increased diversity and a lower nitrite concentration may reduce AOA and AOB diversity. Low temperature also exerted an important effect on the composition of AOA and AOB communities, which exhibited the lowest diversity when exposed to cold water (Urakawa et al., 2008). Different species respond differently to environmental variation, and the results of our investigation revealed the ammonium concentration to be a decisive factor for AOA and AOB community composition.

A stimulation experiment conducted by Prosser and Nicol (2012) suggested that AOA grew faster than AOB at lower ammonia concentrations, as the AOA affinity for ammonium was up to 200-fold that of AOB (Martens-Habbena et al., 2009, 2015). It has also been reported that AOA prefer to inhabit environments with lower ammonia concentrations and have higher amoA gene transcriptional activity than AOB in ammonia-limited water environments. In addition, Takano et al. (2010) discovered that deep-sea archaea adopted the


TABLE 2 | One-way ANOSIM results for the AOA and AOB communities between the sampling locations Sanya Bay (SYT) and Yongxing Island (AT and ST) at the DNA level.

Significant differences (P < 0.05) are indicated in italics and bold.

strategy of recycling membrane lipids between growing cells and the surrounding sediment for saving energy to thrive in low ammonium habitats. However, in ammonia-rich areas, AOB communities would be the dominant component and contribute more to ammoxidation (Zhang Y. et al., 2015). Furthermore, plant species and their densities have crucial roles in determining community composition (Wang et al., 2015; Ettinger et al., 2017). Li et al. (2011) found that the presence of mangroves to increase the abundance of AOA and AOB in the mangrove sediment.

The qPCR quantification results presented in this study suggested that bacterial amoA gene abundance in almost all samples was higher than that of the archaeal amoA gene copy number with the exception of sample SYTL (**Figure 1**). The bacterial amoA gene abundance in the South China Sea was reported to range from 4.24 × 10<sup>4</sup> to 1.99 × 10<sup>6</sup> copies per gram of sediment (wet weight), which was consistent with our results (Cao et al., 2012). In addition, Dang et al. (2010a) found that the abundance of β-AOB was much higher than that of the archaeal amoA gene, and Wang et al. (2015) reported that the abundance of AOB at the transcript level was two orders of magnitude higher than that of AOA in a mangrove ecosystem. These findings were in agreement with our results. Furthermore, plants affected the bacterial community composition and activity by competing with rhizosphere microbes for nutrients, such as ammonium, nitrate, urea, and amino acids as nitrogen sources (Skiba et al., 2011). Foreseeably, microbes in rhizosphere sediment may utilize the low molecular weight compounds diffused from plant roots as carbon sources (Philippot et al., 2013). The ratio of β-AOBamoA/archaeal amoA ranged from 212: 1 to 3090: 1 in deep-sea methane seep sediments of the Okhotsk Sea (Dang et al., 2010b). By comparison, the abundance ratio of AOA to AOB in our study was much lower, ranging from 0.96: 1 to 42.21: 1 (Supplementary Table S2).

#### Diversity of Ammonia-Oxidizing Archaeal and Bacterial Communities

The clusters of archaeal amoA gene sequences obtained in this investigation were mainly from uncultured Thaumarchaeota originating from the marine environment. A chemolithoautotrophic marine crenarchaeote has been isolated, and its role in relation to nitrification has been shown to contribute significantly to global nitrogen and carbon cycles (Könneke et al., 2005). In addition, thaumarchaeotes have been found to play a crucial role in nitrification in both marine and terrestrial environments (Leininger et al., 2006; Wuchter et al., 2006; Beman et al., 2008; Erguder et al., 2009; Martens-Habbena et al., 2009). Moreover, Thaumarchaeota accounted for almost 12% of all archaeal sequences retrieved in Checker Reef sediments, and these organisms preferred oxic rather than anoxic sediments (Rusch et al., 2009; Gaidos et al., 2011; Pester et al., 2011). Beman et al. (2007) analyzed AOA communities associated with coral colonies from nine coral species and four different reef locations in the Gulf of California. Their results showed that amoA sequences were broadly distributed phylogenetically and that their closest relatives were related to sequences from coastal/estuarine sediments and oceanic water column sources. Conversely, they obtained no bacterial amoA gene sequences (Beman et al., 2007).

In our study, the most abundant OTU, ADSZ106 (119 clones at the DNA level and 1 clone at the transcript level), shared 93.96% similarity with uncultured archaeon clone S1– 24 (KC758384) from saltwater aquaria. The bacterial amoA gene in our investigation was primarily affiliated with the cluster of Nitrosomonadaceae at the DNA and transcript levels (**Figures 2A,B**), accounting for 71.42% of all OTUs (**Figure 2B**). Many investigations have showed that most cultured AOB belonged to the family Nitrosomonadaceae, phylum Betaproteobacteria (Koops and Pommerening-Roser, 2001). Moreover, based on the species features, such as affinity for ammonia, and tolerance to salt and nitrite, microbes in this taxon could be further subgrouped into several clusters (Koops et al., 2006).

At the DNA level, the AOA communities were not significantly different between the sampling locations (P > 0.05), while for AOB communities, there were significant difference between SYT and AT (P < 0.05) and between AT and ST (P < 0.05) (**Table 2**). For all ammonia-oxidizing prokaryotes, some OTUs were universally present in all samples, whereas others occurred in only a few samples. For instance, some unique OTUs, e.g., ASYTRS419, ASYTRS312, and SYTRS313, were detected only in the samples collected at SYT. Zhao et al. (2012) reported many human activities, such as overfishing, reef rock digging and tourism activities in Sanya Bay, and all of these factors in combination with climate change have led to a significant decline in coral cover since the 1960s. The seagrass T. hemprichii gradually colonized under these environmental conditions at that location.

#### Higher Transcriptional Activity of AOB Other than AOA

A positive correlation between amoA gene copy numbers and the potential nitrification rate has been recorded (He et al., 2007). Consequently, quantitative assays targeting amoA gene transcripts were carried out in this study to analyze potential

nitrification by AOA and AOB in coral reef ecosystems. The results showed that AOB would contribute more to the first step of nitrification for T. hemprichii (**Figure 1**). Furthermore, the results of an experiment conducted in the mangrove ecosystem were also consistent with our findings (Cao et al., 2015). However, Feng et al. (2016) obtained conflicting results for a marine sponge, for which the abundance of AOA was much higher than that of AOB at the cDNA level.

In our study, 24 OTUs and 14 OTUs were detected at the DNA level for AOA and AOB, respectively, and the number of OTUs decreased to 12 and 6, respectively, at the transcript level. This may be due to different preferences for ammonia. Hence, under the same condition, some species were dormant or below the PCR sensitivity threshold, whereas others exhibited high activity (Feng et al., 2016). For example, one unique bacterial amoA gene, OTU BDSS27, was detected only at the transcript level, and it was found to be related to uncultured ammonia-oxidizing bacterium clone ML-amoA-0 (FJ652557). A similar circumstance had been reported in the study of a marine sponge (Feng et al., 2016). This could be due to low abundance at the DNA level but high transcriptional activity. The most active amoA gene OTU in AOA communities was ADSR93, which was related to the uncultured archaeon clone GZ16110300849 (KY357274) obtained from mangrove sediments. Moreover, the most active bacterial amoA gene was the OTU BRAZ136 with 96 clones at the DNA level and 114 clones at the transcript level. Its closest BLAST hit was the uncultured bacterial clone HaAOB1 (JN177536) originating from the marine sponge Haliclona sp. collected from China East China at the depth of 20 m. Marine sponges are indispensable components of coral reef ecosystems and can help coral reefs thrive in ocean deserts because they absorb the nutrients from seawater and convert them into food for the reef and other marine organisms. Functional gene expression of ammoniaoxidizing microorganisms would also be altered by the health of their hosts. López-Legentil et al. (2010) reported that amoA gene expression was higher in fatally bleached sponges, whereas different patterns were observed in cyclic bleaching corals. Therefore, a higher abundance of AOB would have a more important function in transforming excess ammonia in the niche to maintain the healthy host.

#### CONCLUSION

We herein described the abundance, diversity and transcriptional activity of the AOA and AOB communities of the seagrass T. hemprichii in three coral reef ecosystems at the DNA and cDNA levels. The diversity of AOA communities was higher

#### REFERENCES

Albright, R., Caldeira, L., Hosfelt, J., Kwiatkowski, L., Maclaren, J. K., Mason, B. M., et al. (2016). Reversal of ocean acidification enhances net coral reef calcification. Nature 531, 362–365. doi: 10.1038/nature 17155

than that of AOB, though the abundance of AOB communities was greater than that of AOB, and the community compositions of the sampling locations were distinct. As the focuses of this study were community composition and potential AOA and AOB activity, the amount of ammonia oxidized by AOA and AOB for the growth of seagrass, which is crucial for discerning the roles of ammonia-oxidizing microbes in ecosystems, was not determined. Consequently, the pattern for protein expression pattern of the amoA gene product and the <sup>15</sup>N-isotope method would be used for elucidating their contributions to the seagrass productivity and their nitrogen transfer pathways for future investigations.

#### AUTHOR CONTRIBUTIONS

JL, LjL, and JD conceived the research. JL and XL performed the experiments. JL wrote the manuscript. YaZ and MA edited the manuscript. QY, LL, SZ, YiZ, and CW contributed to sampling or data analysis. All authors reviewed and approved the manuscript.

#### FUNDING

The research was supported by the National Natural Science Foundation of China (41676163, 41406191, 41676107 and 41276114), the Strategic Priority Research Program of the Chinese Academy of Sciences (13020300 and 11020202), National Key Research and Development Program of China (2017YFC0506301), the Guangdong Province Public Welfare Research and Capacity Building Project (2015A020216016), the Science and Technology Planning Project of Guangdong Province, China (2017B030314052), and the China Scholarship Council (201704910156).

#### ACKNOWLEDGMENTS

The authors would like to thank Dr. Lauren Hale at the University of Oklahoma for her helpful edits. They thank all of the members of the Tropical Marine Biological Research Station and Xisha/Nansha Ocean Observation and Research Station in Hainan for their work in the field sample collection.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.00007/full#supplementary-material


in the Gulf of California. ISME J. 2, 429–441. doi: 10.1038/ismej. 2007.118


Z. marina leaf or root microbiomes, vary in relation to distance from patch edge. Peer J. 27:e3246. doi: 10.7717/peerj.3246


fmicb-09-00007 January 23, 2018 Time: 17:52 # 11



**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 Ling, Lin, Zhang, Zhou, Yang, Lin, Zeng, Zhang, Wang, Ahmad, Long and Dong. 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.

# Marine Group II Dominates Planktonic Archaea in Water Column of the Northeastern South China Sea

Haodong Liu<sup>1</sup> , Chuanlun L. Zhang<sup>2</sup> \*, Chunyan Yang1,3, Songze Chen<sup>1</sup> , Zhiwei Cao<sup>4</sup> , Zhiwei Zhang<sup>5</sup> and Jiwei Tian<sup>5</sup>

<sup>1</sup> State Key Laboratory of Marine Geology, Tongji University, Shanghai, China, <sup>2</sup> Department of Ocean Science and Engineering, Southern University of Science and Technology, Shenzhen, China, <sup>3</sup> CNOOC Gas and Power Group, Beijing, China, <sup>4</sup> School of Life Sciences and Technology, Tongji University, Shanghai, China, <sup>5</sup> Physical Oceanography Laboratory, Ocean University of China, Qingdao, China

Temperature, nutrients, and salinity are among the important factors constraining the distribution and abundance of microorganisms in the ocean. Marine Group II (MGII) belonging to Euryarchaeota commonly dominates the planktonic archaeal community in shallow water and Marine Group I (MGI, now is called Thaumarchaeota) in deeper water in global oceans. Results of quantitative PCR (qPCR) and 454 sequencing in our study, however, showed the dominance of MGII in planktonic archaea throughout the water column of the northeastern South China Sea (SCS) that is characterized by strong water mixing. The abundance of ammonia-oxidizing archaea (AOA) representing the main group of Thaumarchaeota in deeper water in the northeastern SCS was significantly lower than in other oceanic regions. Phylogenetic analysis showed that the top operational taxonomic units (OTUs) of the MGII occurring predominantly below 200 m depth may be unique in the northeastern SCS based on the observation that they are distantly related to known sequences (identity ranging from 90–94%). The abundance of MGII was also significantly correlated with total bacteria in the whole column, which may indicate that MGII and bacteria may have similar physiological or biochemical properties or responses to environmental variation. This study provides valuable information about the dominance of MGII over AOA in both shallow and deep water in the northeastern SCS and highlights the need for comprehensive studies integrating physical, chemical, and microbial oceanography.

Keywords: South China Sea, Marine Group II, AOA, planktonic archaea, heterotrophic bacteria

#### INTRODUCTION

Microorganisms are the majority of life in the ocean and play fundamental roles in ecological functions and biogeochemical cycles (Fuhrman, 2009). Advances in genomics have revealed vertical zonation of planktonic microbial communities, which reflects the nature of ocean stratification (Giovannoni et al., 1996; DeLong et al., 2006; Shi et al., 2011). For example, the photic zone is characterized by steep gradients of light, temperature (thermocline), salinity (halocline), and nutrients (neutricline), which dictate the species distribution and function in the upper water column; in the aphotic zone (>200 m depth), decreasing temperature, increasing hydrostatic pressure and lack of light and energy supplies determine the microbial community structure

#### Edited by:

Stefan M. Sievert, Woods Hole Oceanographic Institution, United States

#### Reviewed by:

William D. Orsi, Lüdwig-Maximilians University of Munich, Germany Yonghui Zeng, Aarhus University, Denmark

> \*Correspondence: Chuanlun L. Zhang zhangcl@sustc.edu.cn

#### Specialty section:

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

Received: 16 December 2016 Accepted: 30 May 2017 Published: 15 June 2017

#### Citation:

Liu H, Zhang CL, Yang C, Chen S, Cao Z, Zhang Z and Tian J (2017) Marine Group II Dominates Planktonic Archaea in Water Column of the Northeastern South China Sea. Front. Microbiol. 8:1098. doi: 10.3389/fmicb.2017.01098

**128**

and function of the dark ocean (DeLong et al., 2006). However, other physical processes such as water mass movement and mesoscale eddies have also been reported to control the distribution of microbial populations or activity in the ocean (Galand et al., 2008, 2010; Zhang et al., 2009; Chen et al., 2016).

Planktonic archaea have been recognized to play important roles in global carbon and nitrogen cycles (Karner et al., 2001; Francis et al., 2005; Ingalls et al., 2006). They were initially divided into Marine Group I (MGI; now called Thaumarchaeota) and Marine Group II (MGII) that belong to Euryarchaeota (DeLong, 1992); the latter has generally been observed to dominate the surface ocean in archaeal composition, whereas the former becomes increasingly abundant at greater depths (Massana et al., 2000; Karner et al., 2001; Herndl et al., 2005; Lincoln et al., 2014). While tremendous progress has been made in the physiology, biochemistry, and ecological functions of Thaumarchaeota (Konneke et al., 2005; Ingalls et al., 2006; Martens-Habbena et al., 2009), our understanding of the MGII in the archaeal domain remains fragmented (Zhang C.L. et al., 2015). MGII have been classified into four groups (MGIIA, MGIIB, MGIIC, and MGIID) according to their 16S rRNA gene sequences (Martin-Cuadrado et al., 2015). However, there is no pure culture of MGII at the present. Recently, the metagenomic and transcriptomic studies of MGII have been increasing and gradually unveiling their potential ecological functions in carbon and nitrogen cycling in the ocean. This is exemplified by reports on the capability of MGII in degradation of protein and lipids (Iverson et al., 2012), synthesization of archaeal tetraether lipids (Lincoln et al., 2014), utilization of dissolved protein (Orsi et al., 2015), attachment and utilization of particulate organic matter (Orsi et al., 2016), and harvesting solar energy in the photic zone using proteorhodopsin (Iverson et al., 2012; Li et al., 2015; Orsi et al., 2015, 2016). However, the mechanisms controlling the distribution of MGII in different water columns of the ocean are poorly known.

The South China Sea (SCS) is the largest marginal sea of the northwestern Pacific, which has recently witnessed significant growth in microbial and biogeochemical studies in this oceanographic region (Liu et al., 2007; Moisander et al., 2008; Zhang et al., 2009; Hu et al., 2011; Wei et al., 2011; Jia et al., 2012; Jiao et al., 2014a; Tseng et al., 2015; Xia et al., 2015). The water column dynamics of the SCS is regulated by complex basin topography and water circulations resulting from East Asian Monsoon activities and the Pacific Kuroshio current intrusion (Xu and Oey, 2015; Zhang Z. et al., 2015). Occurrence of surface-confined phototrophic populations in deep waters has been observed in the western Pacific, Luzon Strait, and the SCS, which may be attributed to the active vertical mixing and isopycnal heaving of water associated with internal solitary waves, mesoscale eddies, and/or other physical processes (Jiao et al., 2014b; Chen et al., 2016). A recent study also observed the impact of asymmetrical internal solitary waves on temperature, nutrients, and chlorophyll a in the northern SCS (Dong et al., 2015).

Although the archaeal and bacterial community structures have been reported to be influenced by strong internal waves in the western Pacific Ocean near Luzon Strait (Jiao et al., 2014a), as well as to be affected by mesoscale cyclonic eddies in the western and the central northern SCS, the distributional patterns and niche specificity of archaea in the northeastern SCS are still unknown. The aim of this study was to unveil the relative abundance and distribution of ammonia-oxidizing archaea (AOA) and bacteria (AOB), MGII and total bacteria by targeting the archaeal and bacterial ammonia monooxygenase (amoA) genes and the 16S rRNA genes of MGII and bacteria, respectively, using qPCR. The archaeal community structure was also determined by using 454 sequencing. Overall, we examined that the archaeal community structure showed great similarity between different water depths in the northeastern SCS, which was different from the archaeal community structure presented in western and central regions of the SCS (Zhang et al., 2009). We also observed that the relative abundance of AOA was low throughout the whole water column (**Figures 2**, **3**) compared to other regions of the SCS (Hu et al., 2011). These results collectively suggest that the predominance of MGII in archaeal composition throughout the water column of the northeastern SCS may be caused by strong vertical mixing in this region (Tian et al., 2009; Jiao et al., 2014a).

#### MATERIALS AND METHODS

#### Field Work and Sample Collection

Twenty-four water samples were collected at different depths at the D stations (D3 and D5 with maximum depths ranging from 1800 to 3100 m, between 19◦ 380N and 117◦ 490E and 20◦ 030N and 117◦ 250E) and 49 samples at the B stations (B2, B3, B6, and B7 with maximum depths ranging from 1800 to 3200 m, between 20◦ 450N and 119◦ 480E and 21◦ 510N and 118◦ 260E) in April 2013 in the northeastern SCS (**Figure 1**). All water samples were collected by using 12-liter Niskin bottles attached to a CTD equipment; 1–2 l of sea water were filtered through a 0.22-µm membrane filter (Nitrocellulose Membrane, Millipore GSWP04700) using a vacuum pumping system, which collected both particleassociated and free-living archaea and bacteria. Filters were not exposed to air during filtration. After filtration, the membrane was preserved at −20◦C immediately. Data of depth, temperature, and salinity were recorded by a CTD recorder (model SBE 9-11 Plus, SeaBird Electronics, Inc., United States).

#### DNA Extraction and qPCR

Filters were cut into small pieces using sterilized scissors, which were then transferred to 2 mL tubes. DNA extraction was performed following manufacturer's instructions provided by the FastDNA Spin Kit for Soil (MP Biomedical, Solon, OH, United States). The bacterial and MGII 16S rRNA genes and archaeal and β-AOB amoA genes were quantified on all samples by qPCR (PIKO REAL 96, Thermo Fisher Scientific). The abundance of

and CorelDRAW Graphics Suite X7 (http://www.coreldraw.com/cn/).



each gene from each sample was normalized according to the dilution folds of DNA template and the volume of collected water. The details of primers used for qPCR were shown in **Table 1**. Each 10 µl qPCR solution consisted of 1 µl (∼1 µM) template DNA, 5 µl SYBR Premix Ex TaqTM II (TaKaRa Biotechnology Co.), 0.2 µl each primer (∼1 µM), 0.1 µl Bovine Serum Albumin (BSA, 20 mg/mL) solution (TaKaRa Biotechnology Co.) and 3.5 µl deionized water. The condition was as follows: 95◦C for 30 s; 40 cycles at 95◦C for 5 s, 55◦C for 30 s, and 72◦C for 1 min. All three genes in this study were determined in triplicates for each sample. The amplification efficiency of archaeal amoA gene was around 95% and the R square was greater than 0.99. The amplification efficiencies of MGII and bacterial 16S rRNA genes were ∼96% and ∼85%, respectively, and the R square of them was greater than 0.99 and 0.98, respectively.

#### 454 Pyrosequencing

Pyrosequencing of B7 and D5 samples was conducted with a Roche 454 GS FLX+ Titanium platform (Roche 454 Life Sciences, Branford, CT, United States) at the Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China). The extracted DNA was amplified using a universal primer set ARCH344f/ARCH915r (**Table 1**). Unique barcodes for each sample were added at the 5<sup>0</sup> -end of both the forward and reverse primers to demultiplex sequences. The resulting sequences were processed using split\_libraries.py-split libraries<sup>1</sup> in QIIME (version 1.8.0). Sequences with quality scores greater than 20 were kept. Sequences containing ambiguous base calls or being shorter than 200 bp or with homopolymers longer than six nucleotides were discarded. Chimeric sequences were identified and removed

<sup>1</sup>http://qiime.org/scripts/split\_libraries.html

using UCHIME (version 4.2.40<sup>2</sup> ). Then the processed sequences were clustered using the USEARCH (UCLUST) and operational taxonomic units (OTUs) were defined at the 97% similarity cutoff by using the UPARSE (version 7.1<sup>3</sup> ). OTU representative sequences were then selected and the taxonomy was assigned using the ribosomal database project (RDP) classifier algorithm against the SILVA (SSU115) 16S rRNA database using confidence threshold of 70% in the QIIME program. Singleton and bacterial sequences were removed. After these quality control procedures, 43,429 sequences were removed from 203,712 sequences and 160,283 high-quality sequences were produced with an average length of ∼380 bp. The OTU table was rarefied to equal sequence number for each sample basing on the least sequencing depth (n = 2000) sample B7\_300 m (98.35% sequencing coverage) that was iterated 1000 times. The alpha diversity was calculated at the 97% identity level in QIIME, which included Shannon, Simpson, Chao1, and ace.

## Accession Numbers

The sequence data generated in this study were deposited at the Sequence Read Archive (SRA) in the National Center for Biotechnology Information (NCBI) under the BioProject accession no. SRP072671 with BioSamples SRS2071738-SRS2071762<sup>4</sup> .

## Methods for Data Analysis and Figure Generation

After removing singletons, 388 OTUs were obtained. The top 30 OTUs were selected to compare against blast online and their most similar reference sequences were picked out. Then these sequences were aligned by using ClustalW with default parameter settings. The Neighbor-Joining tree was constructed based on these aligned sequences with 1000 bootstrap values using the MEGA software<sup>5</sup> . The Bray–Curtis similarity matrix analyses were performed using the PAST<sup>6</sup> , with the outcome being displayed by using the HemI<sup>7</sup> .

## RESULTS

Temperature and salinity profiles were similar among the B and D stations (**Supplementary Figure S1**), which are typically observed in the SCS. In this study, we focused on describing the microbiological variation in Thaumarchaeota, MGII and bacteria among these stations. The abundance of Thaumarchaeota was estimated using the archaeal amoA gene based on the consensus that the ratio of archaeal amoA gene to Thaumarchaeotal 16S rRNA gene is between 1 and 2 in the open Ocean (Church et al., 2010; Hu et al., 2011; Lund et al., 2012).

#### Variation in Abundance of Thaumarchaeota, MG II, and Bacteria with Depth

The abundance of archaeal amoA gene was low (104–10<sup>5</sup> copies per liter seawater) at the surface and increased to 106–10<sup>7</sup> copies per liter seawater all around 100 m at the B and D stations (**Figure 2**). It then decreased to 103–10<sup>4</sup> copies per liter seawater below 1000–1500 m at the B and D stations; the exception was at B3, which showed consistent trend of increase again in the archaeal amoA gene abundance below 1500 m (**Figure 2**). The bacterial amoA gene depth profile was not shown because only one out of 60 samples contained detectable bacterial amoA gene in this study.

The MGII and bacterial 16S rRNA gene copies showed general decreasing trend with depth at all stations; exceptions were at B3 and B7, which showed increase again in abundance of both genes below about 1500 m, and at D5, which showed increase again from 2200 to 2800 m (**Figure 2**). These two genes had maximal abundance (10<sup>8</sup> copies per liter seawater) at surface (5 m) or subsurface (50–100 m) and one to three orders of magnitude lower values (105–10<sup>7</sup> gene copies per liter seawater) at the bottom of each station; the exception again was at B7, which reached the maximal abundance of bacterial 16S rRNA gene at the bottom. The minimal copies of each gene, however, occurred at different depths at different locations.

#### Distribution of Different Groups of Archaea with Depth

The 16S rRNA gene sequencing analysis showed that MGII accounted for the most abundant proportion of archaea throughout the water column at both B7 and D5 stations, which were followed by MGIII; these two groups all together accounted for 91.8–99.4% of total archaeal sequences (**Figure 3** and Supplementary Table S1). The dominant clades of MGII were MGIIB and MGIIA in our study. Thaumarchaeota remained below 3.5% at all depths at these two stations, except at the 3200 m depth at B7, which exceeded 5.0% of total archaeal sequences (**Figure 3** and Supplementary Table S1).

The MBG-A (≤1.5% of total archaeal sequences), and Woesearchaeota (up to 1.8% of total archaeal sequences) were also detected at low relative abundance at the B7 and D5 stations. Other unclassified archaea collectively accounted for less than 2–4% of total archaeal sequences at these stations (Supplementary Table S1).

#### Distribution of Different Groups of Archaea at the OTU Level at Different Depths

The total number of archaeal sequences was 94,495 for station B7 samples and 65,788 for D5 samples. As a result, the OTUs (at 97% cut off) were 354 for B7 samples and 339 for D5 samples. Because samples at the B7 and D5 stations showed similar vertical distribution in archaeal community composition, the sequences were combined. The top 30 archaeal OTUs from the combined B7 and D5 stations were selected for the construction

<sup>2</sup>http://drive5.com/usearch/manual/uchime\_algo.html

<sup>3</sup>http://drive5.com/uparse/

<sup>4</sup>https://www.ncbi.nlm.nih.gov/Traces/study/?acc=SRP072671

<sup>5</sup>http://www.megasoftware.net/

<sup>6</sup>http://folk.uio.no/ohammer/past

<sup>7</sup>http://hemi.biocuckoo.org/

of the phylogenetic tree, which represented 67% of total archaeal sequences that were similar to the percentage of the top 30 OTUs calculated separately for B7 samples (65% of total archaeal sequences) and D5 samples (70% of total archaeal sequences).

The top 30 OTUs from the combined B7 and D5 stations were all from MGII (22) or MGIII (8). The 22 MGII OTUs represented 90,826 sequences, which accounted for 78.0% of the total MGII sequences (116,391). The 8 MGIII OTUs represented

populations.

fmicb-08-01098 June 13, 2017 Time: 18:9 # 6

16,559 sequences that accounted for 43.8% of the total MGIII sequences (37,781).

Within the MGII, OTU-1 had the largest number of sequences (35,903) that accounted for 39.5% of total MGII sequences. OTUs-2,-3,-4, and -5 had 4.5–8.0% of total MGII sequences with a total percentage of 24.5%. The remaining OTUs had 1.2–3.5% of total MGII sequences.

Operational taxonomic units-1 and -2 occurred predominantly (>96% of sequences) below the photic zone (300–3200 m). OTUs-4 and -5 occurred largely (>69% of sequences) within the photic zone but also had substantial presence (18.3–20.7% of sequences) at deeper (1200–2000 m or 2200–3200 m) water depths (**Figure 4**). OTU-3 occurred more or less evenly throughout the water column. Other OTUs belonged to one of the above three categories. For example, OTUs-23, -26, -27, and -28 overwhelmingly (>99% of sequences) occurred in the photic zone, particularly the shallower (<75 m) water depths; OTUs-14, -20, and -21 occurred predominantly below 300 m, particularly in the 2200–3200 m depth interval (**Figure 4**).

Most of the MGII OTUs were closely affiliated with sequences identified from other environments (the Gulf of Mexico, the Mediterranean Sea, pelagic oxygen minimum zone, Pacific surface water, the Pearl River estuary, or the Arabian Sea). Nine of the 22 OTUs, including OTUs-1, -2, and -3, however, formed a cluster that was distantly related to sequences from other environments (**Figure 4**). These OTUs also showed a varying distribution with depth, with some of them predominantly occurring in shallow water depths, others in deeper water depths and still others occurring more or less evenly through the water column (**Figure 4**).

Within the MGIII, each OTU represented 5.8–19.1% of total sequences with an average of 12.5 ± 4.5% per OTU. Similar to the distribution of MGII OTUs, MGIII OTUs-9, -18, -19, -22, and -30 had sequences mostly occurring in the deep water, particularly in the 2200–3200 m depth interval, whereas OTUs-7 and -16 had sequences mostly occurring in the shallow water; OTU-13 had sequences occurring more or less throughout the water column, although the deeper waters tended to have greater numbers of sequences (**Figure 4**). These OTUs were similar to reference sequences from other open oceans and, unlike some of the MGII OTUs, didn't form any unique cluster.

Results of the Bray–Curtis analysis also demonstrated the similarity in OTUs between shallower and deeper depth intervals for both MGII and Thaumarchaeota. For example, at B7, the distribution of MGII OTUs from above 300 m showed 40–60% similarity to those from either 2000 or 3200 m depth, whereas the distribution similarity between 800 and 3000 m appeared to be much greater (50–80%). At D5, the distribution of MGII OTUs from 900 m showed about 80% similarity with the distribution of MGII OTUs from 3100 m; whereas, the distribution of MGII OTUs from 700 m showed 60–80% similarity with that from 1200 to 2800 m (**Figure 5A**). In general, the distribution similarity matrixes of OTUs in Thaumarchaeota at both stations are similar to those in MGII (**Figure 5B**), which may be attributed to similar influence by water mixing or organic matter properties, or both.

It is worth mentioning that OTUs of MGII in the top 100 m (B7) depth intervals showed greater distribution similarity than those from most depths below; OTUs of MGII from the top 5 m at D5, however, did not show distribution similarity with those from any depth below (**Figure 5A**), which suggests that surface water or water in the upper photic zone had less mixing with water from deeper depths.

We also performed the Bray–Curtis analyses on other groups (MGIII, Woesearchaea, MBG-A) of archaea, with most of the groups showing similarly well mixing features across the water column, either from surface (5 m) or below 100 m as defined above (**Supplementary Figure S2**).

#### Correlation between MGII and Bacterial 16S rRNA Gene Copies

A total of 73 samples in 6 stations were included to compare the relationship between MGII and bacterial 16S rRNA gene copies. Significant correlation between the logarithmic values of the two kinds of genes existed from the surface to bottom water in the northeastern SCS (R <sup>2</sup> > 0.76, P < 0.01). The average logarithmic value of bacterial 16S rRNA gene copies per liter seawater was 7.23 and the average logarithmic value of MGII 16S rRNA gene copies per liter seawater was 6.39.

FIGURE 4 | Phylogenetic tree based on the 16S rRNA gene from the Northeastern SCS (A: B7 and D5 stations). Support values, with 1000 replicates for Neighbor-Joining (NJ) analyses, were shown in the order of NJ at nodes (values lower than 50% are not shown). The numbers of environmental sequences of top OTUs recovered in this study were shown in the brackets. Dark blue OTUs indicate sequences mainly distributed in the photic zone; red OTUs indicate sequences mainly distributed below the photic zone; and cyan OTUs indicate sequences distributed relatively evenly in the photic and aphotic zones. Reference sequences from NCBI database were shown in bold. The scale bar indicates 0.02 nucleotide substitutions per site. The distribution of the dominant OTUs of MGII and MGIII is shown at right, where the circle size indicates the relative abundance of sequences in each OTU at different depth intervals.

## DISCUSSION

The ocean is characteristically stratified, which is reflected in much stronger difference in microbial community structure vertically than horizontally (DeLong et al., 2006; Shi et al., 2011). In this study, the distribution of Thaumarchaeota, MGII, and MGIII in the northeastern SCS (B and D stations) has distinct patterns from that in other areas of the SCS (Zhang et al., 2009; Tseng et al., 2015; Xia et al., 2015). In particular, the relatively abundant MGII and MGIII distribution with depth at the B and D stations has not been reported in previous studies that mostly describe MGII being predominant in surface water whereas Thaumarchaeota or MGIII in deep water (López-García et al., 2001; Herndl et al., 2005; DeLong et al., 2006; Galand et al., 2009; Tseng et al., 2015; Zhang C.L. et al., 2015).

Jiao et al. (2014a) have reported the presence of surface picoplankton (Prochlorococcus) in deep waters in Luzon Strait in the western Pacific Ocean, who dismissed aggregation, particle packing through grazing and egestion, or winter ventilation as the mechanisms for picoplankton transport to the deep water. It is also counterintuitive that organic matter is enriched in falling particles with increasing depth because the concentration of POC has been observed to decrease with depth (Dai et al., 2009). However, it is possible that the lability of organic matter differed with changing depths at these stations, as the type of organic matter can certainly influence the physiological properties of these proposed heterotrophs. More likely, the transportation of surface water microorganisms was attributed to multiple physical processes such as internal solitary waves, meso-scale eddies and turbulent mixing in the Luzon Strait and the northeastern SCS

(Jiao et al., 2014a). A recent report (Chen et al., 2016) also indicated that internal solitary waves can enhance heterotrophic bacterial growth in the northern SCS. These studies highlight the importance of physical processes in controlling the distribution of planktonic microorganisms in the SCS.

The same physical processes reported by Jiao et al. (2014a) may be responsible for the transport of shallow water MGII groups down to the deep water as well as the abundant presence of MGIII throughout the water column in the northeastern SCS, which can be inferred by the similar depth profiles and similarity matrixes of different archaeal groups. Furthermore, from the available data used in the construction of the phylogenetic tree of MGII and MGIII, sequences from the SCS are more frequently affiliated with those from the Pacific Ocean than from other regions (**Figure 4**), suggesting the impact of mixing between the Pacific and the SCS waters on the community structure of planktonic archaea in the latter. On the other hand, we cannot exclude the contribution of the gravitational falling of particleattached microbes on the occurrence of MGII at greater depths.

The low abundance of autotrophic AOA (Thaumarchaeota) throughout the whole column in the northeastern SCS is in contrast to the peak abundance of AOA occurring at 50–200 m depths, which is 5- to 10-fold higher than surface AOA observed in other regions of the SCS (Hu et al., 2011). Furthermore, the low abundance of AOA (Thaumarchaeota) compared to MGII and the scarce abundance of AOB in the northeastern SCS may be possibly due to the overall oligotrophic environment that is particularly depleted in ammonium. Hu et al. (20110) also showed the scarce AOB in SCS, in which bacterial amoA gene was only detected in seven out of 26 samples (Hu et al., 2011). In our study, only one out of 60 samples contained detectable bacterial amoA genes. AOB are commonly found in soils, freshwater, estuaries, hot springs, and marine environments (Francis et al., 2005). They appear to be more abundant than AOA in coastal settings where ammonium was relatively higher than in the oligotrophic ocean (Fan et al., 2015). AOA on the other hand are more adapted to oligotrophy in the ocean because they possess much higher affinity for ammonia than AOB (Martens-Habbena et al., 2009). It has been reported that the concentration of ammonia concentration in the area near our study sites ranged from 0.08 to 0.38 µM, which is lower than other regions of the SCS (Ling, 2011). The threshold of ammonia concentration for the growth of AOA is as low as 10 nM while the minimum ammonia concentration for the growth of AOB is greater than 1 µM observed under culture conditions (Martens-Habbena et al., 2009). Thus, the lack of ammonia probably was the main reason why AOB were almost absent in our study area. On the other hand, the low abundance of AOA throughout the whole water column in the northeastern SCS may not be due to the low abundance of ammonia alone; vertical mixing could be another reason, which could homogenize low AOA surface water with relatively more AOA abundant deeper water, a hypothesis that can be tested in future studies.

It is intriguing to observe the significant correlation (R <sup>2</sup> = 0.76, **Figure 6**) between the abundance of MGII 16S rRNA gene copies and the abundance of bacterial 16S rRNA gene copies in the northeastern SCS. Significant correlation between MGII and total bacteria has also been observed in western and northern central SCS water column (Zhang et al., 2009) and our work in the western Pacific also supports this observation (unpublished data). Thus it is likely this correlation holds true in other regions of the global ocean. At the moment, however, we can only speculate the possible mechanisms underlying this correlation. One possibility is that the occurrence of MGII and bacteria is controlled by a common variable, for example the affinity to particles. This may be supported by the observation of Orsi et al. (2015), which showed that both MGII and heterotrophic bacterial groups prefer to attach to particles in the ocean waters. Unfortunately our study only used 0.2 µm filters for sample collection, which included possibly both free-living and particle attached bacteria and MGII. Future research is needed to test this hypothesis.

In summary, our results showed that the archaeal distributional patterns in the northeastern SCS water column were distinct from other marine regions. The most abundant archaeal MGII subgroups were MGIIA (OTU 23, 26, 27 in **Figure 4**) and MGIIB (other MGII OTUs in **Figure 4**), with the former clade being mainly present at shallow waters (<100 m) while the later clade being widely present in both shallow and deep waters. In general, the heterotrophic archaea represented by MGII were much more abundant than the autotrophic archaea Thaumarchaeota throughout the whole water column in the northeastern SCS. The exact mechanisms controlling the particular archaeal distribution patterns and community structure remain unclear because of lacking direct physical and chemical measurements as well as RNA analysis. Future studies will need to couple microbiological sampling with measurements of physical and chemical properties in time series in the SCS, which should shed light on or strengthen our understanding of how heterotrophic and autotrophic archaea respond to the changing marine environment.

#### AUTHOR CONTRIBUTIONS

fmicb-08-01098 June 13, 2017 Time: 18:9 # 10

CZ and HL developed the idea and designed the study. HL, CY, and SC processed and analyzed the data. HL and CZ wrote the manuscript. ZZ and JT contributed to the discussion on physical processes in the SCS. ZC contributed to the data analysis.

#### FUNDING

This research was supported by the Ministry of Science and Technology Award No. 2013CB955703 and 2016YFA0601101, the National Science Foundation of China Award Nos. 41530105, and 91428308, the Tongji Interdisciplinary Program No # 1350219165, and the "National Thousand Talents" program through the State Key Laboratory of Marine Geology at Tongji

## REFERENCES


University. This study is also a contribution to the international IMBER project.

#### ACKNOWLEDGMENTS

We thank Fengfeng Zheng, Weiyan Wu, Xiaotong Tang and Liang Dong for sampling. Francisco Rodriguez-Valera and Meng Li provided valuable comments on an earlier version of the manuscript. The comments provided by the two reviewers and the handling editor are greatly appreciated.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.01098/full#supplementary-material

FIGURE S1 | The temperature, salinity and depth profiles of B and D stations. FIGURE S2 | OTU similarity matrixes of MGIII, Woesearchaeota and MBGA using the Bray–Curtis method.

to the green non-sulfur bacteria. Proc. Natl. Acad. Sci. U.S.A. 93, 7979–7984. doi: 10.1073/pnas.93.15.7979


assemblages in the open ocean. ISME J. 5, 999–1013. doi: 10.1038/ismej. 2010.189


**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 Liu, Zhang, Yang, Chen, Cao, Zhang and Tian. 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.

fmicb-08-01098 June 13, 2017 Time: 18:9 # 11

# Evaluating Production of Cyclopentyl Tetraethers by Marine Group II Euryarchaeota in the Pearl River Estuary and Coastal South China Sea: Potential Impact on the TEX<sup>86</sup> Paleothermometer

Jin-Xiang Wang1, 2†, Wei Xie3†, Yi Ge Zhang<sup>4</sup> , Travis B. Meador <sup>1</sup> and Chuanlun L. Zhang3, 5 \*

*<sup>1</sup> MARUM-Center for Marine Environmental Sciences, University of Bremen, Bremen, Germany, <sup>2</sup> Department of Marine Sciences, University of Georgia, Athens, GA, United States, <sup>3</sup> State Key Laboratory of Marine Geology, Tongji University, Shanghai, China, <sup>4</sup> Department of Oceanography, Texas A&M University, College Station, TX, United States, <sup>5</sup> Department of Ocean Science & Engineering, Southern University of Science and Technology, Shenzhen, China*

#### Edited by:

*Stefan M. Sievert, Woods Hole Oceanographic Institution, United States*

#### Reviewed by:

*Florence Schubotz, University of Bremen, Germany Anitra E. Ingalls, University of Washington, United States*

#### \*Correspondence:

*Chuanlun L. Zhang zhangcl@sustc.edu.cn*

*† These authors have contributed equally to this work.*

#### Specialty section:

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

Received: *31 May 2017* Accepted: *10 October 2017* Published: *31 October 2017*

#### Citation:

*Wang J-X, Xie W, Zhang YG, Meador TB and Zhang CL (2017) Evaluating Production of Cyclopentyl Tetraethers by Marine Group II Euryarchaeota in the Pearl River Estuary and Coastal South China Sea: Potential Impact on the TEX86 Paleothermometer. Front. Microbiol. 8:2077. doi: 10.3389/fmicb.2017.02077*

TEX<sup>86</sup> [TetraEther indeX of glycerol dialkyl glycerol tetraethers (GDGTs) with 86 carbon atoms] has been widely applied to reconstruct (paleo-) sea surface temperature. Marine Group I (MG-I) *Thaumarchaeota* were thought to be the primary source of GDGTs constituting the TEX<sup>86</sup> formula; however, recent research has suggested that Marine Group II (MG-II) *Euryarchaeota* may also contribute significantly to the GDGT pool in the ocean. Little is known regarding the potential impact of MG-II *Euryarchaeota*-derived GDGTs on TEX<sup>86</sup> values recorded in marine sediments. In this study, we assessed the relationship between distributions of GDGTs and MG-II *Euryarchaeota* and evaluated its potential effect on the TEX<sup>86</sup> proxy. Lipid and DNA analyses were performed on suspended particulate matter and surface sediments collected along a salinity gradient from the lower Pearl River (river water) and its estuary (mixing water) to the coastal South China Sea (SCS, seawater). TEX86-derived temperatures from the water column and surface sediments were significantly correlated and both were lower than satellite-based temperatures. The ring index (RI) values in these environments were higher than predicted from the calculated TEX86-RI correlation, indicating that the GDGT pool in the water column of the PR estuary and coastal SCS comprises relatively more cyclopentane rings, which thereby altered TEX<sup>86</sup> values. Furthermore, the abundance of MG-II *Euryarchaeota* 16S rRNA gene in the mixing water was two to three orders of magnitude higher than those observed in the river or seawater. Significant linear correlations were observed between the gene abundance ratio of MG-II *Euryarchaeota* to total archaea and the fractional abundance of GDGTs with cyclopentane rings. Collectively, these results suggest that MG-II *Euryarchaeota* likely produce a large proportion of GDGTs with 1–4 cyclopentane moieties, which may bias TEX<sup>86</sup> values in the water column and sediments. As such, valid interpretation of TEX<sup>86</sup> values in the sediment record, particularly in coastal oceans, should consider the contribution from MG-II *Euryarchaeota*.

Keywords: Marine Group II, Euryarchaeota, GDGTs, TEX86, ring index, South China Sea

#### INTRODUCTION

TEX<sup>86</sup> is a popular temperature proxy applied in paleoclimatological studies, which is based on the relative distribution of cyclopental rings among isoprenoid glycerol dialkyl glycerol tetraether (GDGT; **Figure S1**) lipids produced by archaea in marine and terrestrial environments (see review by Schouten et al., 2013). Global core-top calibrations of TEX<sup>86</sup> values were empirically correlated with the annual mean sea surface temperature (SST; Schouten et al., 2002; Kim et al., 2008, 2010). However, mounting evidence indicates anomalies of TEX86-derived SST in coastal seas and the open ocean, which have been attributed to multiple inputs of GDGTs from terrestrial (e.g., Weijers et al., 2006) or bathypelagic sources (e.g., Lee et al., 2008), as well as production in marine sediments (e.g., Liu X. L. et al., 2011).

A great deal of effort has been made to assess TEX<sup>86</sup> accuracy in marine and lake sediments. For example, application of the TEX<sup>86</sup> proxy is cautioned under the following circumstances: a branched and isoprenoid tetraether (BIT) index > 0.2 (Zhu et al., 2011), a ratio of GDGT-2/crenarchaeol > 0.4 (Weijers et al., 2011a), a Methane Index > 0.5 (Zhang et al., 2011), a ratio of GDGT-0/crenarchaeol > 2 (Blaga et al., 2009), or when %GDGT-2 > 45 (Sinninghe Damsté et al., 2012). Recently, based on an assessment of the relationship between the weighted average number of cyclopentane rings in GDGTs (ring index, RI) and published TEX<sup>86</sup> data from core-top sediments, Zhang et al. (2016) established a significant correlation between TEX<sup>86</sup> and RI [RI = 3.32 × (TEX86) <sup>2</sup> − 0.77 × TEX<sup>86</sup> + 1.59; ±2σ ∼ 0.3]. This relationship was expected and reflects the physiological response of marine archaea to synthesize GDGTs with more rings (higher RI values) at higher temperatures (higher TEX<sup>86</sup> values). Deviations from this relationship suggest that temperature is not a dominant factor governing GDGT ring distribution, or, alternatively, the relationship between GDGT ring distribution and temperature is different from the modern analog as defined by the global core-top dataset.

The TEX86-related GDGTs (GDGTs-1, -2, -3 and the crenarchaeol regioisomer) in the water column are thought to primarily derive from the Marine Group I (MG-I) Thaumarchaeota (e.g., Schouten et al., 2008; Pitcher et al., 2011a,b), as it is one of the dominant groups of planktonic archaea in the ocean (Karner et al., 2001). In particular, crenarchaeol, containing one cyclohexane and four cyclopentane moieties, is accepted as a specific biomarker for MG-I Thaumarchaeota (Sinninghe Damsté et al., 2002; Schouten et al., 2008). Marine Group II (MG-II) Euryarchaeota are another group of planktonic archaea that predominantly inhabit coastal water and (near) surface waters of the open ocean (e.g., DeLong, 1992; Galand et al., 2010; Hugoni et al., 2013). Recently, this cosmopolitan group was also invoked as another major source of GDGTs (including crenarchaeol) in the ocean (Lincoln et al., 2014a), which supports an earlier hypothesis about GDGT-producing MG-II Euryarchaeota (Turich et al., 2007). However, concrete evidence of crenarchaeol production by MG-II is lacking due to the inability to obtain a pure culture or isolate, and the exact composition of GDGTs produced by these

To further evaluate the relationship between MG-II Euryarchaeota and the distribution of GDGTs, we quantified the abundance of MG-II 16S rRNA gene, determined the distribution of GDGT core and intact polar lipids (CL and IPL, respectively), and assessed TEX<sup>86</sup> and Ring Index (RI) values from suspended particulate matter and surface sediments collected from the lower Pearl River (PR) and its estuary to the coastal South China Sea (SCS). Our results provide a mechanistic explanation for deviations of the TEX<sup>86</sup> paleothermometer and have important implications for the sources of GDGTs in marine environments and probing past changes in global climate.

#### MATERIALS AND METHODS

#### Sample Collection

Sampling locations and other information for suspended particulate matter (SPM) and surface sediments are shown in **Figure 1** and **Table 1**. SPM samples (n = 18) and surface sediments (n = 8) were collected along a salinity gradient from the lower Pearl River and its estuary to the coastal South China Sea in the summer of 2011. SPM samples were collected from the surface (stations R1–R6) and the bottom (stations R1 and R2) water in the lower Pearl River, from three water depths (surface, middle, and bottom) and during three tidal periods (high tide, slack tide, and low tide) at station M located in the PR estuary, and at four water depths (surface, subsurface, middle, and bottom) at station S in the seawater of the coastal SCS (**Figure 1**). The depth of the sampling layers in the water column is given in **Table 1**. About 4–103 liters of water were filtered onto combusted (450◦C, overnight) glass-fiber filters (Whatman GF/F, 0.7µm, 142 mm diameter) using an in situ submersible pump. The pH, temperature, salinity, and depth were determined in situ by a Horiba instrument (W-20XD, Kyoto, Japan; **Table 1**). Surface sediments (top ca. 10 cm) were collected at all stations (**Figure 1**; **Table 1**) using a grab sampler. All samples were frozen immediately in liquid nitrogen and kept at −80◦C in the laboratory before analysis.

#### GDGT Extraction and Separation

The SPM samples (n = 18) and surface sediments (n = 8) were freeze-dried and extracted using a modified Bligh and Dyer method (Bligh and Dyer, 1959); the separation of core lipids and intact polar lipids followed the procedure described in Weijers et al. (2011b). Briefly, the total lipid extract (TLE) was obtained by ultrasonic extraction (10 min each, 6 times) of SPM (1 filter) or sediments (5 g) with a single-phase solvent mixture of methanol, dichloromethane (DCM), and phosphate buffer (2:1:0.8, v/v/v; pH 7.4). The TLE was separated over an activated silica gel column eluted with n-hexane/ethyl acetate (1:1, v/v) and methanol for CL and IPL, respectively. For GDGT quantification, a known amount of an internal C46 GDGT standard was added into the CL fraction or IPL fraction (Huguet et al., 2006). The CL fraction was directly measured. The IPL fraction was determined by measuring the CL after cleavage of the polar head groups via acid or base hydrolysis (Pitcher

et al., 2009; Weijers et al., 2011b). Briefly, 1/3 IPL fraction (nonhydrolyzed IPL fraction) was directly condensed; another 1/3 IPL fraction was hydrolyzed (2 h) in 1.5 N HCl in methanol, which was called the acid-hydrolyzed IPL fraction (total IPL). DCM and MilliQ water were added, and the DCM fraction was collected (repeated 4 times). The DCM fraction was rinsed (6 times) with MilliQ water in order to remove acid and dried under N<sup>2</sup> gas. The last 1/3 IPL fraction was subjected to base hydrolysis (2 h) in a 1N KOH in methanol/H2O mixture (95:5, v/v), which was called the base-hydrolyzed IPL fraction (phospho IPL). Together with the condensed CL fraction, the CL fraction, two fractions of IPL-derived core lipids and non-cleaved IPL fraction were dissolved in n-hexane/isopropanol (99:1, v/v), and filtered using PTFE filters (pore diameter of 0.45µm). The acidhydrolyzed IPL reflected total IPL, which, prior to hydrolysis,

R2 include two water layers (surface and bottom). Surface sediments were collected at each sampling site.

were attached to both phosphatidyl and glycosidic head groups; The base-hydrolyzed IPL represented IPLs with phosphatidyl head groups only (phospho IPL; Weijers et al., 2011b). Analysis of the non-hydrolyzed IPL fraction was performed to determine any carryover of CL into the IPL fraction.

#### GDGT Analysis

GDGTs from all treatments were analyzed using high performance liquid chromatography-atmospheric pressure chemical ionization-tandem mass spectrometry (HPLC-APCI-MS/MS), which was performed as described by Zhang et al. (2012), using an Agilent 1200 LC equipped with an automatic injector and coupled to a QQQ 6460 MS; peaks were evaluated using Mass Hunter LC-MS manager software. Separation was achieved using a Prevail Cyano column (2.1 × 150 mm, 3µm;


October 2017 | Volume 8 | Article 2077

*bFor the SPM samples collected from the water column, the depth is referred to the sampling water depth; For the sediments, the depth indicates the river water depth.*

*cCL, core lipids; total-IPL, intact polar lipid (IPL) derived core lipids upon acid (H) hydrolysis; phospho-IPL,*

*dRI2* =

*e-, data are not available or not examined.*

*([GDGT-1]* +

*2*\**[GDGT-2]* +

*3*\**[GDGT-3]* +

*4*\**[GDGT-4]* +

*4*\**[Cren.iso])/100.*

 *IPL-derived core lipids derived upon base (OH) hydrolysis.*

Alltech Deerifled, IL, USA) with n-hexane (solvent A) and a mixture of n-hexane/isopropanol 90/10 (v/v; solvent B). The (M+H)<sup>+</sup> ions of each core isoprenoid GDGT (m/z 1,302, 1,300, 1,298, 1,296, 1,294, 1,292) were monitored via selected ion monitoring (SIM) mode (Schouten et al., 2007).

#### GDGT-Based Indices

Indices based on the fractional abundance of GDGTs were calculated as follows:

$$\text{TEX\_{86}} = \text{([}\_{GDGT} - 2] + [\text{GDGT} - 3] + [\text{Cren.iso}])$$

$$\text{/([}\_{GDGT} - 1] + [\text{GDGT} - 2] + [\text{GDGT} - 3]$$

$$+ [\text{Cren.iso}]) \text{ (Schouten et al., 2002)}\qquad \text{(1)}$$

$$\text{(TM)} \qquad \text{(LODCT)} \quad \text{1} + \text{2} \qquad \text{(LODCT)} \quad \text{21} + \text{2}$$

$$\begin{aligned} \text{Ring Index}\_1(\text{RI}\_1) &= ([\text{GDGT}-1] + 2 \times [\text{GDGT}-2] + 3) \\ &\times [\text{GDGT}-3] + 4 \times [\text{Cren.}] + 4 \end{aligned}$$

× [Cren.iso])/100 (Zhang et al., 2016) (2)

$$\begin{aligned} \text{Ring Index} \text{(RI}\_2\text{)} &= \{ [GDGT - 1] + 2 \times [GDGT - 2] + 3 \\ &\quad \times [GDGT - 3] + 4 \times [GDGT - 4] + 4 \\ &\quad \times [Cren.iso] \} / 100 \end{aligned}$$

with the GDGT numbers corresponding to the GDGT structures in **Figure S1**. Note that RI<sup>2</sup> was originally developed for the current study and modified from RI<sup>1</sup> (Zhang et al., 2016), in which the fractional abundance of crenarchaeol was replaced by GDGT-4 in order to eliminate the influence of crenarchaeol on weighted average number of cyclopentane rings in GDGTs.

#### DNA Extraction and the Quantitative Polymerase Chain Reaction (qPCR)

The SPM samples (n = 12) from station R1 (river water), station M (mixing water), and station S (seawater) were selected for the DNA analysis. The frozen filters were washed 3 times by phosphate buffered saline (pH 7.4). The supernatants were centrifuged under 11,000 g for 10 min. The DNA was extracted following the protocol of FastDNA SPIN Kit. The DNA samples were dissolved with a final dilution in 100-µL deionized water and preserved at −80◦C until further processing. The DNA concentrations were quantified in duplicate with a Nano-Drop spectrophotometer (Thermo Fisher Scientific Inc., Wilmington, DE, USA). The quantitative PCR primers were Arch\_334F (5′ACGGGGCGCAGCAGGCGCGA3′ )/Arch\_ 518R (5′ATTACCGCGGCTGCTGG3′ ) for total archaeal 16S gene quantification (Bano et al., 2004), GII-554 f (5′GTCGMTTTTATTGGGCCTAA3′ ), and Eury806-r (5′CACAGCGTTTACACCTAG3′ ) for MG-II Euryarchaeota 16S gene quantification (Massana et al., 1997; Teira et al., 2004). The qPCR analysis was performed at 95◦C for 30 s and 40 cycles at 94◦C for 30 s, 55◦C for total archaea and 53◦C for MG-II Euryarchaeota for 30 s and 68◦C for 1 min. Triplicate measurements were run for each sample and standard.

PCR bands of 16S rRNA gene and MG-II gene were amplified from SPM samples in Station M. They were recovered by a Gel Extraction Kit (omega) and sequenced on the 3730 sequencing platform. The sequences were annotated as the corresponding target genes, which demonstrated the specificity of the chosen qPCR primers. A dilution series of purified DNA from those positive clones were used as standards. A melting curve analysis was performed to demonstrate that the fluorescent signal obtained in a given reaction was consistent with the expected profile for specific PCR products. The R 2 values of standard curves were >0.99. The efficiency of each qPCR was between 87 and 99%.

#### Amplicon Sequencing of the Archaeal 16S rRNA Gene

SPM samples collected in river water, mixing water and seawater were selected to conduct MiSeq pyrosequencing targeting the archaeal 16S rRNA gene. In contrast to the qPCR primers, these primers targeted longer sequences to increase the precision of phylogenetic analysis. The primers were Arch\_344F (5′ACGGGGCGCAGCAGGCGCGA3′ ) and Arch\_915R (5′GTGCTCCCCCGCCAATTCCT3′ ; Gantner et al., 2011). The MiSeq sequencing was conducted on the MiSeq platform (2 × 250 PE, Illumina) at the Shanghai Personalbio Biotechnology (Shanghai, China). Mothur (version 1.29.2; Schloss et al., 2009) was applied to filter the raw pyrosequencing data. The selected sequences were analyzed using the QIIME standard pipeline (Caporaso et al., 2010). Taxonomy was assigned according to the Ribosomal Database Project (RDP) classifier 2.2 (minimum confidence of 80%; Cole et al., 2009). The GenBank accession numbers are PRJNA38421 for those archaeal 16S rRNA genes.

## Satellite-Derived Surface Water Temperature (SWT)

The satellite-derived SWT was determined with a spatial resolution of 4 km from the NOAA advanced very-highresolution radiometer (AVHRR; version 5.2; http://www.nodc. noaa.gov/SatelliteData/pathfinder4km/). The June mean SWT was obtained from the daily averaged values of 30 days in June 2011 (sampling month). The annual mean SWT and winter mean SWT represented 8-year mean values of annual mean temperature (2004–2011) and monthly mean temperature (December–February), respectively, as the surface sediment (top ca. 10 cm) collected in this study might represent a deposition of 6–10 years, based on an estimation from Strong et al. (2012).

## RESULTS AND DISCUSSION

## TEX86-Derived Temperature and Ring Index

The TEX86-derived temperature was calculated based on the calibration of Kim et al. [SST = 68.4 × log (TEX86) + 38.6; (Kim et al., 2010)]. The CL-TEX<sup>86</sup> temperatures derived from either the SPM or the surface sediments were close to the satellitebased annual mean SWT in the river water and mixing water, station R and M, respectively; whereas CL-TEX<sup>86</sup> temperatures for the seawater station S were lower than the winter mean SWT (**Figure 2**). The correspondence between the CL-TEX<sup>86</sup> temperatures derived from SPM and sediment samples (R <sup>2</sup> = 0.70, P < 0.01) along the salinity gradient indicated that the TEX<sup>86</sup> signal in the sediment predominantly reflected archaea from the water column, which is consistent with previous studies

in the PR estuary (Wang et al., 2015) and other coastal settings (Herfort et al., 2006; Zell et al., 2014).

Total IPL-TEX<sup>86</sup> temperatures in the water column were consistently lower than both the June mean SWT and in situ measurements, although the sampling season was summer (**Figure 2**). In the mixing water and seawater, the phospho IPL-TEX<sup>86</sup> was lower than the total IPL-TEX<sup>86</sup> whereas it was very close to the CL-TEX<sup>86</sup> (**Figure 2**; **Table 1**). Since phosphatidyl head groups can be degraded faster than the glycosidic head groups, the phospho IPL is considered to be a better reflection of the living microorganisms (Harvey et al., 1986; Schouten et al., 2010), which may explain the deviation between these IPL pools. Relatively rapid conversion of phospho IPL to CL may result in more similar ring distributions and thus TEX<sup>86</sup> values between phosphor IPL and CL. Furthermore, in the same study area, the variability in TEX<sup>86</sup> was suggested to be due to changes in archaeal community composition in the water column, in which the unusually low TEX86-derived temperature in the coastal SCS was speculated to be linked to MG-II Euryarchaeota (Wang et al., 2015).

Since TEX<sup>86</sup> can be influenced by factors other than temperature, the ring index was proposed to evaluate the accuracy of TEX<sup>86</sup> in marine sediments (Zhang et al., 2016). Here, CL-, phosphor IPL-, and total IPL-TEX<sup>86</sup> values were plotted against RI<sup>1</sup> (Equation 2; **Figure 3**) using data derived from the SPM and surface sediment samples collected during the current and previous studies (Wei et al., 2011; Ge et al., 2013; Zhang et al., 2013; Wang et al., 2015). Most SPM and surface sediment samples from the open South China Sea plotted within the RI1-TEX86-confined zone [RI<sup>1</sup> = 3.32 × (TEX86) <sup>2</sup> − 0.77 × TEX<sup>86</sup> + 1.59, ±2σ ∼ 0.3; (Zhang et al., 2016)], whereas the majority of samples from coastal SCS and the PR estuary fell above the calibration zone (**Figure 3**), exhibiting higher ring index values than those from the open SCS. This implies that the GDGT pool in the water column of the PR estuary and coastal SCS comprised relatively more cyclopentane rings than predicted from the measured SWT. If it is assumed that GDGTs produced only by Thaumarchaeota underpin the relationship between SWT and TEX<sup>86</sup> (e.g., Schouten et al., 2002), then it follows that either Thaumarchaeota in the PR estuary respond differently to temperature than marine strains, or there is another source of cycloalkyl-containing GDGTs in the estuary.

To further assess the distribution of RI values in the SPM and to explore other contributor(s) to the cyclopentyl GDGT pool in the study area, mean values of CL-, total IPL-, and phospho IPLring index were examined (**Figure 4**). Note that crenarchaeol, a biomarker for MG-I Thaumarchaeota, was excluded from the ring index calculation (RI2, Equation 3) in order to limit its overwhelming influence on the index. The re-defined RI<sup>2</sup> equation is more sensitive to the variation of cyclopentanecontaining GDGTs that might be contributed by other archaea. Compared with the river water and seawater, the highest RI<sup>2</sup> value for either CL (avg. 0.39 ± 0.08) or IPL (avg. 0.48 ± 0.07 for total IPL; avg. 0.47 ± 0.10 for phospho IPL) occurred at station M in the mixing water (**Figure 4**; **Table 1**), suggesting that the PR estuary (station M) appeared to be a hot spot of production of GDGTs with cyclopentane moieties. Further confirmation came from the comparison of %GDGT 1–4 in different water

of MG-II *Euryarchaeota* to total archaea along the salinity gradient from the river water to seawater. RI (Equation 3) was calculated from CL (red bars), total-IPL (yellow bars), and phosphor-IPL (blue bars). Details are shown in Table 1.

settings (Table S1), which showed that the sum of the fractional abundances of the GDGTs with 1–4 cyclopentane moieties in the mixing water was significantly higher than that in the seawater or river water. In the mixing water station, the mean values of total IPL-RI and phospho IPL-RI were not significantly different; both were higher than the CL-RI (**Figure 4**). In the seawater station, however, the total IPL-RI (0.34 ± 0.07) was more elevated than the phospho IPL-RI (0.22 ± 0.01) and the CL-RI (0.24 ± 0.04; **Figure 4**). This ring index distribution pattern at station S corresponded to the TEX86-temperature distribution in the SPM and sediment samples (**Figure 2**), suggesting that cyclopentanecontaining GDGTs altered the TEX<sup>86</sup> record in the water column, and the vertical transportation of GDGTs from the water column appeared to be a predominant source to the sediment.

#### Relationship between MG-II Euryarchaeota and Cyclopentane-Containing GDGTs

Previous studies reported that a significant proportion of MG-II Euryarchaeota was diversely present in the estuarine and coastal regions, including the Pearl River estuary (Liu et al., 2014; Wang et al., 2015), the Yangtze River estuary (Liu M. et al., 2011), and the Jiulong River estuary (Hu et al., 2015). In this study, the MG-II Euryarchaeota 16S rRNA gene averaged 5.4 ± 5.9 × 10<sup>8</sup> copies L −1 (n = 6) at the mixing water station, which was two to three orders of magnitude higher than that in river water station (avg. 1.5 ± 2.1 × 10<sup>5</sup> copies L−<sup>1</sup> , n = 2) and seawater station (avg. 4.9 ± 9.4 × 10<sup>6</sup> copies L−<sup>1</sup> , n = 4; **Figure 4**). Considering the heterotrophic life style of MG-II, which have been demonstrated by the former genetic analysis (Iverson et al., 2012; Li et al., 2015; Martin-Cuadrado et al., 2015) and cultivation experiment (Orsi et al., 2015, 2016), the high abundances of MG-II in the mixing water station seem to be due to the high phototrophs that enhanced by nutrients input from upper river (Gan et al., 2014).

The ratio of MG-II Euryarchaeota 16S rRNA gene abundance to total archaeal 16S rRNA gene abundance ([MG-II 16S]/[Archaea 16S]) in mixing water (avg. 0.25 ± 0.08, n = 6) was significantly higher than in the seawater (avg. 0.10 ± 0.004, n = 4; P < 0.01), whereas it was negligible (avg. <0.0001) in the river water (**Figure 4**). This observation was also supported by pyrosequencing analysis, which exhibited a linear correlation with the qPCR-based ratio of [MG-II]/[Archaea 16S] (**Figure S2**; Table S2). These results further confirmed that the PR estuary (mixing zone, salinity avg. 16.6) provided a habitat to sustain a natural enrichment of planktonic MG-II Euryarchaeota.

The presence of (more labile) phospho IPL-GDGTs implied in situ production of isoprenoidal GDGTs in the water column along the entire salinity gradient from the Pearl River to coastal SCS. The elevated phospho IPL-RI in the mixing water

FIGURE 5 | RI2 (A), fractional abundance of GDGTs (B–E, G) and Crenarchaeol and isomer (H,F, respectively), and the ratio of GDGT-2 to GDGT-3 (I) vs. the ratio of the MG-II *Euryarchaeota* 16S rRNA genes to the total archaeal 16S rRNA genes. The black points represent SPM samples collected from the lower Pearl River, the PR estuary, and the coastal SCS.

Wang et al. MG-II Production of Cyclopentyl GDGTs

additionally suggests that higher relative proportions of GDGTs with 1–4 cyclopentane moieties were produced in the PR estuary. A linear regression analysis confirmed the positive relationship between phospho IPL-derived RI and the [MG-II 16S]/[Archaea 16S] ratio in the water column along the salinity gradient (R 2 = 0.72, P < 0.01; **Figure 5A**). Therefore, it is reasonable to hypothesize that MG-II Euryarchaeota preferentially synthesized GDGTs with 1–4 cyclopentane moieties in this region. Moreover, production of GDGTs-1 to -4 by MG-II Euryarchaeota could represent the missing source needed to explain the elevated value of RI and deviation from the RI-TEX<sup>86</sup> relationship driven by the preservation of Thaumarchaeota GDGTs in global core-top sediments.

To further constrain the relationship between MG-II Euryarchaeota and cyclopentane-containing GDGTs, linear regression analysis was conducted between the fractional abundance of GDGTs and the [MG-II 16S]/[Archaea 16S] ratio in the SPM along the salinity gradient. Results exhibited a significantly positive linear correlation between the [MG-II 16S]/[Archaea 16S] ratio and the fractional abundance of phospho IPL-based GDGT with cyclopentane moieties (**Figures 5B–F**). In contrast, we observed no correlation between the [MG-II 16S]/[Archaea 16S] ratio and the fractional abundances of phospho IPL-based GDGT-0 or crenarchaeol (**Figures 5G,H**; **Table 2**). Similar trends of the linear correlations were also observed between the [MG-II 16S]/[Archaea 16S] ratio and the CL- and total IPL-GDGTs with 1–4 cyclopentane moieties, with a less significant correlation between the [MG-II 16S]/[Archaea 16S] ratio and the total IPL-based crenarchaeol regioisomer (**Table 2**). Although MG-II Euryarchaeota were suggested to be an alternative source of crenarchaeol in the ocean (Lincoln et al., 2014a), our study showed an absence of a significant correlation between the distribution of MG-II Euryarchaeota and crenarchaeol (**Figure 5H**). However, members of MG-II Euryarchaeota living in the North Pacific Subtropical Gyre (i.e., those targeted by Lincoln et al., 2014a) may be different from those living in the coastal zone of the PR estuary. This is consistent with the phylogenetic distribution of MG-II reported by Wang et al. (2015), which showed diverse groups of MG-II living in this region.

By comparison of the R 2 values and slopes of the regression equations (**Figures 5B–F**), GDGT-1 exhibited not only the strongest correlation with MG-II Euryarchaeota in the study area, but also largest relative enrichment (i.e., slope of 14.06). If MG-II Euryarchaeota preferentially synthesized GDGT-1, additional contribution of GDGT-1 to the water column of the PR estuary and the coastal SCS would be reflected by an increase in RI values and a substantial decrease in TEX<sup>86</sup> values. Offsets in TEX<sup>86</sup> values have been similarly proposed to result from the decreased ratio of GDGT-2 to GDGT-3 in the surface sediments of this area (Wang et al., 2015), whereas the increased ratio of GDGT-2 over GDGT-3 in the deep-water column seems to be responsible for a warm bias of TEX86 derived temperature in other marine environments (Taylor et al., 2013; Hernandez-Sanchez et al., 2014). In this study, however, no correlation is exhibited between GDGT-2/3 ratio and [MG-II 16S]/[Archaea 16S] ratio (**Figure 5I**). A recent study by Kim et al. TABLE 2 | Regression analysis between the ratio of MG-II 16S rRNA genes to Archaeal 16S rRNA genes vs. the fractional abundance of GDGTs, ring index (RI2), and the ratio of GDGT-2 to GDGT-3.


*a [MG-II/Archaea], the 16S rRNA gene ratio of MG-II to archaeal. CL, core lipids; total-IPL, intact polar lipids derived upon acid hydrolysis; phospho-IPL, intact polar lipids derived upon base hydrolysis.*

(2015) suggested that coincident increases in GDGT-2 and the crenarchaeol regioisomer and decreases in GDGT-1 and GDGT-3 shifted TEX86-derived temperatures toward higher values in the deep-water surface sediments of the Mediterranean Sea. In combination with our data (**Figures 4**, **5**), these observations are consistent with the interpretation that planktonic Euryarchaeota have the potential to bias TEX<sup>86</sup> by changing the distribution of TEX86-related GDGTs (especially those with 1–3 cyclopentane rings) in different marine environments.

## CONCLUSIONS

This study assessed the relationship between TEX86-related GDGTs and MG-II Euryarchaeota along a salinity gradient from river water to seawater. The fractional abundance of MG-II Euryarchaeota was correlated with %GDGTs with cyclopentane moieties as well as ring index values, implying that MG-II Euryarchaeota may contribute ringed GDGTs to the total GDGT pool. This source would thus increase the ring index value and potentially bias the TEX<sup>86</sup> proxy. However, MG-II Euryarchaeota living in the estuary and coastal region did not seem to be a significant source of crenarchaeol. These and other findings based on environmental distributions provide indirect evidence of the lipid profile of MG-II Euryarchaeota, which cannot be validated until a pure culture is available.

## AUTHOR CONTRIBUTIONS

JW and CZ designed this study. JW extracted and analyzed lipids. WX extracted and analyzed DNA. JW, WX, YZ, TM, and CZ wrote the paper.

## ACKNOWLEDGMENTS

We thank Huangmin Ge, Geng Wu, and Chao Li for helping with the sampling. Songze Chen helped performing the qPCR analysis. We appreciate Dr. Ding He (Zhejiang Uni.) for commenting on an earlier version of the manuscript. An early version of this manuscript was discussed online at the Biogeosciences Discussion forum (doi: 10.5194/bgd-12-12455-2015); however, the final paper was declined for publication by Biogeosciences. This research was supported by the National Key Basic Research Program of China grant #2013CB955703 and 2016YFA0601101 (CZ), and the National Science Foundation of China Grant # 41530105 and 41673073 (CZ). This study is also a contribution to the international IMBER project. Comments from the two reviewers and the editor significantly improved the quality of the paper and are greatly appreciated.

#### REFERENCES


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2017.02077/full#supplementary-material

Figure S1 | Structures of archaeal core GDGTs described in the text.

Figure S2 | Relationship of fractional abundances of MG-II *Euryarchaeota* derived from quantitative polymerase chain reaction (qPCR) and pyrosequencing. % MG-II (qPCR) is calculated based on the ratio of MG-II 16S rRNA gene to Archaea 16S rRNA gene; % MG-II (sequencing) refers to the fractional abundance of the MG-II OTUs to the total archaeal OTUs.


II from the mediterranean deep chlorophyll maximum. ISME J. 9, 1619–1634. doi: 10.1038/ismej.2014.249


impalications for SST reconstructions. Global Planet. Change 108, 158–174. doi: 10.1016/j.gloplacha.2013.06.011


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

The reviewer FS declared a shared affiliation, with no collaboration, with several of the authors to the handling Editor.

Copyright © 2017 Wang, Xie, Zhang, Meador and Zhang. 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.

# Diversity and Structure of Diazotrophic Communities in Mangrove Rhizosphere, Revealed by High-Throughput Sequencing

Yanying Zhang1,2,3, Qingsong Yang1,4, Juan Ling<sup>1</sup> , Joy D. Van Nostrand<sup>3</sup> , Zhou Shi<sup>3</sup> , Jizhong Zhou<sup>3</sup> and Junde Dong1,2 \*

<sup>1</sup> CAS Key Laboratory of Tropical Marine Bio-resources and Ecology, Guangdong Provincial Key Laboratory of Applied Marine Biology, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, China, <sup>2</sup> Tropical Marine Biological Research Station in Hainan, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Sanya, China, <sup>3</sup> Department of Microbiology and Plant Biology, Institute for Environmental Genomics, University of Oklahoma, Norman, OK, United States, <sup>4</sup> University of Chinese Academy of Sciences, Beijing, China

#### Edited by:

Hongyue Dang, Xiamen University, China

#### Reviewed by:

Wei Xie, Tongji University, China Radha Prasanna, Indian Agricultural Research Institute (ICAR), India

> \*Correspondence: Junde Dong dongjd@scsio.ac.cn

#### Specialty section:

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

Received: 31 March 2017 Accepted: 04 October 2017 Published: 18 October 2017

#### Citation:

Zhang Y, Yang Q, Ling J, Van Nostrand JD, Shi Z, Zhou J and Dong J (2017) Diversity and Structure of Diazotrophic Communities in Mangrove Rhizosphere, Revealed by High-Throughput Sequencing. Front. Microbiol. 8:2032. doi: 10.3389/fmicb.2017.02032 Diazotrophic communities make an essential contribution to the productivity through providing new nitrogen. However, knowledge of the roles that both mangrove tree species and geochemical parameters play in shaping mangove rhizosphere diazotrophic communities is still elusive. Here, a comprehensive examination of the diversity and structure of microbial communities in the rhizospheres of three mangrove species, Rhizophora apiculata, Avicennia marina, and Ceriops tagal, was undertaken using high-throughput sequencing of the 16S rRNA and nifH genes. Our results revealed a great diversity of both the total microbial composition and the diazotrophic composition specifically in the mangrove rhizosphere. Deltaproteobacteria and Gammaproteobacteria were both ubiquitous and dominant, comprising an average of 45.87 and 86.66% of total microbial and diazotrophic communities, respectively. Sulfate-reducing bacteria belonging to the Desulfobacteraceae and Desulfovibrionaceae were the dominant diazotrophs. Community statistical analyses suggested that both mangrove tree species and additional environmental variables played important roles in shaping total microbial and potential diazotroph communities in mangrove rhizospheres. In contrast to the total microbial community investigated by analysis of 16S rRNA gene sequences, most of the dominant diazotrophic groups identified by nifH gene sequences were significantly different among mangrove species. The dominant diazotrophs of the family Desulfobacteraceae were positively correlated with total phosphorus, but negatively correlated with the nitrogen to phosphorus ratio. The Pseudomonadaceae were positively correlated with the concentration of available potassium, suggesting that diazotrophs potentially play an important role in biogeochemical cycles, such as those of nitrogen, phosphorus, sulfur, and potassium, in the mangrove ecosystem.

Keywords: mangroves, microbial community, diazotrophs, nifH, high-throughput sequencing, sulfate-reducing bacteria

## INTRODUCTION

fmicb-08-02032 October 16, 2017 Time: 12:42 # 2

Mangroves are unique intertidal ecosystems along tropical and subtropical coastlines and play an essential role in maintaining sea levels and protecting coasts in tropical and subtropical regions (Duke et al., 2007). In tropical marine environments, mangroves are thought to be important as primary producers of organic matter, providing the basis of a large and complex food web (Holguin et al., 2001). Although mangrove ecosystems are rich in organic matter, in general they are nutrient-deficient (Sengupta and Chaudhuri, 1991; Holguin et al., 1992; Alongi et al., 1993; Vazquez et al., 2000). Microorganisms are an important component of the mangrove ecosystem, and there is increasing evidence that microbes are crucial to the biogeochemical productivity of the mangrove ecosystem (Holguin et al., 2001; Thatoi et al., 2013).

In tropical mangroves, bacteria and fungi constitute 91% of the total microbial biomass, whereas algae and protozoa represent only 7 and 2%, respectively (Alongi, 1988). Bacterial communities play an important role in nutrient transformation in mangrove ecosystems (Holguin et al., 2001). Nitrogen fixation [converting gaseous nitrogen (N2) to biologically available forms such as ammonia (NH3)] by diazotrophs is considered to be the major source of combined nitrogen input in mangrove forest habitats (Kyaruzi et al., 2003). Free-living diazotrophs are widely distributed within the mangrove ecosystem, with high rates of nitrogen fixation detected in association with dead and decomposing leaves, pneumatophores, rhizosphere soil, tree bark, cyanobacterial mats covering the surface of the sediment, and the sediments themselves (Zuberer and Silver, 1978, 1979; Hicks and Silvester, 1985; Holguin et al., 1992, 2001; Lugomela and Bergman, 2002). The high productivity of mangrove ecosystems might be partially attributable to the high rate of biological nitrogen-fixing activity of diazotrophs in sediments and in the rhizosphere of mangrove trees (Holguin et al., 2001).

In recent years, high-throughput sequencing has offered a more comprehensive perspective on microbial communities and has been employed to study the bacterial community associated with mangroves (Dos Santos et al., 2011; Andreote et al., 2012; Gomes et al., 2014; Loganathachetti et al., 2015; Alzubaidy et al., 2016; Wu et al., 2016). The functional diversity and structure of the microbial communities in mangrove wetlands is largely shaped by environmental variables, and each habitat harbors unique microbial functional communities (Bai et al., 2013). Mangrove trees can influence the growth and distribution of microbial communities by enriching the organic carbon pool and changing the redox conditions of the sediments (Holguin et al., 2001). Previous results obtained using acetylene reduction and molecular methods [denaturing gradient gel electrophoresis (DGGE) and terminal restriction fragment length polymorphism (T-RFLP)] showed that the composition and activity of diazotrophs in mangrove ecosystems are strongly influenced by both root–bacterial interactions (Zuberer and Silver, 1978, 1979; Ravikumar et al., 2004; Flores-Mireles et al., 2007) and geochemical parameters (Zhang et al., 2008; Romero et al., 2012). However, the functional nifH gene, a marker for diazotrophs, has seldom been examined using high-throughput sequencing in the mangrove rhizosphere (Jing et al., 2014), and little is known comprehensively about the roles that mangrove tree species and environmental variables play in shaping mangove rhizosphere diazotrophic communities. Therefore, it is necessary to examine the diversity, composition, and structure of sediment communities based on both 16S rRNA and nifH genes and their links with environmental factors in order to improve our understanding of mangrove ecosystem functioning.

In this study, high-throughput Illumina sequencing was used to investigate microbial communities in the rhizosphere sediments of three species of urban mangrove trees located in Sanya River Mangrove Nature Reserve: Rhizophora apiculata, Avicennia marina, and Ceriops tagal. The aims were to: (i) investigate the diversity and abundance of microbial taxa associated with these three mangrove habitats based on identification of the 16S rRNA and nifH genes; (ii) determine the differences in the total microbial composition and the diazotrophic composition among different mangrove species; and (iii) explore the possible relationships between the 16S rRNA and nifH gene communities and environmental variables, given the important roles of these microbes in driving biogeochemical cycles in mangrove ecosystems.

## MATERIALS AND METHODS

## Study Site and Sampling Collection

Three species of mangrove trees were investigated, R. apiculata, A. marina, and C. tagal, which are the dominant species in Sanya River Mangrove Nature Reserve, a typical tropical urban mangrove ecosystem located in the southernmost part of Hainan Island in China. In June 2013, six different sediment cores were collected randomly at each sampling point during low tide. Sediments adjacent to the rhizosphere were collected down to approximately 10 cm. After roots were removed, the sediment was packed on-site into sealed polythene bags. Samples were maintained on ice until transfer to the laboratory. The wet sediments of each core were thoroughly mixed, and subsamples used for nucleic acid extraction were stored at −20◦C prior to DNA extraction. Subsamples for environmental parameter analysis were stored at 4◦C prior to analysis. Environmental variables including total carbon (TC), total nitrogen (TN), total phosphorus (TP), and available potassium (AK) were measured as described by Bao (1999).

#### DNA Extraction, Amplification, Sequencing, and Data Processing

Total community DNA was extracted from 1.0 g of wet sediment using an E.Z.N.A. <sup>R</sup> Soil DNA Kit (Omega Bio-tek, Norcross, GA, United States). The DNA was then purified with a Promega Wizard DNA Clean-Up System (Madison, WI, United States). DNA concentration was measured by Pico Green using a FLUOstar OPTIMA fluorescence plate reader (BMG Labtech, Jena, Germany).

Two genes were amplified for each sample. To characterize the general 16S rRNA gene community (total microbial

composition), the V4 region of the 16S rRNA gene was amplified with the primer pair 515F (5<sup>0</sup> -GTGCCAGCMGCCGCGGTAA-3 0 ) and 806R (5<sup>0</sup> -GGACTACHVGGGTWTCTAAT-3<sup>0</sup> ). The diazotrophic microbial community was characterized using the nifH gene, which was amplified with primers PolF and PolR (Poly et al., 2001). Both sets of amplicons were modified with Illumina adapter and barcode sequences (Caporaso et al., 2012). Sample libraries were generated from purified PCR products. The Miseq 300 Cycle Kit was used for paired-end sequencing on a Miseq benchtop sequencer (Illumina, San Diego, CA, United States).

The 16S rRNA and nifH gene sequences were separated by sample based on their barcodes, permitting up to one mismatch. Quality trimming was done using Btrim (Kong, 2011). Forward and reverse reads were merged into full-length sequences using FLASH (Magoc and Salzberg, 2011 ˇ ). Sequences were removed if they were too short or contained ambiguous bases. For the 16S rRNA gene, operational taxonomic units (OTUs) were classified using UCLUST at the 97% similarity level. Samples were rarefied to 20,000 sequences per sample. OTUs that were only present in a single sample were removed. Taxonomic assignment was conducted using the RDP classifier (release 5.0) (Wang et al., 2013).

For the nifH gene, the sequences were analyzed using the FRAMEBOT program (Wang et al., 2013). Sequences having frameshift errors were removed. Error-free sequences were then translated into conceptual protein sequences. The NifH protein sequences were grouped into OTUs using the DOTUR program with a 0.05 sequence distance cutoff (Schloss and Handelsman, 2005; Dang et al., 2009, 2013; Zhou et al., 2016). Samples were rarefied to 12,000 sequences per sample, and singletons were removed. Taxonomic assignment for nifH OTUs was carried out by searching representative sequences against reference nifH sequences with known taxonomic information. A neighborjoining (NJ) phylogenetic tree was built by the molecular evolutionary genetics analysis (MEGA6) software (Tamura et al., 2013) for all NifH protein sequences together with selected reference sequences from different diazotrophic groups.

#### Statistical Analysis

All analyses were performed using the package vegan in R (R Foundation for Statistical Computing, Vienna, Austria) or our R-based pipeline<sup>1</sup> . Total microbial and diazotrophic species richness and diversity were calculated using the Chao1, Shannon–Wiener (H<sup>0</sup> ), and Simpson evenness (E) indices. Principal coordinates analysis (PCoA) was used to visualize changes in overall microbial and diazotrophic community structure. Three non-parametric tests [multiple-response permutation procedure (MRPP), permutational multivariate analysis of variance (Adonis), and analysis of similarity (ANOSIM)] were performed based on Bray–Curtis distances to test dissimilarity of 16S rRNA and nifH gene communities among mangrove species. Analysis of variance (ANOVA) was performed to identify significant variation in 16S rRNA and nifH gene groups among mangrove species. Redundancy analysis (RDA) and the Mantel test were performed to determine the relationships between 16S rRNA and nifH gene communities and environmental parameters. All sequences obtained from this study were deposited in the NCBI sequence read archive (SRA) under accession number SRP103888.

## RESULTS

## Environmental Characteristics

The environmental characteristics of the rhizosphere sediment samples from three mangrove species are shown in **Table 1**. All of the measured environmental characteristics including TC, TN, TP, and AK were higher in R. apiculata rhizosphere samples than in those from the other two mangrove species. The sediment samples from the A. marina rhizosphere had the lowest concentrations of TC, TN, and TP, and had the highest nitrogen to phosphorus ratio (N/P). The sediment samples from the C. tagal rhizosphere had the lowest N/P (**Table 1**). The concentrations of TN and AK were higher than observed in our previous investigation (Zhang et al., 2008). The concentration of TP was almost the same as that reported in the Sungei Mandai mangrove ecosystem of Singapore (Jing et al., 2014).

#### 16S rRNA Gene Composition, Diversity, and Community Structure

A total of 406,272 high-quality 16S rRNA gene sequences with lengths of 245–260 bp were obtained. Samples were rarefied to 20,000 sequences per sample. All sequences obtained could be assigned to 24,015 OTUs using UClust (grouped based on 97% similarity). These sequences were classified into 27 bacterial and 2 archaeal phyla (**Figure 1A**). Sequences related to bacteria within the phylum Proteobacteria were the most abundant. Within the Proteobacteria, Gammaproteobacteria, and Deltaproteobacteria sequences made up an average of 25 and 21% of the samples, respectively. Bacteroidetes was the second most abundant phylum observed in this study, followed by Chloroflexi, Acidobacteria, and Firmicutes (**Figure 1A**).

The 16S rRNA gene communities were highly diverse. The number of OTUs ranged from 8,329 to 14,359 per sample, with A. marina harboring the fewest OTUs among the three mangrove species. For individual samples the Shannon–Wiener (H<sup>0</sup> ) index ranged from 6.99 to 8.04, and Simpson evenness (E) ranged from 0.02 to 0.14. The average Shannon–Wiener (H<sup>0</sup> ) of 16S rRNA gene sequence diversity from R. apiculata was higher than those

TABLE 1 | Environmental characteristics of rhizosphere sediment samples (n = 6) from the three mangrove species (expressed as mean value and standard error, SE).


<sup>1</sup>http://www.ou.edu/ieg/tools/data-analysis-pipeline.html

from A. marina and C. tagal (**Table 2**). Three non-parametric tests (MRPP, Adonis, and ANOSIM) were performed using the Bray–Curtis dissimilarity index and consistently showed that microbial communities were significantly different among mangrove species (P < 0.02) (**Table 3**). PCoA was also used to compare microbial communities among mangrove species, and the results confirmed that the microbial communities could be divided into three groups corresponding to the mangrove species (**Figure 2A**).

#### Phylogenetic and Taxonomic nifH Gene Composition, Diversity, and Community Structure

After processing, 216,000 high-quality nifH sequences (283–323 bp) were retrieved from the mangrove rhizosphere sediments of R. apiculata, A. marina, and C. tagal. Samples were rarefied to 12,000 sequences per sample. All sequences obtained could be assigned to 1,334 OTUs at the 95% protein sequence similarity level (Dang et al., 2009, 2013; Zhou et al., 2016). The 1,334 unique nifH protein sequences shared 51–100% sequence identity with the top-match sequences obtained from GenBank. Among these, 591 unique protein sequences (comprising up to 82.82% of total sequences) shared quite high sequence identity (>90%) with nifH sequences of known bacteria or archaea, such as Deltaproteobacteria, Acidithiobacillia, Alphaproteobacteria, Gammaproteobacteria, Bacteroidetes, Cyanobacteria, Chlorobi, Deferribacteres, Epsilonproteobacteria, Firmicutes, Spirochaetes, Verrucomicrobia, and Euryarchaeota. Phylogenetic types of nifH genes were defined according to Zehr et al. (2003). The deduced nifH sequences were affiliated with four major groups in the reconstructed nifH phylogenetic tree (**Figure 3** and Supplementary Figure S1). The nifH community was dominated by sequences belonging to Cluster I and Cluster III nifH clades, which accounted for 35.33 and 64.61% of the total captured sequences and 16.56 and 82.23% of the total OTUs, respectively. Only seven OTUs accounting for 0.52% of total sequences were found to belong to Cluster II, and nine OTUs accounting for 0.67% of total sequences were found in Cluster IV (**Figure 3** and Supplementary Figure S1). R. apiculata harbored significantly more nifH sequences belonging to Cluster I and fewer nifH sequences belonging to Cluster III than did A. marina and C. tagal (Supplementary Tables S1, S2).

Family or higher taxonomic information was then assigned to the 1,334 nifH OTUs, according to their nearest taxonomic matches, for further analyses. All nifH sequences were classified into 1 archaeal and 11 bacterial phyla (**Figure 1B**). Sequences related to diazotrophs within the phylum Proteobacteria were the most abundant. Within the Proteobacteria, Deltaproteobacteriarelated sequences were the most abundant group, making up an average of 60.59% of diazotrophic sequences. Gammaproteobacteria were the second major diazotrophic group, making up an average of 26.07% diazotrophic sequences (**Figure 1B**). The number of diazotrophic OTUs ranged from 490 to 715 across all the samples, with C. tagal harboring the highest number of OTUs among the three mangrove species. The Shannon–Wiener (H<sup>0</sup> ) index ranged from 4.51 to 5.21, and Simpson evenness (E) ranged from 0.05 to 0.12 for individual samples. The average Shannon–Wiener (H<sup>0</sup> ) index of diazotrophic diversity from A. marina was relatively lower than those of A. marina and C. tagal (**Table 2**). Dissimilarity tests showed that the diazotrophic communities were significantly different among mangrove species (P < 0.01) (**Table 3**). PCoA also showed significant variations in the diazotrophic communities from different mangrove species (**Figure 2B**).

## Comparison of 16S rRNA and nifH Gene Composition among Samples from the Three Mangrove Species

The effect of mangrove tree species on 16S rRNA and nifH gene distribution was further investigated using the most dominant 16S rRNA (n = 17) and nifH (n = 15) gene groups (**Figure 4**). The results showed that most of the dominant microbial groups had no significant differences in relative abundance in the communities from the three

TABLE 2 | Diversity indices of 16S rRNA gene and nifH sequences from rhizosphere sediments (n = 6) of three mangrove species (expressed as mean value and standard error, SE).


TABLE 3 | Non-parametric analyses to test dissimilarity of 16S rRNA and nifH gene communities between any two mangrove rhizosphere sediments (RA, R. apiculata; AM, A. marina; CT, C. tagal).


Significant differences (P < 0.05) are indicated in italics.

of variation explained by each axis is shown. RA, R. apiculata. AM, A. marina. CT, C. tagal.

mangrove species. Of these dominant 16S rRNA gene groups, Epsilonproteobacteria, Deferribacteres, and Euryarchaeota were highly abundant in rhizosphere sediment from R. apiculata, while Actinobacteria dominated in sediment from A. marina. However, Gammaproteobacteria and Deltaproteobacteria, the two most dominant bacterial groups across all of the mangrove species, were almost invariable (**Figure 4A**).

In contrast to the total microbial pattern at the phylum level, most of the dominant diazotrophic groups were significantly different among the three mangrove species (P < 0.05). The abundance of the two most dominant diazotrophic groups, Desulfobacteraceae and Desulfovibrionaceae, belonging to the Deltaproteobacteria, was significantly different among the mangrove species (P < 0.05), and these groups accounted for an average of 33.02 and 18.29% of total diazotrophic sequences, respectively. Desulfobacteraceae were highly abundant in rhizosphere sediment from C. tagal, while Desulfovibrionaceae dominated in A. marina. The Pseudomonadaceae of the Gammaproteobacteria, the third most dominant diazotrophic group, were almost invariable, accounting for an average

of 12.38% of total diazotrophic sequences. Two dominant Betaproteobacteria diazotrophic groups, Rhodocyclaceae and Comamonadaceae, were relatively highly abundant in rhizosphere sediment from R. apiculata. The diazotrophic groups belonging to the Verrucomicrobia and Firmicutes were almost invariable (**Figure 4B**).

TABLE 4 | Monte Carlo permutation test of relationship between environmental attributes and 16S rRNA and nifH gene high-throughput sequencing data.

significant differences among sediments of the three mangrove species (P < 0.05).


Significant differences (P < 0.05) are indicated in italics.

## Relationship between 16S rRNA and nifH Gene Communities and Environmental Factors

Redundancy analysis was performed to explore the relationship between 16S rRNA and nifH gene community structures with different environmental characteristics. The results showed that the composition of diazotrophic communities was significantly correlated with all investigated environmental factors (Monte Carlo test, P < 0.05; **Table 4**) except the concentration of TN. The total 16S rRNA gene community did not significantly correlate with TN or N/P ratios of mangrove sediments (**Table 4**). However, the diversity of both 16S rRNA and nifH genes did not significantly correlate with investigated environmental factors (**Figure 5**).

The RDA biplot showed that the abundance of sequences from the 16S rRNA groups Chloroflexi, Betaproteobacteria, Alphaproteobacteria, Firmicutes, and Actinobacteria was positively correlated with the N/P ratio. Gammaproteobacteria, Cyanobacteria, Acidobacteria, and Deltaproteobacteria were positively correlated with TP. In addition, Epsilonproteobacteria, Deferribacteres, and Euryarchaeota were positively correlated with the C/N ratio, TC, and AK (**Figure 5A**).

The diazotrophic groups Deltaproteobacteria, Desulfobulbaceae, and Desulfuromonadaceae were positively correlated with TP. The diazotrophic gammaproteobacterial families Pseudomonadaceae, Cellvibrionaceae, and Chromatiaceae were positively correlated with the C/N ratio, TC, TN, and AK. The diazotrophic groups Deltaproteobacteria, Desulfovibrio, and Desulfobacca were negatively correlated with the C/N ratio, TC, TN, and AK. Diazotrophic Deltaproteobacteria and Desulfobacter were positively correlated with the N/P ratio (**Figure 5B**).

## DISCUSSION

Although high-throughput sequencing approaches are now commonly applied to investigate bacterial community structures in mangroves (Dos Santos et al., 2011; Andreote et al., 2012; Gomes et al., 2014; Loganathachetti et al., 2015; Alzubaidy et al., 2016; Wu et al., 2016), they have not been widely used to target nitrogen-fixing functional genes (nifH) to explore diazotrophic communities in urban mangrove rhizospheres (Jing et al., 2014). Compared with previous high-throughput sequencing studies on diazotrophic communities from the mangroves along the coastline of Singapore (Jing et al., 2014), we found a much greater diversity of bacteria and archaea having the potential to fix nitrogen from the three rhizosphere sediments from R. apiculata, A. marina, and C. tagal, which might be attributable to differences in mangrove species, environmental conditions, and primers used for analysis. The primers we used have a very broad coverage

for both bacterial and archaeal nifH genes, and have been widely used in various environmental studies (Bernard et al., 2014; Hoppe et al., 2014; Keshri et al., 2015; Zhang et al., 2015; Newell et al., 2016; Penton et al., 2016; Tu et al., 2016; Rodrigues et al., 2017).

## 16S rRNA and nifH Gene Community Structures from Rhizosphere Sediments of R. apiculata, A. marina, and C. tagal

Previous studies have reported Deltaproteobacteria and Gammaproteobacteria as the dominant bacterial groups in rhizosphere sediments from many mangrove species (Gomes et al., 2014; Loganathachetti et al., 2015; Alzubaidy et al., 2016; Wu et al., 2016). In this present study, Deltaproteobacteria and Gammaproteobacteria were the dominant groups of both total bacterial and diazotrophic communities from all three mangrove species studied. These two groups constituted an average of 45.87 and 86.66% of total 16S rRNA and nifH gene communities, respectively (**Figure 1**), suggesting that these groups play an important functional role in the anaerobic conditions of the mangrove rhizosphere. Furthermore, there was a high relative abundance of diazotrophic groups from the Deltaproteobacteria in this study, such as Desulfobacteraceae, Desulfovibrionaceae, and Desulfuromonadaceae, belonging to the sulfate-reducing bacteria, which are known to play key roles in sedimentary cycling of N, C, and S (Lyimo et al., 2002; Varon-Lopez et al., 2014; Romero et al., 2015). Sulfate-reducing bacteria were also reported to be prevalent in pristine, anthropogenic, and oil-contaminated mangrove sediments (Santos et al., 2011; Jing et al., 2014; Wu et al., 2016), which suggests that sulfate-reducing bacteria may contribute substantially to both nitrogen fixation and sulfate reduction in the mangrove rhizosphere. Furthermore, our findings reinforce the prominence of sulfate-reducing bacteria as the main diazotrophic group in mangrove samples. In addition, diazotrophs from the Gammaproteobacteria were reported to be widespread in tropical and subtropical oceans (Bird et al., 2005; Jing et al., 2014). The family Pseudomonadaceae is a common constituent of mangrove rhizosphere diazotrophs (Zhang et al., 2008; Liu et al., 2012; Jing et al., 2014), and it was detected as the dominant group of the class Gammaproteobacteria in this study.

In addition, Bacteroidetes and Firmicutes, including a high abundance of sulfate reducers and methanogens, were dominant in sediments of A. marina (Alzubaidy et al., 2016). Bacteroidetes are very frequent in tidal mudflats or near-shore sediments, and an increased abundance of Bacteroidetes in the rhizospheres of mangroves has been noted for Rhizophora mangle, Avicennia schaueriana, and Laguncularia racemosa located in Guanabara Bay (Rio de Janeiro, Brazil) (Gomes et al., 2010). Actinobacteria, including mostly soil-borne microbes, were reported to be enriched in mangrove sediments in the Red Sea (Alzubaidy et al., 2016). Chloroflexi was the second most dominant phylum in three mangrove species in Beilun Estuary (Wu et al., 2016). Consistent with previous reports, bacteria from the phyla Bacteroidetes, Chloroflexi, Acidobacteria, and Firmicutes were also widespread in the rhizosphere sediments of R. apiculata, A. marina, and C. tagal.

#### Influence of Mangrove Species and Environmental Factors on 16S rRNA and nifH Gene Composition from Rhizosphere Sediments

Several factors shaping mangrove rhizosphere microbial communities have been proposed (Alzubaidy et al., 2016). Mangrove tree species and geochemical parameters were widely reported to be influential (Flores-Mireles et al., 2007; Zhang et al., 2008; Gomes et al., 2010, 2014; Romero et al., 2012; Jing et al., 2014; Wu et al., 2016). Mangrove roots have been suggested to be able to impose a selective force on the mangrove rhizosphere microbial communities. This phenomenon appeared to be plant

species-specific (Gomes et al., 2014), resulting in the mangrove tree species playing an important role in shaping the rhizosphere microbial communities (Gomes et al., 2010, 2014; Pires et al., 2012; Wu et al., 2016). Mangrove root exudates may not only stimulate microbial respiration and create suboxic and anoxic microenvironments to facilitate nitrogen fixation, but also directly stimulate nitrogen fixation by providing metabolizable organic matter as carbon and energy sources to the diazotrophs (Dang and Lovell, 2016). The significant influence of the mangrove tree species on the rhizosphere microbial community was further confirmed in this study. Three non-parametric tests based on the Bray–Curtis distance matrix showed that both the total 16S rRNA and the nifH gene communities were significantly different among mangrove species (P < 0.02) (**Table 2**). The PCoA results also confirmed that the 16S rRNA and nifH gene communities could be divided into three groups corresponding to the respective mangrove species (**Figure 2**). The differences in 16S rRNA gene community composition from the three investigated mangrove species showed that the dominant 16S rRNA gene groups Epsilonproteobacteria, Actinobacteria, Deferribacteres, and Euryarchaeota were significantly different in abundance in association with different mangrove species (**Figure 4A**). Our results are consistent with the idea that root exudates select for specific groups at both the taxonomic level and the functional level (Alzubaidy et al., 2016). Most of the dominant diazotrophic groups were significantly different among mangrove species (P < 0.05). The possible reason for plant species-specific diazotrophic communities is that mangroves of different species and under different physiological conditions may secrete different types of organic matter, which selects different diazotrophic species to be functional (Dang and Lovell, 2016).

The importance of geochemical parameters in structuring 16S rRNA and nifH gene communities has been previously shown in mangrove ecosystems (Flores-Mireles et al., 2007; Zhang et al., 2008; Romero et al., 2012; Jing et al., 2014). Our results showed that both total 16S rRNA and nifH gene communities significantly correlated with most of the investigated environmental factors (**Table 4**). Mangrove root exudates provide a valuable source of carbon, while microorganisms that colonize the rhizosphere help plants acquire phosphorus and potassium; enhance nitrogen uptake; or even help the plants to cope with infection, toxic compounds, and other sources of stress (Singh et al., 2004; Kristensen et al., 2008). In this study, the fact that the highest 16S rRNA and nifH gene diversity was found in samples from the R. apiculata rhizosphere is not surprising, as these samples have a relatively high TC concentration. It has been reported that long-term fertilization with nitrogen and phosphorus not only affects the community structure and activity of diazotrophs, but also potentially plant–microbe interactions (Romero et al., 2012, 2015). Diazotroph diversity is reduced or completely eliminated by high levels of ammonia caused by anthropogenic activities, while relatively high phosphorus concentrations provide favorable conditions for nitrogen fixation (Jing et al., 2014). The abundance of the dominant diazotroph Desulfatibacillum in this study was positively correlated with TP, but negatively correlated with the N/P ratio. Since plants require potassium for numerous physiological processes such as growth and development, and protein synthesis, there is interest in bacteria such as Pseudomonas, which are capable of solubilizing potassium to an accessible form in the soil (Alzubaidy et al., 2016). In the present study, Pseudomonadaceae abundance was positively correlated with the concentration of AK. Among the family Pseudomonadaceae, most protein sequences shared quite high sequence identity (>90%) with those of the known bacteria Pseudomonas stutzeri, which further hints at the important role that diazotrophic Pseudomonas play in potassium cycling. It is notable that more than half of the variability (>50%) was unexplained by the host species or environmental variables investigated in this study, for both the 16S rRNA and nifH gene communities. Other unknown factors such as redox, pH, oxygen content, organic matter content, and salinity may be important in influencing the diversity, abundance, structure, and spatial distribution of the sediment nifH gene communities (Dang et al., 2009, 2013; Zhou et al., 2016).

## CONCLUSION

Our study comprehensively characterized the diversity and structure of 16S rRNA and nifH gene communities in the rhizosphere of three mangrove species. Both the mangrove species and various environmental variables played important roles in shaping these communities. Most of the dominant diazotrophs were significantly different among mangrove species. However, our study was based on DNA abundance, so activity of nitrogenase cannot be confirmed. Therefore, expressionbased studies such as mRNA-based microarray hybridization and metagenomic studies, together with in situ nitrogen-fixing measurements, are required to elucidate the role that diazotrophs play in the mangrove rhizosphere.

## AUTHOR CONTRIBUTIONS

YZ, JL, and JD conceived the research. YZ, QY, and JL performed the experiments. YZ wrote the manuscript. JVN and JZ edited the manuscript. QY, JL, and ZS contributed sampling or data analysis pipelines. All authors reviewed and approved the manuscript.

## FUNDING

The research was supported by the National Natural Science Foundation of China (41676107, 41276114, 41676163, and 41406191), National Key Research and Development Program of China (2017YFC0506301), Strategic Priority Research Program of the Chinese Academy of Sciences (XDA13020300), Guangdong Province Public Welfare Research and Capacity Building Project (2015A020216016), and Science and Technology Planning Project of Guangdong Province, China (2014B030301064).

## ACKNOWLEDGMENT

fmicb-08-02032 October 16, 2017 Time: 12:42 # 10

We thank all of the members of the Tropical Marine Biological Research Station in Hainan for their assistance in field sample collection.

## REFERENCES


Bao, S. (1999). Analysis of Agricultural Soil. Beijing: China Agriculture Press.


## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2017.02032/full#supplementary-material

functional diversity of enriched petroleum hydrocarbon-degrading consortia. FEMS Microbiol. Ecol. 74, 276–290. doi: 10.1111/j.1574-6941.2010.00962.x



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

Copyright © 2017 Zhang, Yang, Ling, Van Nostrand, Shi, Zhou and Dong. 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.

# Community Composition of Nitrous Oxide Consuming Bacteria in the Oxygen Minimum Zone of the Eastern Tropical South Pacific

Xin Sun\*, Amal Jayakumar and Bess B. Ward

Department of Geosciences, Princeton University, Princeton, NJ, United States

The ozone-depleting and greenhouse gas, nitrous oxide (N2O), is mainly consumed by the microbially mediated anaerobic process, denitrification. N2O consumption is the last step in canonical denitrification, and is also the least O<sup>2</sup> tolerant step. Community composition of total and active N2O consuming bacteria was analyzed based on total (DNA) and transcriptionally active (RNA) nitrous oxide reductase (nosZ) genes using a functional gene microarray. The total and active nosZ communities were dominated by a limited number of nosZ archetypes, affiliated with bacteria from marine, soil and marsh environments. In addition to nosZ genes related to those of known marine denitrifiers, atypical nosZ genes, related to those of soil bacteria that do not possess a complete denitrification pathway, were also detected, especially in surface waters. The community composition of the total nosZ assemblage was significantly different from the active assemblage. The community composition of the total nosZ assemblage was significantly different between coastal and off-shore stations. The low oxygen assemblages from both stations were similar to each other, while the higher oxygen assemblages were more variable. Community composition of the active nosZ assemblage was also significantly different between stations, and varied with N2O concentration but not O2. Notably, nosZ assemblages were not only present but also active in oxygenated seawater: the abundance of total and active nosZ bacteria from oxygenated surface water (indicated by nosZ gene copy number) was similar to or even larger than in anoxic waters, implying the potential for N2O consumption even in the oxygenated surface water.

Keywords: N2O consuming bacteria, nosZ gene, microarray, oxygen minimum zone, Eastern Tropical South Pacific

#### INTRODUCTION

N2O is a major ozone-depleting substance and a greenhouse gas whose radiative forcing per mole is 298 times that of carbon dioxide (IPCC, 2007; Ravishankara et al., 2009). Oxygen minimum zones (OMZs) are the most intense marine sources of N2O and are hot spots of rapid N2O cycling (Martinez-Rey et al., 2015). OMZs are marine regions with a strong O<sup>2</sup> gradient (oxycline) overlying an oxygen deficient zone (ODZ) where O<sup>2</sup> concentration is low enough to induce anaerobic processes. The global expansion and intensification of OMZs, which are

Edited by:

Hongyue Dang, Xiamen University, China

#### Reviewed by:

Lisa Y. Stein, University of Alberta, Canada James T. Hollibaugh, University of Georgia, United States

> \*Correspondence: Xin Sun xins@princeton.edu

#### Specialty section:

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

Received: 02 May 2017 Accepted: 12 June 2017 Published: 28 June 2017

#### Citation:

Sun X, Jayakumar A and Ward BB (2017) Community Composition of Nitrous Oxide Consuming Bacteria in the Oxygen Minimum Zone of the Eastern Tropical South Pacific. Front. Microbiol. 8:1183. doi: 10.3389/fmicb.2017.01183

predicted to result from global warming, further stress the importance of understanding N2O cycling in these regions (Codispoti, 2010). N2O production and consumption are driven by marine bacteria (Naqvi et al., 2000). The dominant microbial process for N2O cycling is denitrification, the sequential reduction of NO<sup>3</sup> <sup>−</sup> to NO<sup>2</sup> <sup>−</sup>, NO, N2O and finally to N<sup>2</sup> (Zumft, 1997; Naqvi et al., 2000). Denitrification could stop at an intermediate step before N<sup>2</sup> if O<sup>2</sup> concentration exceeds the threshold for the latter step or if electron donors are depleted (Ward et al., 2008; Dalsgaard et al., 2012; Babbin et al., 2014). Thus O<sup>2</sup> concentration or electron donor availability could also control the N2O budget. N2O concentrations and net N2O production rates (N2O production minus N2O consumption) were found to peak at the oxic-suboxic interface in OMZs, due to excess production from nitrification and incomplete denitrification (Nicholls et al., 2007; Ji et al., 2015; Trimmer et al., 2016). However, while multiple processes can produce N2O, reduction by N2O consuming bacteria is the only known biological N2O sink.

N2O consumption is the final step of denitrification, and is the least O<sup>2</sup> tolerant step (Bonin et al., 1989; Körner and Zumft, 1989). N2O consumption rates have been measured in ODZs at depths where O<sup>2</sup> concentration ranged from very low to below the detection limit (Wyman et al., 2013; Babbin et al., 2015). N2O consumption by denitrification and genes involved in N2O reduction have also been detected in oxygenated seawater (Farías et al., 2009; Wyman et al., 2013). Characterizing the distribution and environmental regulation of this step is necessary for a complete quantification of the oceanic N2O budget and will improve our ability to predict oceanic N2O emissions under global climate change.

N2O consumption is catalyzed by the enzyme nitrous oxide reductase, encoded by the nosZ gene. A recent study of nosZ genes found a lower diversity of nosZ genes in ODZ waters than in the upper oxycline of the OMZ (Castro-González et al., 2015). The distribution of nosZ genes was related to O<sup>2</sup> concentration, which suggested that the quantity and composition of nosZ genes and the diversity of denitrifying bacteria might influence the microbial potential for N2O consumption.

We aimed to determine the distribution and community composition of total and transcriptionally active (abbreviated as 'active' hereafter) nosZ assemblages based on the presence (DNA) and expression (RNA) of nosZ genes in the OMZ of the Eastern Tropical South Pacific (ETSP), one of the three major OMZs in the world ocean. Three hypotheses were tested in the study: (1) the community compositions of total and active nosZ assemblages differ between coastal and off-shore stations because the quantity and quality of the nutrients at the two stations differ due to different contributions from land and sediment; (2) quantities and composition of total and active nosZ assemblages are related to O<sup>2</sup> concentration because N2O consumption is the least O<sup>2</sup> tolerant step in conventional denitrification; and (3) the distribution of the active nosZ assemblage is more related to N2O concentration than that of total nosZ assemblage because the former indicates live and active organisms.

## MATERIALS AND METHODS

#### Experimental Sites and Sampling

Samples were collected on the R/V Nathaniel B. Palmer during June to July 2013 (cruise NBP 1305) in the OMZ of the ETSP at the off-shore station (BB1; 14.0◦ S, 81.2◦W) and the coastal station (BB2; 20.50◦ S, 70.70◦W) (Supplementary Figure 1). Particulate material was collected in Niskin bottles mounted on the standard conductivity-temperature-depth (CTD) rosette system (Seabird Electronics, Seattle, WA, United States) at four depths at each station (BB1: 60, 130, 300, and 1000 m; BB2: 60, 115, 300, and 1000 m) and concentrated by filtration (up to 4 L) through Sterivex filters (0.22 µm). Filters were flash frozen in liquid nitrogen onboard and stored at −80◦C until DNA and RNA extraction was performed.

Temperature, salinity, sigma theta, bottom depth and pressure at each station were measured on the SBE 911+ CTD system. Fluorescence, representing chlorophyll a, was measured using a single channel fluorometer (Wet labs, Philomath, OR, United States) mounted on the CTD. Oxygen distributions were determined using the STOX sensor (detection limit = 10 nM) mounted on the CTD rosette (Revsbech et al., 2009). Ammonium, nitrite and nitrate concentrations were measured using standard colorimetric protocols (UNESCO, 1994). N2O concentration was determined using mass spectrometry (Ji et al., 2015). N<sup>∗</sup> is the deviation of measured dissolved inorganic nitrogen (DIN = nitrate + nitrite + ammonia) from predicted DIN by Redfield ratio and the worldocean nitrogen to phosphate regression relationship (Deutsch et al., 2001). Environmental data were reported by Ji et al. (2015) and are provided in Supplementary Table 1.

#### DNA and RNA Extractions

Both DNA and RNA were extracted from eight Sterivex filters using the plant tissue protocol of the All Prep DNA/RNA Mini Kit (50) using a QIAcube (Qiagen). Reverse transcription from RNA to cDNA was performed using SuperScript <sup>R</sup> III First-Strand Synthesis System for RT-PCR (InvitrogenTM by Life TechnologiesTM). Excess RNA was removed by RNase at the end of the synthesis.

#### Quantitative PCR Assays

The abundance of total and active nosZ assemblages were estimated by quantitative PCR (qPCR) using SYBR <sup>R</sup> Green based assays using protocols described previously (Jayakumar et al., 2013). Primers nosZ1F and nosZ1R (Henry et al., 2006) were used to amplify a 259-bp conserved fragment of the nosZ gene. Known quantities (∼20–25 ng) of DNA and cDNA samples were assayed along with a minimum of five serial dilutions of plasmids containing nosZ gene, no template controls and no primer controls, all in triplicate on the same plate. To maintain continuity and consistency among qPCR assays, a subset of samples from the first qPCR assay was run with subsequent assays and fresh standard dilutions were prepared for each assay. DNA, cDNA and the concentrated standards were quantified prior to every assay using PicoGreen fluorescence (Molecular Probes, Sun et al. nosZ Genes in ETSP OMZ

Eugene, OR, United States) calibrated with several dilutions of phage lambda standards, to account for DNA loss due to freeze thaw cycles. qPCR assays were run on a Stratagene MX3000P (Agilent Technologies, La Jolla, CA, United States). Automatic analysis settings were used to determine the threshold cycle (Ct) values. The copy numbers (number of copies of the gene sequence detected in the sample) were calculated according to: Copy number = (ng <sup>∗</sup> number/mole)/(bp <sup>∗</sup> ng/g ∗ g/mole of bp) and then converted to copy number per ml seawater filtered, assuming 100% extraction efficiency.

#### Microarray Experiments

fmicb-08-01183 June 23, 2017 Time: 14:46 # 3

DNA and cDNA qPCR products were used as targets for microarray experiments to characterize the community composition of total and active nosZ assemblages, respectively. Triplicate qPCR products from each depth were pooled. nosZ gene targets were purified and extracted from agarose gels using the QIAquick gel extraction kit (Qiagen). Purified DNA qPCR products from eight depths and cDNA qPCR products from seven depths were used to prepare targets for microarray analysis.

Microarray targets were prepared from the qPCR products following the protocol of Ward and Bouskill (2011). Briefly, dUaa was incorporated into purified DNA and cDNA during linear amplification using the BioPrime kit (InvitrogenTM). The dUaa-Klenow product was labeled with Cy3 (dissolved in dimethyl sulfoxide), purified using QIAquick columns (Qiagen) and quantified by Nanodrop 2000 (Thermo Scientific). Duplicate Cy3 products for each sample were hybridized at 65◦C overnight (16 h) onto replicate microarrays under ozone free conditions. Hybridized microarrays were washed and scanned with an Axon 4300 laser scanner.

#### nosZ Microarray

The microarray (BC016) contains 114 nosZ archetype probes. Each probe is a 90-bp sequence comprised of a 70-bp nosZ gene fragment and a 20-bp control region. Each archetype probe represents, and hybridizes with, all nosZ sequences with >85% identity, based on published sequences available in 2013. There are 71 NosZ archetypes, which represent typical or Clade I nosZ genes, and 43 WNZ archetypes, which represent the atypical or Clade II nosZ genes. The development of the microarray is described in Jayakumar et al. (in preparation) and the sequences are shown in Supplementary Table 2.

#### Data Analyses

Fluorescence signal intensities for nosZ probes hybridized to the microarrays were obtained using GenePix Pro 7 software. The fluorescence ratio (FR) of each feature is defined as the ratio Cy3/Cy5 (70-mer probe/20-mer standard for each feature). The FR for each nosZ archetype was calculated as the average of probe signal intensities for duplicate features on the same microarray. Normalized fluorescence ratio (FRn) was calculated by dividing the FR of each nosZ probe by the maximum nosZ FR on the same microarray. FRn is the proxy of the relative abundance of each archetype and was used for further analyses.

Detrended correspondence analysis (DCA) was performed to analyze the overall microbial community composition. A dissimilarity test was performed using Permutational Multivariate Analysis of Variance (adonis). α-diversities (Shannon diversity indices) of total and active nosZ assemblages were calculated. β-diversities (Bray–Curtis dissimilarities) of total and active nosZ assemblages between different sites (i.e., depths) were calculated to perform a Mantel test. The Mantel test was used to determine significant environmental variables correlated with microbial community composition. These analyses were carried out using the vegan package in R (version 3.3.1). A maximum likelihood phylogenetic tree was built from aligned archetype sequences with MEGA 7 software. FRn values for each archetype at different depths from both stations were visualized on the phylogenetic tree by iTOL<sup>1</sup> . The copy number of nosZ genes at each depth is given as mean (± standard error) of the qPCR triplicates.

#### RESULTS

#### Abundance and Depth Distribution of Total and Active nosZ Assemblages

At stations BB1 and BB2, the continuously undetectable O<sup>2</sup> concentration, the local nitrite maximum and the nitrate deficit at intermediate depths (130–370 m at BB1; 75–400 m at BB2) all indicated the presence of ODZs (gray areas in **Figure 1**). Sampling depths were chosen to represent water column features defined by oxygen concentration, as measured with the in situ STOX sensor: oxygenated surface water, upper oxycline [characterized by sharp O<sup>2</sup> concentration gradient ranging from saturation to below detection limit (<10 nM)], top of the ODZ (O<sup>2</sup> concentration <10 nM), core of the ODZ (O<sup>2</sup> concentration <10 nM) and lower oxycline (O<sup>2</sup> concentration >10 nM). The abundance of the total nosZ genes ranged from 24.1 (±1.4) copies mL−<sup>1</sup> in a sample from the lower oxycline to 636.4 (±28.3) copies mL−<sup>1</sup> in a sample from the ODZ (**Figure 1**). As for the active nosZ assemblage, the lowest abundance of active nosZ genes was 5.1 (±0.5) copies mL−<sup>1</sup> in a sample from the lower oxycline and the highest abundance was 604.6 (±103.7) copies mL−<sup>1</sup> in a sample from the surface water.

The abundance of total and active nosZ assemblages showed different distribution patterns at the two stations (**Figure 1**). At station BB1, the abundance of both total and active nosZ genes was highest in a sample from the surface water, and decreased with depth. At station BB2, the abundance of both total and active nosZ genes peaked in samples from the ODZ and was lowest in samples from the lower oxycline. The active nosZ genes were most abundant in the sample from 300 m. However, the total nosZ genes were most abundant in the sample from 115 m, where the abundance of the active nosZ genes was only 3% of the total.

<sup>1</sup>http://itol.embl.de/

#### Diversity and Dominant Archetypes of Total and Active nosZ Assemblages

The distribution of FRn of the total or active archetypes was similar across all depths within the same station (**Figure 2**). The average α-diversity was not significantly different (student's t-test, P = 0.102) between the total assemblages (3.21) and the active assemblages (2.60) (**Table 1**). The least diverse total assemblage was from the lower oxycline (1000 m of station BB2), but the two least diverse active assemblages were from the ODZs (130 m of BB1 and 300 m of BB2).

The FRn distribution of nosZ archetypes showed that a very limited number of archetypes dominated the total or the active nosZ assemblages (**Figure 2**). Dominant archetypes were affiliated with bacteria from various environments, including salt marsh, soil, marine sediment, marine hot spring and activated sludge of a wastewater treatment plant (Supplementary Tables 3, 4). The FRn of the top five most abundant archetypes accounted for 48.9 to 83.3% of the total nosZ hybridization signal (**Figure 3A** and Supplementary Table 3). Notably, the highest percentage (83.3%) was from the sample from the lower oxycline (1000 m) at station BB2 and the most abundant archetype (NosZ42, an uncultured clone of nosZ gene derived from salt marsh sediments; Kearns et al., 2015) accounted for 31.6% of total FRn. However, this archetype was not among the top five archetypes of the active nosZ assemblage in the same sample (**Figure 3B** and Supplementary Table 4). The two most dominant typical nosZ archetypes in the total assemblage were NosZ6, derived from an uncultured clone from salt marsh sediments (Kearns et al., 2015), which is closely related to Marinobacter, and NosZ65, derived from Marinobacter sp. BSs20148 from marine sediment (Song et al., 2013) (**Figure 3A** and Supplementary Table 3). In contrast, NosZ6 and NosZ65 were not dominant in the active assemblage (**Figure 3B** and Supplementary Table 4).

The sample from the ODZ (130 m) at station BB1 illustrates the contrasts observed between total and active nosZ assemblages. The FRns of the top three dominant archetypes (WNZ21, WNZ16 and NosZ65) in the total nosZ assemblage were comparable to each other, and they constituted 56.1% of the total community (Supplementary Table 3). However, NosZ65 was nearly undetectable in the active assemblage and the top two dominant archetypes (WNZ21 and WNZ16) accounted for 87.8% of the active assemblage in the same sample (Supplementary Table 4). The total nosZ assemblage was much more diverse than the active assemblage at this depth (**Table 1**). The representative sequences of WNZ21 and WNZ16 archetypes were derived from nosZ gene sequences of Anaeromyxobacter dehalogenans strain DCP18 (Chee-Sanford et al., unpublished) and an uncultured bacterium clone obtained from agricultural soils (Sanford et al., 2012), respectively. WNZ21 and WNZ16 archetypes were not only dominant in the active assemblage in one sample from the ODZ, but were among the top five abundant archetypes of both total and active assemblages in almost all samples (**Figure 3**).

TABLE 1 | α-diversities of total (DNA) and active (RNA) nosZ genes at off-shore station BB1 and coastal station BB2.


#### Community Composition of Total and Active nosZ Assemblages

Functional gene microarrays were used to describe the community composition of nosZ assemblages. FRn values from duplicate microarrays replicated well (r <sup>2</sup> = 0.802–0.997) (Supplementary Figure 2) and each pair of duplicates clustered together in the DCA plots (**Figure 4**).

The two-dimensional DCA model including both DNA and RNA microarray results explained 43.7% of the community composition of nosZ assemblages with 31.6% explained by the first axis and 12.1% explained by the second axis (**Figure 4A**). The clearest pattern was the clear separation of total (filled symbols) and active (open symbols) nosZ assemblages, indicating that they were different from each other. The significance (P < 0.001) of the difference between total and active assemblages was confirmed by the dissimilarity test (**Table 2**). Therefore, the community composition of total and active nosZ assemblages was further analyzed by two DCA models separately to better examine other patterns.

The DCA model of DNA microarray results explained 63.9% of the composition of the total nosZ assemblage (**Figure 4B**).

The total nosZ community composition revealed site difference and O<sup>2</sup> dependence. Samples from the same station clustered together in the DCA model, indicating community composition was different between the off-shore station (BB1) and the coastal station (BB2). The site difference was statistically significant (P < 0.001) based on the dissimilarity test (**Table 2**). Besides the geographical pattern, composition of the total nosZ assemblage was also affected by O<sup>2</sup> concentration (**Figure 4B**). Samples from the ODZs clustered together, while samples with higher O<sup>2</sup> concentrations were distinct from the ODZ samples and different from each other. O<sup>2</sup> concentration of the seawater might not be the most important driver of microbial community composition, however, since the O<sup>2</sup> pattern was not captured by either axis of the DCA model.

The DCA model of RNA microarray results explained 40.2% of the community composition of the active nosZ assemblage (**Figure 4C**). Significant site difference (P = 0.039) of the active community composition was also revealed by the DCA model (**Figure 4C**) and the dissimilarity test (**Table 2**). However, the clustering based on O<sup>2</sup> concentration that was observed in the total nosZ assemblage was not observed for the active nosZ assemblage.

#### Environmental Variables Correlated with the Community Composition

The composition of total and active nosZ assemblages was correlated with different environmental variables based on a Mantel test (**Table 3**). Relative depth, nitrate concentration, temperature, density (sigma theta) and pressure were significantly related to the β-diversity of the total nosZ assemblage. However, the β-diversity of the active nosZ assemblage was significantly related to N2O concentration, nitrite concentration and fluorescence. Bottom depth, which was dramatically different between two stations and is a proxy for important ecological differences between the two sites, was a significant factor for both total and active nosZ assemblages.

## DISCUSSION

#### Abundance and Diversity of nosZ Assemblages

Oxygen minimum zones are sites of high N2O flux to the atmosphere (Law and Owens, 1990; Arévalo-Martínez et al., 2015). N2O consuming organisms are the only biological sink for N2O. Hence their abundance and community composition in the OMZ may be important in understanding the N2O flux. The abundance of total and active N2O consuming bacteria in the OMZ of the ETSP was estimated by measuring nosZ gene copy number (**Figure 1**). The relationship between abundance of N2O consuming bacteria and depth in this study differed from that of denitrifiers indicated by nirS gene copy number at the same stations (Ji et al., 2015): the abundances of the total and active N2O consuming bacteria in the surface water were similar to or higher than those in the ODZs (**Figure 1**), but the abundance of denitrifiers in the surface water was two orders of magnitude smaller than that in the ODZs. This difference suggests that the two genes represent functionally different groups.

nirS and nosZ also differed in their absolute abundance. The highest abundance of N2O consuming bacteria was only a few hundred copies mL−<sup>1</sup> , which was three orders of magnitude smaller than the highest abundance of denitrifiers measured at the same stations (Ji et al., 2015). It is assumed that both genes are present in the genome as single copy genes, although there are exceptions for nosZ (Sanford et al., 2012). One possible explanation for the differences in both distribution and abundance of the nosZ assemblage and the nirS assemblage is that not all N2O consuming bacteria contain the complete denitrification gene sequence (Sanford et al., 2012). The atypical nosZ genes are associated with bacteria that lack the other steps in the conventional denitrification pathway. Notably, bacteria with only nosZ genes but no other denitrification genes were overrepresented in the genomes of marine bacteria compared to other ecosystems (Graf et al., 2014). nirS, however, was



preferentially associated with bacteria that contained a complete denitrification pathway (Graf et al., 2014).

Another contributing factor may be the specificity or bias of the PCR primers. The nosZ primers used in this study were optimized to amplify all known nosZ sequences as of 2006, and should therefore represent the large database of both terrestrial and marine sequences available at the time. However, it is clear that they might underrepresent the atypical N2O consuming bacteria, which were not known at the time. The nirS primers used in the previous analysis of these samples (Ji et al., 2015) are potentially biased toward marine sequences (Braker et al., 1998) and may underrepresent more diverse sequences now available from other environments. One way to improve the nosZ coverage is to use multiple primer sets targeting different groups of nosZ archetypes.

The N2O consuming bacteria are a small component of the total microbial assemblage, but are still quite diverse (Jones et al., 2013), so they are difficult to characterize by pure culture or metagenomics. The microarray, which was designed specifically to target N2O consuming bacteria using more than 100 nosZ gene probes (**Figure 2**), may be a better tool to capture these underrepresented organisms without cultivation or detection of rare sequences in complex metagenomic datasets. The high reproducibility of microarrays reported previously (Bulow et al., 2008) was confirmed in this study in that duplicates for each sample run on two different microarrays clustered together



<sup>1</sup>Relative depth was calculated by dividing measured depth by the bottom depth of each station. <sup>2</sup>Sigma Theta was density calculated with in situ salinity and potential temperature at zero pressure. Bolded P values indicate significant correlation (P < 0.05).

in DCA (**Figure 4**) and had high r <sup>2</sup> of linear regressions (Supplementary Figure 2).

Based on the FRn values of diverse nosZ archetypes determined by microarray hybridization (**Figure 2**), a very limited number of archetypes dominated the total or the active assemblages. Moreover, the top five active archetypes accounted for larger percentage of the assemblage than that of the total archetypes (**Figure 3**), consistent with the less diverse active assemblage compared to the total assemblage (**Table 1**). These findings imply that although the total nosZ assemblage is very diverse, a few nosZ archetypes might be the major contributors of N2O consumption at the study sites. The relative abundance of active archetypes, however, might uncouple that of enzymes of different archetypes and/or the contribution of different archetypes to the N2O consumption rate due to different stabilities of enzymes from different archetypes.

#### Total and Active Community Compositions and Their Controlling Environmental Variables

The active nosZ assemblage was different from the total assemblage in both abundance profiles and community composition as detected by qPCR and nosZ microarray hybridization analysis, respectively. The highest abundance of the total N2O consuming bacteria indicated by nosZ DNA copy number was 636.4 (±28.3) copies mL−<sup>1</sup> in the sample from the ODZ (115 m) at station BB2, but the abundance of the active bacteria in the same sample was only 21.1 (±5.2) copies mL−<sup>1</sup> (**Figure 1**). In the sample from oxygenated surface seawater (60 m) at station BB1, the abundance of the active N2O consuming bacteria was 604.6 (±103.7) copies mL−<sup>1</sup> , but the abundance of the total bacteria was only 357.2 (±12.5) copies mL−<sup>1</sup> (**Figure 1**).

Significant differences between the community composition of total and active N2O consuming bacteria were indicated by the results of DCA and dissimilarity test (**Figure 4A**; **Table 2**). The different community composition was attributed to the differences in the distribution of FRn of more than 100 archetypes, especially the dominant ones. NosZ6 and NosZ65 were dominant in the total nosZ assemblage but were minor components in the active assemblage. NosZ42 was the most abundant archetype in the total nosZ assemblage from the lower oxycline (1000 m) at station BB2, but was not among the top five archetypes of the active assemblage. These differences between active and total nosZ communities are consistent with observations from soil and salt marsh sediments in which the active component of the microbial assemblage was apparently more responsive to environmental conditions (Barnard et al., 2013; Kearns et al., 2016).

Despite the differences between the total and the active N2O consuming assemblages, their distributions both depended on geographic location. The composition of total and active nosZ assemblages was significantly different between the coastal and the off-shore stations as indicated by the results of DCA and dissimilarity test. Since the two stations shared similar dominant nosZ archetypes, the geographical divergence mainly reflected differences among the large number of rare archetypes between the two stations. The more negative N<sup>∗</sup> at the coastal station BB2 (Supplementary Table 1) indicates more intense nitrogenloss fueled by more organic matter. The different amount of organic matter, which supports the metabolism of heterotrophic nosZ bacteria, might partially contribute to the geographical differences of nosZ assemblages. Geographical differences might also result from different nutrient sources at the two stations, since their distance to the sediment (bottom depth) and to the shore were dramatically different. The dependence of geographic location was also observed for ammonia oxidizing archaea (Peng et al., 2013) and for nirS denitrifiers (Jayakumar et al., 2013) in the ETSP and Arabian Sea OMZs.

In addition to geographical patterns, the total and active assemblages were correlated with different environmental variables (**Table 3**). Depth and environmental parameters that co-varied with depth (including temperature, density and pressure) were major drivers of the β-diversity of the total N2O consuming assemblages, implying different organisms coexist in the water column by occupying different ecological niches. However, N2O concentration difference was a major driver of the β-diversity of the active nosZ assemblages, implying the active nosZ community was a better indicator for N2O consumption potential.

#### nosZ Assemblage in Oxygenated Seawater

The role of the nosZ assemblage in oxygenated seawater has been ignored because N2O consumption is considered the least oxygen tolerant anaerobic step in the conventional denitrification pathway (Zumft, 1997). However, nosZ genes were abundant in oxygenated surface water in the Southern Indian Ocean (Raes et al., 2016) and nosZ mRNAs were detected in the

oxic regions in the Arabian Sea (Wyman et al., 2013). Our study confirmed that a nosZ assemblage was not only present but also active in oxygenated surface water in the OMZ of the ETSP. In particular, atypical nosZ archetypes, usually associated with N2O consuming bacteria lacking a complete denitrification pathway, were present and active in surface waters. In addition, the most abundant archetypes of total and active nosZ communities were both atypical nosZ archetypes (WNZ21 and WNZ16), implying the significant contribution of atypical archetypes to the nosZ communities and the necessity to consider atypical archetypes while analyzing the potential of N2O consumption.

N2O reductase enzymes from denitrifiers had very low O<sup>2</sup> tolerance (Bonin et al., 1989; Körner and Zumft, 1989); on the contrary, nosZ assemblages were detected in the oxygenated surface waters and O<sup>2</sup> concentration was not significantly correlated with the active microbial community, as indicated by DCA and Mantel test. The survival of N2O consuming bacteria in oxic layers and their O2-independence might be attributed to anoxic micro-environments created by phytoplankton microaggregates or particles. Free-living and particle-associated microbes from the same seawater sample can have different community compositions (Delong et al., 1993). More specifically, a recent study in the OMZ of the ETSP showed that nosZ mRNAs were 28-fold more abundant on particles (>1.6 µm) compared to free-living microbes (0.2–1.6 µm) (Ganesh et al., 2015). Additionally, nosZ mRNA co-occurred with the cyanobacterium Trichodesmium in oxic water in the Arabian Sea (Wyman et al., 2013). Consistently, fluorescence, a proxy for chlorophyll a, was significantly correlated with the β-diversity of the active nosZ assemblages in this study (**Table 3**).

The active nosZ community in oxygenated surface water might capture N2O produced in deeper seawater and thus reduce the flux into the atmosphere. Thus, evaluating the nosZ community is essential to the prediction of the oceanic N2O emissions. Moreover, the oceanic N2O emissions represent net fluxes, which are controlled by both N2O production and N2O consumption. Some N2O flux models (Suntharalingam and Sarmiento, 2000; Martinez-Rey et al., 2015; Trimmer et al., 2016) do not parameterize N2O consumption, and other models either consider N2O consumption only in suboxic or anoxic waters (Cornejo and Farías, 2012; Babbin et al., 2015) or estimate N2O consumption assuming it is constrained by O<sup>2</sup> concentration (Zamora et al., 2012). Failing to consider the O2-independent, non-denitrification N2O consumption potential in these O<sup>2</sup> forcing models might contribute to their uncertainty and the variation among different models. Additionally, the N2O consuming organisms have not been fully investigated. Besides denitrifiers and atypical N2O consuming bacteria analyzed in this study, other organisms (Trichodesmium and Crocosphaera) also exhibited N2O consuming capacity under laboratory conditions (Farías et al., 2013), suggesting that their significance in the environment warrants further investigation.

## CONCLUSION

The results described above support two (1 and 3) of the initial hypotheses. (1) Compositions of total and active nosZ assemblages were different between the coastal station and the off-shore station mainly due to their dramatic differences of distance to the sediment and to the shore, which are very likely to result in different environmental conditions (i.e., different phytoplankton assemblages, different nutrients and organic matter). (2) The abundances of total and active nosZ assemblages in oxygenated seawater were similar to or larger than those in the ODZs, implying the potential for N2O consumption even in oxygenated surface water. Atypical nosZ archetypes, which may lack a complete denitrification pathway, dominated both total and active nosZ assemblages. (3) The total and active nosZ assemblages were significantly different from each other. The community composition of the total nosZ assemblage showed O<sup>2</sup> dependence and shifted along depth gradients and environmental gradients associated with depth, but fluorescence, N2O and nitrite concentration were significantly correlated with the composition of the transcriptionally active community. We conclude that the difference between active and total nosZ assemblages may be related to differential response to environmental conditions by different components of the diverse natural assemblage and that the presence of nosZ assemblage in surface waters should be investigated to determine their actual N2O reduction capabilities.

#### AUTHOR CONTRIBUTIONS

XS and BW designed the experiments. AJ and BW collected samples. XS and AJ performed experiments. XS analyzed the data. XS and BW wrote the paper.

## FUNDING

This paper was supported by an NSF grant to BW and AJ (OCE-1029951).

## ACKNOWLEDGMENT

We would like to acknowledge all scientists and the crew of the R/V Nathaniel B. Palmer for assistance in sample collection.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.01183/full#supplementary-material

#### REFERENCES

fmicb-08-01183 June 23, 2017 Time: 14:46 # 10


oxygen minimum zones in the ocean. Aquat. Microb. Ecol. 70, 245–259. doi: 10.3354/ame01654



Pacific: insights from the MEMENTO database. Biogeosciences 9, 5007–5022. doi: 10.5194/bg-9-5007-2012

Zumft, W. G. (1997). Cell biology and molecular basis of denitrification. Microbiol. Mol. Biol. Rev. 61, 533–616.

**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 Sun, Jayakumar and Ward. 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.

# Sulfur Metabolism of *Hydrogenovibrio thermophilus* Strain S5 and Its Adaptations to Deep-Sea Hydrothermal Vent Environment

Lijing Jiang1, 2, 3†, Jie Lyu1, 2, 3† and Zongze Shao1, 2, 3, 4 \*

*<sup>1</sup> Key Laboratory of Marine Genetic Resources, Third Institute of Oceanography, State Oceanic Administration, Xiamen, China, <sup>2</sup> Fujian Key Laboratory of Marine Genetic Resources, Xiamen, China, <sup>3</sup> Fujian Collaborative Innovation Center of Marine Biological Resources, Xiamen, China, <sup>4</sup> Laboratory for Marine Biology and Biotechnology, Qingdao National Laboratory for Marine Science and Technology, Qingdao, China*

#### *Edited by:*

*Hongyue Dang, Xiamen University, China*

#### *Reviewed by:*

*Meng Li, Shenzhen University, China Guangyi Wang, Tianjin University, China Maggie Lau, Princeton University, United States*

*\*Correspondence:*

*Zongze Shao shaozz@163.com †Co-first authors.*

#### *Specialty section:*

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

*Received: 03 July 2017 Accepted: 04 December 2017 Published: 13 December 2017*

#### *Citation:*

*Jiang L, Lyu J and Shao Z (2017) Sulfur Metabolism of Hydrogenovibrio thermophilus Strain S5 and Its Adaptations to Deep-Sea Hydrothermal Vent Environment. Front. Microbiol. 8:2513. doi: 10.3389/fmicb.2017.02513* *Hydrogenovibrio* bacteria are ubiquitous in global deep-sea hydrothermal vents. However, their adaptations enabling survival in these harsh environments are not well understood. In this study, we characterized the physiology and metabolic mechanisms of *Hydrogenovibrio thermophilus* strain S5, which was first isolated from an active hydrothermal vent chimney on the Southwest Indian Ridge. Physiological characterizations showed that it is a microaerobic chemolithomixotroph that can utilize sulfide, thiosulfate, elemental sulfur, tetrathionate, thiocyanate or hydrogen as energy sources and molecular oxygen as the sole electron acceptor. During thiosulfate oxidation, the strain produced extracellular sulfur globules 0.7–6.0µm in diameter that were mainly composed of elemental sulfur and carbon. Some organic substrates including amino acids, tryptone, yeast extract, casamino acids, casein, acetate, formate, citrate, propionate, tartrate, succinate, glucose and fructose can also serve as carbon sources, but growth is weaker than under CO<sup>2</sup> conditions, indicating that strain S5 prefers to be chemolithoautotrophic. None of the tested organic carbons could function as energy sources. Growth tests under various conditions confirmed its adaption to a mesophilic mixing zone of hydrothermal vents in which vent fluid was mixed with cold seawater, preferring moderate temperatures (optimal 37◦C), alkaline pH (optimal pH 8.0), microaerobic conditions (optimal 4% O2), and reduced sulfur compounds (e.g., sulfide, optimal 100µM). Comparative genomics showed that strain S5 possesses more complex sulfur metabolism systems than other members of genus *Hydrogenovibrio*. The genes encoding the intracellular sulfur oxidation protein (*DsrEF*) and assimilatory sulfate reduction were first reported in the genus *Hydrogenovibrio*. In summary, the versatility in energy and carbon sources, and unique physiological properties of this bacterium have facilitated its adaptation to deep-sea hydrothermal vent environments.

Keywords: *Hydrogenovibrio*, *Thiomicrospira*, sulfur oxidation, *DsrEF*, assimilatory sulfate reduction, hydrothermal vent

## INTRODUCTION

The environments of deep-sea hydrothermal vents are characterized by steep gradients of physical and chemical parameters in the mixing zones between hot vent fluids and cold deep-sea water. Despite these extreme conditions, vent ecosystems develop quite well based on primary production conducted by chemolithoautotrophic microbes that are either free-living or associated with invertebrates as symbionts (Sievert et al., 2008). In these ecosystems, carbon fixation driven by sulfur oxidation is a major process of primary production. In black chimney ecosystems, hydrogen sulfide in hydrothermal fluids generated by seawater-rock interactions in the sub-seafloor (Jannasch and Mottl, 1985) serves as the major energy source for chemolithoautotrophs (McCollom and Shock, 1997). In addition to hydrogen sulfide, elemental sulfur, thiosulfate and polysulfide can be found in both the mixing zones and far away from the vents (Mullaugh et al., 2007; Gartman et al., 2011; Beinart et al., 2015). These partially oxidized inorganic sulfur compounds can be further oxidized by chemolithoautotrophic sulfur-oxidizing bacteria (CSOB).

Hydrogenovibrio bacteria are a type of CSOB that were first identified around the Galapagos Rift vents, and originally described as strains of the genus Thiomicrospira (Ruby et al., 1981). However, members of the genus Thiomicrospira from deep-sea hydrothermal vents were reclassified to the genus Hydrogenovibrio based on phylogeny, physiology and morphology during preparation of this manuscript (Boden et al., 2017). To date, several bacteria belonging to this genus have been isolated from deep-sea hydrothermal vent environments and characterized, including strain L-12 (Ruby and Jannasch, 1982), strain TH-55 (Jannasch et al., 1985), strain MA2-6 (Brinkhoff and Muyzer, 1997), strain MA-3 (Wirsen et al., 1998), strain XCL-2 (Ahmad et al., 1999), strain I78 (Takai et al., 2004), strain SP-41(Hansen and Perner, 2014), and strain EPR85 (Houghton et al., 2016). However, only Thiomicrospira crunogenus and Thiomicrospira thermophilus have been subjected to species description, (Jannasch et al., 1985; Takai et al., 2004), which resulted in their recently being renamed as Hydrogenovibrio crunogenus and Hydrogenovibrio thermophilus (Boden et al., 2017). In addition, culture-independent methods revealed that bacteria of this genus were predominant members of communities in hydrothermal vent samples, such as the Lost City carbonate chimney and sulfide chimneys in the Southwest Indian Ridge (SWIR) (Brazelton and Baross, 2010; Cao et al., 2014). Thus, members of this genus are proposed to play an important role in sulfur and carbon cycling in hydrothermal vent systems. To date, only one deep-sea hydrothermal vent bacterium in the genus (H. crunogenus XCL-2) has had its genome completely sequenced (Scott et al., 2006). Genomic analysis indicated that strain XCL-2 was an obligate chemolithoautotroph, possessing genes encoding a carbonconcentrating mechanism and Calvin-Benson-Bassham (CBB) cycle. Oxidation of reduced sulfur compounds by strain XCL-2 relied on the Sox system and sulfide:quinone reductase (SQR). Moreover, Brazelton and Baross (2010) reported metagenomic sequences derived from a carbonate chimney of the Lost City vent field on the MAR that were highly similar to the genome of strain XCL-2, even though they inhabit different niches in two different hydrothermal vent systems (a black chimney and a white chimney) in two separate oceans. These findings imply that these organisms are highly adaptable in hydrothermal vent systems.

The SWIR is recognized as an ultraslow spreading ridge that has not been thoroughly characterized in terms of geology, geochemistry and ecology. In a previous study, metagenomics analysis revealed that members of the genus Thiomicrospira, now reclassified as Hydrogenovibrio, were relatively abundant in the black chimney sulfides of the SWIR (Cao et al., 2014). However, defining their roles in situ requires more studies of their physiology, metabolism, and adaptation to the deep-sea hydrothermal vents. In this report, we isolated a bacterium of Hydrogenovibrio from an active hydrothermal vent chimney on the SWIR that was closely related to H. thermophilus (Takai et al., 2004; Houghton et al., 2016). Further analyses were conducted to characterize its physiology, annotate its genome and investigate its sulfur oxidation processes. The results denote its role and adaption to micro-niches in hydrothermal vent environments.

#### MATERIALS AND METHODS

#### Sample Collection, Enrichment, and Isolation

During the COMRA DY30 oceanic research cruise (March 2014), black chimney samples were collected from an active hydrothermal vent with a remotely operated vehicle (ROV) "Hailong II" from a depth of 2,742 m on the SWIR (49◦ 39′E, 37◦ 47′ S; Site 30III-S005-ROV01). Aboard the research vessel Da-yang Yi-hao, samples were immediately transferred into 100 ml glass bottles under a gas phase of 100% N<sup>2</sup> (100 kPa), containing 50 ml sterilized MJ synthetic seawater (Takai et al., 1999) and 0.05% (w/v) sodium sulfide. The suspended slurry was inoculated into MMJHS medium (Takai et al., 2003) under a gas phase mixture of 80% H2, 18% CO2, and 2% O<sup>2</sup> (200 kPa), after which the culture was incubated at 28◦C. Cells were then purified with the dilutionto-extinction technique using the same medium. The purity of the culture was confirmed by microscopic examination and 16S ribosomal RNA (rRNA) gene sequencing.

## Growth Characteristics

The physiological characterization of the isolate was detected on MMJS medium (Takai et al., 2003). After autoclaving, the medium was dispensed into 50 ml serum bottles, then sealed with a butyl-rubber stopper under a gas phase mixture of 80% N2, 18% CO<sup>2</sup> and 2% O<sup>2</sup> (200 kPa). Unless otherwise stated, the experiments were conducted in triplicate. Bacterial growth was measured by spectrophotometry and direct cell counting using a phase contrast microscope (Eclipse 80i, Nikon, Japan). The optimum growth conditions were tested under various parameters, including different temperatures (20◦C, 25◦C, 28◦C, 30◦C, 35◦C, 37◦C, 40◦C, and 45◦C), salt concentrations (0, 1, 2, 3, 4% (w/v) NaCl), pH (5.5, 7.0, and 8.0), and oxygen concentrations (gas phase settings of 0, 2, 4, 6, 8, 10, and 20%). In the case of oxygen absence, 10 mM nitrate was added as a potential electron acceptor.

Heterotrophic growth was tested in a NaHCO3-minus MMJS medium containing potential organic carbon sources under a gas phase of 4% O2: 0.1% (w/v) peptone, yeast extract, tryptone, starch, casein and casamino acids, 5 mM of acetate, formate, citrate, tartrate, succinate, propionate and pyruvate, 5 mM each of 20 amino acids, 0.02% (w/v) sucrose, galactose, glucose, lactose, fructose, maltose and trehalose.

#### Oxidation of Inorganic Sulfur Compounds and Hydrogen

The ability for sulfur oxidation was tested in the MMJS medium using various sulfur compounds at different concentrations as the sole energy source, including thiosulfate (5, 10, 15, 20, 30 mM), sodium sulfide (50, 100, 200, 400, 800µM, 1, 2 mM), sulfite (5 mM), elemental sulfur (1% w/v), thiocyanate (5 mM) or tetrathionate (5 mM). Growth characteristics along with sulfur oxidation were examined under the optimal growth conditions with 10 mM Na2S2O<sup>3</sup> and 100µM Na2S as the sole energy source, respectively. When grown with thiosulfate or sodium sulfide, Hydrogenovibrio bacteria produced colloidal elemental sulfur, which was usually influenced by culture conditions such as the pH of the medium, which was consistent with previous reports (Javor et al., 1990; Houghton et al., 2016). Therefore, the effects of pH on the accumulation of extracellular elemental sulfur were determined in the MMJS medium, which was adjusted to pH 5.5 with citrate buffer, left neutral (pH 7.0) or adjusted to pH 8.0 with Tris-HCl buffer, respectively. The effects of oxygen concentration on the accumulation of elemental sulfur were also tested under different oxygen concentrations (2, 4, 6, 8, 10, and 20%) in the gas phase. In addition, hydrogen oxidation was tested in MMJS medium without any reduced sulfur compounds under a gas phase mixture of 80% H2, 18% CO<sup>2</sup> and 2% O<sup>2</sup> (200 kPa).

#### Analysis of Sulfur Compounds

Thiosulfate, sulfate and sulfite were determined by ion chromatography (ICS-3000, Dionex, USA). The cultures were initially processed by centrifugation (6,000 × g, 15 min), then filtered through a 0.45µm membrane filter as previously described (Jiang et al., 2009). Filtered samples were analyzed immediately. The sample was then purified through an analytical column (Dionex IonPac AS11-HC, 4.6 × 250 mm) using NaOH (20 mM) as an eluent. The flow rate, injection volume, column temperature and suppressor current were 0.8 ml/min, 25 µl, 30◦C, and 40 mA, respectively. Elemental sulfur generated during the oxidation of thiosulfate was determined using the method developed by Li et al. (2006), which included spectrophotometric analysis for indirect quantification of sulfur produced by microorganisms. Briefly, sulfur was extracted with chloroform, then evaporated to dryness. Subsequently, the residue was dissolved in ethanol and treated with excess bisulfite to convert thiosulfate. The thiosulfate produced was then reacted with excess iodine (I2), after which the remaining I<sup>2</sup> was determined by a spectrophotometric method.

#### Electron Microscopic Analysis of Sulfur Globules

Extracellular sulfur globules were observed using scanning electron microscopy (SEM). Samples were prepared using a modified procedure from Bae et al. (2006). Cells were centrifuged at 2,000 rpm for 30 min after cultivation, then rinsed in phosphate buffer solution (PBS) (pH 7.4). The supernatant was subsequently removed and the cell pellets were fixed with 2.5% (w/v) glutaraldehyde in 0.1 M PBS (pH 7.4) for 2 h. Next, the pellets were dehydrated with ethanol stepwise with increasing concentrations (30, 50, 70, and 90%) over 10 min intervals. Finally, the cells were resuspended in absolute ethanol for 20 min. Samples were dried for 24 h to remove ethanol before SEM examination (Hitachi S-4800, Japan) at 5.0 KV.

#### DNA Extraction, Genome Sequencing, and Phylogenetic Analysis

Genomic DNA was extracted using the method described by Jiang et al. (2009). The quality and quantity of the extracted DNA were then determined using agarose gel electrophoresis and a NanoDrop system (Thermo NanoDrop 2000, Wilmington, Delaware, USA). The complete genome of strain S5 was sequenced by Shanghai Majorbio Biopharm Technology Co., Ltd. (Shanghai, China) using a combination of Illumina Hiseq 4000 (2 × 150 bp) and Illumina MiSeq (2 × 250 bp) platforms (Illumina, USA). Generated raw reads were first filtered to remove adapters and low-quality reads. The draft genome sequence was then assembled based on clean data generated from the Illumina Hiseq platform. The Illumina Miseq reads were used to fill in gaps, correct potential base errors and increase the consensus quality. Gaps were then filled in by sequencing the PCR products using a capillary sequencer (ABI 3730XL, ABI, USA). The complete genome sequence was assembled with SOAPdenovo (version 2.04) (http://soap.genomics.org.cn/). Gene annotation was performed by Rapid Annotation using Subsystem Technology (RAST) server (Aziz et al., 2008) and the NCBI Prokaryotic Genomes Automatic Annotation Pipeline (PGAAP). The gene functions and metabolic pathways were analyzed by searching against the Kyoto Encyclopedia of Genes and Genomes and Clusters of Orthologous Groups databases. The phylogenetic relationships of the retrieved 16S rRNA gene sequences were identified by BLAST searches (http://www.ncbi. nlm.nih.gov/BLAST) of the GenBank database. The 16S rRNA gene sequences of closely related taxa obtained from the GenBank database were aligned using CLUSTAL X1.83 (Thompson et al., 1997). Phylogenetic analysis was conducted using the MEGA 6.0 Program (Tamura et al., 2013). Distance matrices were calculated based on the Kimura two-parameter method (Kimura, 1980). Phylogenetic trees were inferred using the neighborjoining method (Saitou and Nei, 1987) and Bootstrap values were determined based on 1,000 replications.

## RESULTS

#### Purification and Morphology

Enrichment cultures were grown in liquid MMJHS medium with CO<sup>2</sup> as the carbon source, thiosulfate and hydrogen as energy sources, and oxygen as a terminal electron acceptor. After 2 days of incubation, growth was observed at 28◦C. The enriched culture consisted of dense populations of small, short rods and produced colloidal elemental sulfur. The cells were subsequently purified three times using the dilution-to-extinction technique at 28◦C. This culture was designated as strain S5. The cells were Gramnegative with slightly curved rods about 1.5–2.5µm long and 0.4–0.7µm wide and motile with a polar flagellum.

#### Phylogenetic Analyses

A BLAST search of the obtained 16S rRNA gene sequence (1,553 bp) showed that strain S5 was most closely related to H. thermophilus strain I78 (Takai et al., 2004; Boden et al., 2017) and H. thermophilus strain EPR85 (Houghton et al., 2016) with 100% 16S rRNA gene sequence similarity. Strain S5 exhibited similarities <98% with other isolates of H. crunogenus (Boden et al., 2017), including strain TH-55 (Jannasch et al., 1985), strain SP-41 (Hansen and Perner, 2014) and strain XCL-2 (Scott et al., 2006). The 16S rRNA gene phylogeny suggests that strain S5 clusters with bacteria of H. thermophilus, which was supported by a high bootstrap value of 100% (**Figure 1**).

#### Physiological Characteristics

Growth tests revealed that the isolate grew in a range of temperatures (20◦C−45◦C), salinities (1–4% NaCl) and oxygen concentrations (2%−20%), but that optimum growth occurred at 37◦C, 4% O<sup>2</sup> and 3% NaCl. No growth was observed in the absence of oxygen, and only weak growth occurred under 20% O<sup>2</sup> in the gas phase, indicating that strain S5 preferred microaerobic growth conditions. After 24 h of incubation, the cells grew under both alkaline and acidic conditions, with the highest biomass occurring at pH 8.0 (2.0 × 10<sup>8</sup> cells ml−<sup>1</sup> ), followed by pH 7.0 (biomass of 1.4 × 10<sup>8</sup> cells ml−1) and the lowest biomass (1.1× 10<sup>8</sup> cells ml−<sup>1</sup> ) being observed for acidic cultures (pH 5.5).

The chemoautotrophic growth tests showed that strain S5 could utilize thiosulfate, sulfide, elemental sulfur, tetrathionate or thiocyanate as the sole energy source, but that it could not utilize sulfite, and elemental sulfur and thiocyanate only supported weak growth. Molecular hydrogen could also be used as an energy source. Heterotrophic growth tests showed that strain S5 was unable to grow using any of the tested organic carbons as the sole energy source, but when reduced sulfur compounds such as thiosulfate were present it grew with various organic carbons sources, including each kind of 20 amino acids, tryptone, yeast extract, casamino acids, casein, acetate, formate, citrate, propionate, tartrate, succinate, glucose, and fructose. Relatively high biomasses were obtained when cells utilized tryptone, methionine, arginine, glutamate or hydroxyproline as a carbon source, but the biomasses were less than those obtained under chemoautotrophic conditions. Other organic compounds, including peptone, starch, pyruvate, sucrose, galactose, lactose maltose and trehalose, failed to support its growth. These results confirm that strain S5 prefers a chemolithoautotrophic lifestyle.

#### Oxidation of Reduced Sulfur Compounds

The growth on reduced sulfur compounds was further determined in more detail using different concentrations of thiosulfate (5, 10, 15, 20, 30 mM) and sulfide (50, 100, 200, 400, 800µM, 1, 2 mM) as sole energy sources. Strain S5 grew with thiosulfate at all tested concentrations from 5 to 30 mM, with 10 mM supporting maximal growth (1.16 × 10<sup>8</sup> cells ml−<sup>1</sup> ). Additionally, growth was observed in the presence of sulfide at 1 mM and below, with optimal growth (1.19 × 10<sup>8</sup>

Jiang et al. Sulfur Metabolism of *Hydrogenovibrio thermophilus*

cells ml−<sup>1</sup> ) occurring in the presence of 100µM sodium sulfide. No growth was observed when the sulfide concentration was 2 mM. The growth rate and sulfur oxidation rate were 0.35 h−<sup>1</sup> and 1.045 mM h−<sup>1</sup> , respectively, under optimal cultivation conditions with 10 mM thiosulfate (**Figure 2**). When strain S5 grew at pH 5.5 (in buffered medium) and pH 7.0 (unbuffered), elemental sulfur accumulated in the culture. These findings suggested that strain S5 performed incomplete thiosulfate oxidation, while sulfite was not detected as an intermediate during this process. However, alkaline conditions lead to further oxidation of elemental sulfur, such as at pH 8.0. Elemental sulfur generation from thiosulfate was also affected by oxygen concentrations in the gas phase. The maximum accumulation of elemental sulfur occurred in 4% oxygen, which was also the optimal condition for growth.

Elemental sulfur occurred in the culture at the start of the mid-exponential growth phase, but more accumulated at the end of the exponential phase and during the stationary phase (**Figure 2**). Morphological observations under SEM showed that elemental sulfur was in the form of globules outside the cell, with the size varying from 700 nm to 6.0µm (**Figure 3**). Energy dispersive X-ray spectroscopic analysis further revealed that the sulfur globules mainly contained carbon and sulfur. Elemental sulfur was the main component, accounting for 51.5–71.3% of the total (**Figure 3**).

#### Genomic Features

The complete genome of strain S5 comprised 2.77 Mb with a G+C content of 50.52 mol%, which was remarkably higher than that of three other Hydrogenovibrio bacteria (**Table 2**). Gene annotation based on RAST predicted 2598 genes, including 2463 protein-encoding genes, 49 tRNA-encoding genes and 12 rRNAencoding genes. We compared the genomes of closely related species available in public databases, including strains XCL-2 (Scott et al., 2006), MA2-6 (unpublished data) and XS5 (Zhang et al., 2016). The genome of strain S5 was similar to those of strains MA2-6 and XS5 in G+C content and coding density, but distinctly different from that of strain XCL-2 (**Table 2**). The average nucleotide identity (ANI) between strain S5 and other Hydrogenovibrio bacteria was similar at the genome level (**Table 2**). The ANI of strain S5 with strain MA2-6 was highest (96.94%), followed by that of strain XS5 (90.10%), and a much lower ANI (73.17%) with strain XCL-2 (**Table 2**), indicating that the genome of strain S5 was most similar to that of strain MA2-6.

#### Central Metabolism

#### Oxidation of Sulfur and Hydrogen

The genome of strain S5 contained homologs for genes soxAYXBCD, all of which are necessary for assembling a Sox system that oxidizes reduced sulfur compounds (**Table 2**). Strangely, the soxZ gene is absent in the strain S5 genome, but present in all other members of this genus (**Table 2**). Genes encoding the Sox components are not organized in a single cluster in strain S5, which seems common in this genus (**Table 2**). Strain S5 also contained a homolog of the sqr gene, which usually oxidizes sulfide to elemental sulfur, resulting in the deposition of sulfur outside the cells. Unlike H. crunogens strain XCL-2, strain S5 contained more genes involved in sulfur metabolism, including flavocytochrome-C sulfide dehydrogenase (FCC), intracellular sulfur oxidation protein (DsrEF), sulfite reductase (CysIJ) and sulfate adenylyltransferase (CysDH) (**Table 2**). All genes essential for the assimilation of sulfate reduction were

present in the genome of strain S5. No genes involved in dissimilatory sulfate reduction, such as sulfite reductase (DsrAB), electron transfer protein (DsrC) or membrane-bound electron-transporting complex (DsrMKJOP), were detected in this bacterium (**Table 2**).

Genomic analysis revealed that strain S5 contained two hydrogenase genes, [NiFe]-Hydrogenases Group 1 (hyaAB) and Group 2b (hupUV), which were organized into two operons. According to a recent classification of the hydrogenases in the genus Thiomicrospira (Hansen and Perner, 2016), Group 1 hydrogenase of strain S5 can be further categorized into cluster I, together with those from strain MA2-6 and H. marinmus. These were distinctly different from cluster II hydrogenases, which exclusively contains those from H. crunogenus (**Table 2**). In addition, Group 2b hydrogenase was also detected in strain MA2-6, but absent from the genome of H. crunogens XCL-2. Neither [NiFe]-Hydrogenases Group 1 nor Group 2b was found in the non-vent bacterium strain XS5.

#### Carbon Metabolism

All enzymes essential for central carbon metabolism are encoded by strain S5 (**Table 2**). Strain S5 was most similar to strain XS5 with respect to CO<sup>2</sup> fixation, and not its closest bacterium strain MA2-6 (**Table 2**). Like other members of the genus Hydrogenovibrio, strain S5 has a carbon-concentrating mechanism that facilitates the uptake of dissolved inorganic carbon. Similar to strains MA2-6 and XS-5, strain S5 possesses three types of carbonic anhydrases (α-, β-, and γ-class). This contrasts with H. crunogenus strain XCL-2, which only has two classes (α- and β-class) (**Table 2**). Moreover, strain S5 contains all genes encoding enzymes involved in the CBB cycle. Glycolysis and gluconeogenesis, the tricarboxylic acid cycle and the pentose phosphate pathway are also complete in this genome. Two pathways for acetate assimilation are present in the genome of strain S5: (1) a two-step reaction including acetate kinase and phosphate acetyltransferase and (2) a single step reaction including acetyl coenzyme A synthetase. Strain S5 also contains an acetate permease (ActP) function as an acetate transporter. In contrast, only acetyl coenzyme A synthetase for acetate assimilation was found in the genome of H. crunogenus XCL-2.

#### Respiration

For aerobic respiration, strain S5 contains genes encoding complexes I–IV of the respiratory chain, including NADH ubiquinone oxidoreductase (NuoABCDEFGHIJKLMN; complex I), succinate dehydrogenase (FrdABC; complex II), cytochrome bc1 complex (PetABC; complex III), cytochrome c oxidoreductase (CcoNOP; complex IV). In addition, strain S5 has a bd-type cytochrome oxidase (CydAB), which is absent in H. crunogenus XCL-2 (**Table 2**). A low oxygen adapted cbb3 type cytochrome c oxidase was also detected in strain S5. The presence of genes atpABCDEFGH in this genome indicates that the respiratory chain is linked to a F0F1 ATPase that generates ATP. Strain S5 cannot utilize nitrate as an electron acceptor based on physiological analyses. Similarly, no genes encoding nitrate reductase (NasA) or nitrite reductase (NirBD) were found in this genome or in strain MA2-6. However, both the nasA and nirBD genes were present in the genomes of the other two isolates (**Table 2**).

#### DISCUSSION

Hydrogenovibrio bacteria, previously described as members of the genus Thiomicrospira, have frequently been found surrounding deep-sea hydrothermal vents (Ruby and Jannasch, 1982; Jannasch et al., 1985; Brinkhoff and Muyzer, 1997; Wirsen et al., 1998; Ahmad et al., 1999; Takai et al., 2004; Brazelton and Baross, 2010; Cao et al., 2014; Hansen and Perner, 2014; Houghton et al., 2016). Our previous study, which was based on metagenomics, revealed that this group was one of the dominant populations in hydrothermal vent chimneys of the SWIR (Cao et al., 2014). These organisms likely play an important role in the biogeochemical cycles of sulfur; however, the mechanisms underlying their adaptations to such harsh environments remain unclear. In this study, a new strain was isolated from a sulfide chimney in the SWIR. On the basis of 16S rRNA gene phylogeny, physiological and genomic properties, the newly isolated strain was designated as Hydrogenovibrio thermophilus strain S5, and further subjected to characterization of sulfur oxidation and environmental adaptations. The results of this study highlight the specifications of Hydrogenovibrio bacteria to adapt to their vent niches, and imply their roles in the carbon and sulfur cycles of deep-sea hydrothermal environments.

#### Adaptation of *Hydrogenovibrio* Bacteria to Deep-Sea Hydrothermal Vents

With the exception of H. thermophilus strain I78, which grows at 55◦C, most Hydrogenovibrio isolates cannot tolerate high temperatures and usually grow at below 40◦C (**Table 1**). It has been implied that Hydrogenovibrio bacteria preferentially inhabit the mesophilic mixing zones of hydrothermal vents. Moreover, thermodynamic calculations have shown that the oxidation of sulfide is most favorable at relatively low temperatures (McCollom and Shock, 1997). Another key trait of deep-sea Hydrogenovibrio bacteria that facilitates its adaptation to vent environments is the requirement of oxygen for growth. Indeed, these organisms were all described as microaerobic bacteria and incapable of growth under anaerobic conditions (**Table 1**). Genomic analyses confirmed that sulfur-oxidizing pathways in all of the available genome-sequenced bacteria of this genus require O<sup>2</sup> as a terminal electron acceptor. Moreover, strain S5 possesses two cytochrome oxidases, a cbb3-type and a bdtype, which have been proposed to have a high affinity for O<sup>2</sup> and therefore play an important role in microaerobic growth (D'Mello et al., 1996; Preisig et al., 1996). Thus, the metabolically habitable areas of Hydrogenovibrio bacteria strictly require O2. H. thermophilus strain S5 can adapt to a wide range of oxygen concentrations, ranging from 2 to 20%, with an optimum of 4% (**Table 1**). The optimal growth pH of strain S5 was pH 8.0, which is consistent with seawater pH, and similar to that of other deepsea Hydrogenovibrio isolates (Wirsen et al., 1998; Takai et al., 2004). However, the optimal growth of strain I78 occurred at pH 6.0 (**Table 1**), even though strain S5 showed the highest 16S rRNA gene sequence similarity (100%) to strain I78 (Takai et al., 2004). Moreover, pH significantly affected the sulfur-oxidizing process of strain S5. At higher pH, strain S5 completely oxidized thiosulfate to sulfate, while thiosulfate was oxidized to elemental sulfur deposits outside of the cells under acidic conditions. The

TABLE 1 | Characteristics differentiating *Hydrogenovibrio thermophilus* strain S5 from other strains within the genus *Hydrogenovibrio* from different deep-sea hydrothermal vents.


*Strains: 1, strain S5; 2, Hydrogenovibrio thermophilus strain I78 (Takai et al., 2004); 3, Hydrogenovibrio crunogenus strain TH-55 (Jannasch et al., 1985); 4, Thiomicrospira sp. strain L-12 (Ruby and Jannasch, 1982); 5, Thiomicrospira sp. strain MA-3 (Wirsen et al., 1998).* +*, positive; –, negative.*


TABLE 2 | Comparison of the genomes between *Hydrogenovibrio thermophilus* strain S5 and its closest relatives based on RAST annotations in this study.

*The complete genome sequence of Hydrogenovibrio crunogenus strain XCL-2 (GenBank accession number: NC\_007520), and the draft genome sequences of Hydrogenovibrio sp. strain MA2-6 (GenBank accession number: NZ\_JOMK01000001) and Hydrogenovibrio sp. strain XS5 (GenBank accession number: LQBO00000000) are available in the NCBI database.* +*, present; –, absent.*

phenomenon was also observed in H. crunogenus (Javor et al., 1990) and H. thermophilus (Houghton et al., 2016). The vent fluids are typically acidic in basalt-hosted hydrothermal vents; hence, it is likely that Hydrogenovibrio species can generate elemental sulfur globules in situ.

H. thermophilus strain S5 can utilize a wide range of inorganic sulfur compounds, including sulfide and partially oxidized inorganic sulfur compounds (e.g., thiosulfate, elemental sulfur, and tetrathionate) (**Table 1**) and prefer moderately low concentrations of sulfide (0.1–1 mM). However, it cannot grow under hydrogen sulfide concentrations of 2 mM or greater, although it can oxidize sulfide to elemental sulfur for detoxification. Hydrogen sulfide is usually rich in the vent fluids of black hydrothermal chimneys, with concentrations typically in the millimolar range (Martin et al., 2008), while partially oxidized inorganic sulfur compounds, such as elemental sulfur, thiosulfate and polysulfide, occur in the areas around the vent and outer layers of the chimneys where the fluid mixes with oxygenic water. The utilization of partially oxidized inorganic sulfur compounds provides ecological advantages for strain S5, which could mitigate exposure to high temperatures and toxic chemicals in venting fluids.

In addition to inorganic sulfur compounds, Hydrogenovibrio species can also utilize H<sup>2</sup> and Fe2<sup>+</sup> as energy sources to fuel carbon fixation (Hansen and Perner, 2016; Barco et al., 2017). This versatile energy metabolism could enable these organisms to be widespread and highly adaptable at inhabiting different hydrothermal vent fields globally, including the East Pacific Rise, Pacific Ocean and SWIR. Comparison of strain S5 with other Hydrogenovibrio isolates revealed its physiological features (**Table 1**) and genomic characteristics (**Table 2**) in energy sources, optimal O<sup>2</sup> concentration for growth, sulfur-oxidizing genes, and hydrogenase. The results further demonstrated the specification and adaptation of Hydrogenovibrio bacteria to different hydrothermal vent environments.

## Sulfur Oxidation Pathways in *H. thermophilus* S5

To examine the metabolic pathways of sulfur oxidation, we generated the complete genome sequence of strain S5. Prior to this study, the only member from deep-sea hydrothermal vents to have its entire genome sequenced was H. crunogenus strain XCL-2 (Scott et al., 2006), which is distantly related phylogenetically and also significantly differs in physiological traits from strain S5 (**Tables 1**, **2**). Based on the genome data and growth tests, we drew a schematic model of sulfur oxidation in Hydrogenovibrio thermophilus strain S5 (**Figure 4**). Hydrogen sulfide is oxidized to elemental sulfur by two pathways, namely SQR and FCC. SQR transfers electrons from sulfide to the respiratory chain via quinone, while FCC donates electrons via cytochrome c (Chen et al., 1994; Visser et al., 1997; Griesbeck et al., 2000). SQR plays a more important role than FCC in phototrophic and chemotrophic sulfide oxidation (Chan et al., 2009). Moreover, SQR was found to be involved in sulfide detoxification. Therefore, it was presumed that SQR could play a more important role than FCC during sulfide oxidation in hydrothermal vents. However, FCC has a high affinity for sulfide, which might be an advantage for cells under very low concentrations of sulfide (Chan et al., 2009). Thiosulfate can be oxidized by a Sox system (SoxABCDXY) in the periplasm of strain S5. Despite the absence of the soxZ gene, strain S5 is still capable of completely oxidizing reduced sulfur compounds to sulfate. SoxZ is usually combined with SoxY as a heterodimeric carried protein (SoxYZ) to shuttle covalently attached intermediates between the enzymes of the Sox pathway (Grabarczyk et al., 2015), which is present in all other genomes of this genus (**Table 2**), which implies that SoxZ is not essential for strain S5 to oxidize thiosulfate. Incomplete oxidation of thiosulfate will lead to the formation of extracellular elemental sulfur, which was also observed in H. thermophilus (Houghton et al., 2016). Usually, SoxCD-containing bacteria cannot generate extracellular sulfur during thiosulfate oxidation; however, Hydrogenovibrio bacteria from deep-sea hydrothermal vent systems seem to be an exception. Elemental sulfur produced by the incomplete oxidation of sulfide or thiosulfate under lower pH was excreted outside of the cells of H. thermophilus strain S5 and formed sulfur globules (**Figure 3**). Generally, elemental sulfur produced in the periplasm can also be transported into the cytoplasm for further oxidation, which is usually performed by the dissimilatory sulfite reductase (DsrAB) or a heterodisulfide reductase (Hdr)-like enzyme (Dahl, 2009; Gregersen et al., 2011). Moreover, sulfur carrier proteins Rhd, TusA and DsrEFH were found to be involved in sulfur transfer from persulfide intermediates to DsrAB (Stockdreher et al., 2014). The protein DsrEFH occurs exclusively in sulfur oxidizers and is essential for the oxidation of sulfur stored in the intracellular sulfur globules of purple sulfur bacteria, such as Allochromatium vinosum (Dahl et al., 2008). In this study, the Dsr system and hdr genes were not detected in this genome. However, the homologs ofrhd and dsrEF genes were first discovered in Hydrogenovibrio isolates (**Table 2**). Taken together, these findings indicate that an unknown sulfur trafficking process might be present in this genus.

Extracellular sulfur globules are generated by a diverse group of chemotrophic and phototrophic bacteria (Dahl and Prange, 2006). However, few studies have investigated extracellular sulfur globules produced by chemoautotrophs (Prange et al., 2002; Dahl and Prange, 2006; Beard et al., 2011). In this study, the properties of sulfur globules secreted by H. thermophilus strain S5 cells were analyzed by SEM and EDX. This is the first report detailing sulfur globules from Hydrogenovibrio bacteria, and also the first report evaluating chemoautotrophic sulfur-oxidizing bacteria from deep-sea hydrothermal vents.

#### Assimilatory Sulfate Reduction in *H. thermophilus* S5

All known genes involved in assimilatory sulfate reduction were present in H. thermophilus strain S5. Sulfate uptake is performed by two sulfate permeases: (1) one belonging to the ArsB/NhaD superfamily and (2) another belonging to a high-affinity transporter SulP. The ArsB/NhaD permease is located in the immediate vicinity of gene clusters involved in assimilatory sulfate reduction. Therefore, the transporter might be involved in sulfate transport for sulfate assimilation.

Once inside the cell, the sulfate is activated to form adenylyl sulfate (APS), which is catalyzed by the enzyme ATP sulfurylase (CysDN). Subsequently, strain S5 directly reduces APS to sulfite by APS reductase (CysH) without further phosphorylation to phosphoadenosine-5'-phosphosulfate (PAPS) as an intermediate. The generated sulfite is further reduced to sulfide by an assimilatory sulfite reductase (CysIJ), which is quite distantly related to their dissimilatory counterparts (DsrAB). The final step is catalyzed by O-acetylserine (thiol)-lyase B (CysM), which synthesizes cysteine from O-acetylserine and sulfide. In addition, some studies have indicated that CysM could also utilize thiosulfate to produce cysteine (Kredich, 1992; Sievert et al., 2008). Hence, it is possible that the CysM in strain S5 performs a similar function for thiosulfate assimilation. This is the first report of the assimilatory sulfate reduction pathway in the genus Hydrogenovibrio. Assimilatory sulfate reduction enables microorganisms to reduce sulfate for the formation of amino acids, nucleic acids and sulfur-containing coenzymes. The result suggests that strain S5 is more versatile in sulfur sources than other members of this genus that do not have an assimilatory sulfate reduction pathway. Specifically, strain S5 can reduce sulfate and/or thiosulfate for the synthesis of cellular material when reduced sulfur compounds are limited in deep-sea hydrothermal vents.

#### CONCLUSIONS

H. thermophilus strain S5 has specifically adapted itself to hydrothermal vent environments, mainly via sulfur oxidation and carbon dioxide fixation. This organism seems to prefer deep-sea vent niches with moderate temperatures (37◦C), alkaline pH (pH 8.0), microaerobic conditions (4% O2), various reduced sulfur compounds (e.g., 100µM of sulfide) and H2. Furthermore, it adapts well to vent environments with steep physical and chemical gradients, where varied inorganic sulfur compounds, low concentrations of carbon dioxide, changing oxygen concentrations and pH values usually occur. A schematic model of sulfur metabolism in strain S5 was generated based on genomic analysis, providing the first detailed description of the genus Hydrogenovibrio. An assimilatory sulfate reduction pathway was found in the genus Hydrogenovibrio for the first time, which might support their growth far from the vent, where sulfide is depleted. In addition, the properties of extracellular sulfur globules, including their morphology, size and elemental composition, were documented in Hydrogenovibrio bacteria. The activity of H. thermophilus strain S5 in sulfur oxidation and the formation of extracellular sulfur globules highlight its role in sulfur cycling in deep-sea hydrothermal vent environments.

## AUTHOR CONTRIBUTIONS

LJ designed the study, analyzed the data and wrote the manuscript. JL did experiments, analyzed the data and wrote the manuscript. ZS designed the study and wrote the manuscript.

#### ACKNOWLEDGMENTS

We would like to thank the Research Vessel Da-Yang Yi-Hao crews for assisting with sample collection. This work was financially supported by the National Natural Science Foundation of China (No. 41672333), COMRA program (No. DY135-B-01), National Program on Key Basic Research Project (973 Program) (No.2012CB417300), Xiamen Ocean Economic Innovation and Development Demonstration Project

#### REFERENCES


(16PZP001SF16), and National Infrastructure of Natural Resources for Science and Technology Program of China (NIMR-2017-9).


**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 Jiang, Lyu and Shao. 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.

# Cultivation-Independent and Cultivation-Dependent Analysis of Microbes in the Shallow-Sea Hydrothermal System Off Kueishantao Island, Taiwan: Unmasking Heterotrophic Bacterial Diversity and Functional Capacity

Kai Tang<sup>1</sup> \* † , Yao Zhang<sup>1</sup>† , Dan Lin<sup>1</sup> , Yu Han<sup>1</sup> , Chen-Tung A. Chen<sup>2</sup> , Deli Wang<sup>1</sup> , Yu-Shih Lin<sup>2</sup> , Jia Sun<sup>1</sup> , Qiang Zheng<sup>1</sup> and Nianzhi Jiao<sup>1</sup> \*

<sup>1</sup> State Key Laboratory of Marine Environmental Science, Institute of Marine Microbes and Ecospheres, Xiamen University, Xiamen, China, <sup>2</sup> Department of Oceanography, National Sun Yat-sen University, Kaohsiung, Taiwan

Shallow-sea hydrothermal systems experience continuous fluctuations of physicochemical conditions due to seawater influx which generates variable habitats, affecting the phylogenetic composition and metabolic potential of microbial communities. Until recently, studies of submarine hydrothermal communities have focused primarily on chemolithoautotrophic organisms, however, there have been limited studies on heterotrophic bacteria. Here, fluorescence in situ hybridization, high throughput 16S rRNA gene amplicon sequencing, and functional metagenomes were used to assess microbial communities from the shallow-sea hydrothermal system off Kueishantao Island, Taiwan. The results showed that the shallow-sea hydrothermal system harbored not only autotrophic bacteria but abundant heterotrophic bacteria. The potential for marker genes sulfur oxidation and carbon fixation were detected in the metagenome datasets, suggesting a role for sulfur and carbon cycling in the shallow-sea hydrothermal system. Furthermore, the presence of diverse genes that encode transporters, glycoside hydrolases, and peptidase indicates the genetic potential for heterotrophic utilization of organic substrates. A total of 408 cultivable heterotrophic bacteria were isolated, in which the taxonomic families typically associated with oligotrophy, copiotrophy, and phototrophy were frequently found. The cultivation-independent and -dependent analyses performed herein show that Alphaproteobacteria and Gammaproteobacteria represent the dominant heterotrophs in the investigated shallow-sea hydrothermal system. Genomic and physiological characterization of a novel strain P5 obtained in this study, belonging to the genus Rhodovulum within Alphaproteobacteria, provides an example of heterotrophic bacteria

#### Edited by:

Stefan M. Sievert, Woods Hole Oceanographic Institution, United States

#### Reviewed by:

Jinjun Kan, Stroud Water Research Center, United States Qi-Long Qin, Shandong University, China

#### \*Correspondence:

Kai Tang tangkai@xmu.edu.cn Nianzhi Jiao jiao@xmu.edu.cn †These authors have contributed equally to this work.

#### Specialty section:

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

Received: 26 April 2017 Accepted: 07 February 2018 Published: 22 February 2018

#### Citation:

Tang K, Zhang Y, Lin D, Han Y, Chen C-TA, Wang D, Lin Y-S, Sun J, Zheng Q and Jiao N (2018) Cultivation-Independent and Cultivation-Dependent Analysis of Microbes in the Shallow-Sea Hydrothermal System Off Kueishantao Island, Taiwan: Unmasking Heterotrophic Bacterial Diversity and Functional Capacity. Front. Microbiol. 9:279. doi: 10.3389/fmicb.2018.00279

**184**

with major functional capacity presented in the metagenome datasets. Collectively, in addition to autotrophic bacteria, the shallow-sea hydrothermal system also harbors many heterotrophic bacteria with versatile genetic potential to adapt to the unique environmental conditions.

Keywords: shallow-sea hydrothermal system, metagenomics, genomics, heterotrophic bacteria, microbial community, Rhodovulum

#### INTRODUCTION

fmicb-09-00279 February 20, 2018 Time: 17:10 # 2

Approximately 50–60 shallow-sea hydrothermal systems at depths of less than 200 m below sea level are currently known, occurring near active coastal or submarine volcanoes, with systems located along arcs, mid-ocean ridges, and in island arcrelated environments and even in continental margins (Tarasov et al., 2005). While extensive microbial surveys of the deepsea hydrothermal systems have been conducted since the first discovery of deep-sea hydrothermal vents nearly 40 years ago (Corliss et al., 1979), attention has also been paid to their shallow-sea counterparts, which are much easier to access and can often be explored via SCUBA diving (Tarasov et al., 2005). The biological data for nearly 30 shallow-sea hydrothermal vent ecosystems have been published (Tang, 2014). Previous surveys of 16S rRNA genes using tag pyrosequencing and clone libraries have revealed the composition of microbial communities in shallow-sea hydrothermal environments, including the shallowsea hydrothermal systems located at Kueishantao Island off NE Taiwan (Zhang et al., 2012; Tang et al., 2013), Eolian Islands (Vulcano Island and Panarea Island) in Italy (Manini et al., 2008; Maugeri et al., 2009, 2010, 2013b), Ambitle Island (Meyer-Dombard et al., 2012), D. João de Castro Bank, Azores (Mohandass et al., 2012), Milos Island in Greece (Brinkhoff et al., 1999; Sievert et al., 1999, 2000; Bayraktarov et al., 2013; Giovannelli et al., 2013; Price et al., 2013), Taketomi Island in Japan (Hirayama et al., 2007), and Eyjafjordur in Iceland (Marteinsson et al., 2001). These investigations showed that there were frequently a high abundance of autotrophs within the classes Gammaproteobacteria and Campylobacteria (previously termed Epsilonproteobacteria) (Waite et al., 2017) in the shallow-sea systems.

The submarine hydrothermal systems harbor chemolithoautotrophic bacteria and archaea communities typically associated with locally reduced gasses sulfide (H2S), methane (CH4), and hydrogen (H2) (Grzymski et al., 2008; McDermott et al., 2015; Brazelton and Baross, 2010), which are considered to support primary productivity through carbon dioxide (CO2) fixation. The sulfur-reducing chemolithoautotrophs Nautiliales-like organisms within Campylobacteria and sulfide-oxidizing chemolithoautotrophs Thiomicrospira-like organisms within Gammaproteobacteria dominated and exhibited distinct zonation within the water columns of the shallow hydrothermal system off Kueishantao Island, Taiwan (Zhang et al., 2012; Tang et al., 2013). Chemolithoautotrophs Nautiliales-like and Thiomicrospiralike organisms are frequently found in the other shallow-sea systems (Hirayama et al., 2007), and were found in some deep-sea hydrothermal system as well (Scott et al., 2006; Campbell et al., 2009; Brazelton and Baross, 2010; Yamamoto and Takai, 2011). They might possess the reductive tricarboxylic acid cycle (rTCA) and the Calvin-Benson-Bassham (CBB) cycle for carbon fixation in response to the available energy source in the form of the oxidation of reduced sulfur compounds and hydrogen in the environment, which is possibly fueled by geochemical energy with hydrogen and reduced sulfur, respectively (Tang et al., 2013). Chemolithoautotrophic organisms were also active within the steep geochemical gradients of the shallow-sea hydrothermal sediments, which were possibly involved in sulfide oxidation and sulfate reduction (Bayraktarov et al., 2013; Giovannelli et al., 2013). The sediments often harbored abundant Campylobacteria (such as genus Sulfurovum that encompasses sulfur- and thiosulfate-oxidizers bacteria) similar to those seen in deep-sea vents, but also other proteobacterial lineages that are distinct from those of deep-sea vents (Giovannelli et al., 2013). These studies provided some indication of a potential biogeochemical function for chemolithoautotrophic organisms in the shallow-sea hydrothermal system with significant ramifications for sulfur and carbon cycles (Zhang et al., 2012; Bayraktarov et al., 2013; Giovannelli et al., 2013; Tang et al., 2013).

Inorganic carbon is the primary carbon source assimilated by autotrophic bacteria in submarine hydrothermal systems, but hydrothermal fluids can also carry elevated concentrations of dissolved and particulate organic matter (Bennett et al., 2011; Yang et al., 2012). In addition, fluids may cool down and change physicochemical fluctuating conditions by mixing rapidly with seawater (Yang et al., 2012). The enrichment of nutrients and the temperature and dissolved oxygen gradients in the mixed hydrothermal fluids and seawater might support active heterotrophic microbes. Recent studies have shown that deep-sea hydrothermal systems are inhabited by versatile heterotrophic Alphaproteobacteria and Gammaproteobacteria, which are significantly distinct from heterotrophic lineages common in the deep-sea environment, some of which have the potential for alkane degradation (Meier et al., 2016). A bacterium closely related to a human pathogenic Vibrio species was isolated from surrounding sulfide chimneys of a hydrothermal vent along the East Pacific Rise, and its genomic information provided new insights on how species adapt to the deep-sea environment (Hasan et al., 2015). Several studies have shown the presence of physiologically, metabolically, and phylogenetically diverse heterotrophic communities in shallow-sea hydrothermal systems (Giovannelli et al., 2013; Meyer-Dombard et al., 2013; Price et al., 2013), in which some seemed mostly to be involved in arsenic and iron redox cycling (Meyer-Dombard et al., 2013; Price et al., 2013). However, in marine hydrothermal systems,

compared to chemolithoautotrophic organisms, heterotrophic bacteria distribution, diversity, and metabolic capacity have still been poorly investigated, and isolates with reference genomes from the shallow-sea hydrothermal system are scarce.

In the present work, we revisited the shallow-sea hydrothermal system near Kueishantao Island and collected samples from the water column above the yellow and white vents, in addition to samples of sandy sediments, rocks, and dead vent deposits near the vent sites. Microbial compositions were characterized using culture-independent methods including fluorescence in situ hybridization (FISH) and 16S rRNA gene amplicon sequencing. Metagenomic analysis was used to determine the functional and genetic potential of bacteria. The various heterotrophic bacteria with potentially different trophic strategies were isolated from the investigated shallow-sea hydrothermal systems. Some of these isolates may be previously undiscovered bacterial species. Further studies aimed to characterize the physiological and genomic features of bacterial strain P5, a bacterioplankton species with a considerable capacity for adaptation to the shallow-sea hydrothermal environment.

#### MATERIALS AND METHODS

#### Sample Collection

Sampling was performed by scuba divers in May 2015 in two different regions of the shallow-sea hydrothermal system located near Kueishantao Island (N 24.834, E 121.962), Taiwan, China: the white vent area and the yellow vent area (**Figure 1**). All necessary permits were obtained for the described field studies. The vents were observed at depths of approximately 10 and 7 m, respectively. Four samples were collected from the following locations (water samples): W\_0m, W\_5m, W\_surface, and W\_outside, which are 0 m above the white vent, 5 m above the white vent, the surface water directly above the vent, and surface water 6 m laterally away from the white vent, respectively. Four more samples were collected from the following locations (water samples): Y\_0m, Y\_5m, Y\_surface, and Y\_outside, which are 0 m above the yellow vent, 5 m above the yellow vent, the surface water directly above the yellow vent, and the surface water 6 m laterally away from the yellow vent. The sandy and rocky sediments nearby the yellow vent were named YS\_S and YS\_R, respectively. The sandy and rocky sediments on the seafloor were named S\_S and S\_R. The rocky sediment collected nearby the dead vent was named DS\_R (an expired vent without the plume and gas discharging). The 2-l water samples were filtered through 3 and 0.2 µm pore-size polycarbonate filters (EMD Millipore Corp., Darmstadt, Germany) for further analysis. Sediment samples were frozen on site and kept frozen during transportation and storage.

#### Abundance of Bacteria Determined by CARD-FISH

Fifty-milliliter water samples were immediately fixed with freshly prepared paraformaldehyde (4% final concentration) and stored at 4◦C overnight prior to filtration through 0.2 µm pore size polycarbonate filters (Millipore, United States). Filtered samples were stored at −20◦C for later analysis by FISH with horseradish peroxidase-labeled oligonucleotide probes (CARD-FISH). Picoplankton abundance was determined by DAPI staining, and bacteria were enumerated by CARD-FISH. Filters were embedded in low-gelling-point agarose and incubated with either lysozyme for the bacteria probe mix (Eub338: 5<sup>0</sup> -GCT GCC TCC CGT AGG AGT-3<sup>0</sup> , Eub338II: 5<sup>0</sup> -GCA GCC ACC CGT AGG TGT-3<sup>0</sup> or Eub338III: 5<sup>0</sup> -GCT GCC ACC CGT AGG TGT-3<sup>0</sup> ) or for the negative control probe (Non338: 5<sup>0</sup> -ACTCCT ACG GGA GGC AGC-3<sup>0</sup> ) (Teira et al., 2004). Filters were cut into four pieces and hybridized with HRP-labeled oligonucleotide probes and tyramide Alexa488 for signal amplification following the previously described protocol (Teira et al., 2004).

#### DNA Extraction

Total DNA of water samples was extracted from the samples with the FastDNA SPIN Kit for Soil (MP Biomedicals, Illkirch, France) according to the manufacturer's instructions. DNA from the sediment samples was isolated using PowerMax Soil DNA Isolation Kits (MoBio Laboratories Inc., United States) according the manufacturer's instructions. The concentration and purity of DNA were evaluated using a NanoDrop spectrophotometer (ND-1000, Thermo Fisher Scientific, Waltham, MA, United States). DNA extracts were stored at −80◦C until further analysis. The following sequencing was performed at the Chinese National Human Genome Center in Shanghai.

## Amplification and Sequencing of the Bacterial 16S rRNA Gene

Amplicons were generated using fusion degenerate primers 343F (Wilson et al., 1990) and 798R (Rochelle et al., 1995) with ligated overhang Illumina adapter consensus sequences. The libraries were prepared in accordance with the instructions included with the Illumina Nextera XT Index kit (Illumina, United States). Pooled amplicons were purified with the Agencourt AMPure XP purification system (Beckman, United States) and analyzed with an Agilent bioanalyzer 2100 (Agilent Technologies, United States) to confirm appropriate amplicon size. Finally, amplicons were diluted to 10 nM, quantified and sequenced on the Illumina MiSeq platform (reagent kit v.3; Illumina, United States).

#### Metagenomes Sequencing

For metagenome sequencing, 1 µg of sample DNA was sheared to 500 bp by Covaris M220 (Covaris, United States). The library was constructed by NEBNext UltraTM DNA Library Prep Kit (NEB, United States). Finally, 10 nM sequencing library was used generate cluster in cBot using TruSeq PE Cluster Kit (Illumina, United States), and sequenced by Illumina Hiseq 2500 for 2 × 250 bp data.

## 16S rRNA Gene Sequence Data Analysis

Acquired Illumina reads were filtered by Meta Genome Rapid Annotation using Subsystem Technology (MG-RAST<sup>1</sup> ) QC

<sup>1</sup>http://metagenomics.anl.gov/

pipelines to remove the replicated reads (Meyer et al., 2008). The filtered reads were used for the following bioinformatic analysis. For taxonomic analysis, the SILVA small subunit (SSU) database implemented in MG-RAST was used as annotation source for 16S rRNA reads to analyze the bacterial populations in samples using an E-value cutoff of 1e-05, minimum identity cutoff of 60%, and minimum alignment length cutoff of 150 bp.

#### Shotgun Metagenomic Sequence Data Analysis

Functional profiles were identified using the SEED subsystems annotation source of the MG-RAST, with 1e-05 as maximum e-value, a minimum identity of 60%, and a minimum alignment length of 15 amino acids. To remove the bias of average genome size on the sampling of genes from a given metagenomic community, the raw functional gene hits were normalized to the number of recA gene encoding recombinase A hits in the respective database. The PRIMER-6 package was used to calculate the Bray–Curtis similarity matrices of metagenomes and generate non-metric multidimensional scaling plots. PERMANOVA implanted in PAST v3.05 was carried out to compare samples from each environment (Hammer et al., 2001). To determine whether the relative abundances of functional genes differed significantly between sample categories, we conducted multiple t-tests with P-values calculated using a Holm–Sidak correction (Shaffer, 1995) for multiple comparisons implemented in Prism v6.

## Strain Isolation and Culture

All reagents used in bacterial cultures were obtained from Sigma–Aldrich (United States) unless otherwise specified. For the cultivation of bacteria, 200 µL of the water sample was spread onto at least one of the agar plates. The widely used medium for routine culture of marine bacteria were selected (Joint et al., 2010), including low-nutrient R2A (Difco, United States), and nutrient-rich 2216E (Becton-Dickinson, United States), RO (1 g peptone, 1 g yeast extract, 1 g natrium aceticum, 1 g sodium acetate per liter artificial seawater (ASW) with vitamins and trace elements, pH 7.8–8.0), SYPG (30 g NaCl, 0.5 g yeast extract, 0.25 g tryptic peptone, 0.1 g glutamic acid monosodium salt per liter, pH 7.5), SC (1 g D-glucose, 1 g alkapolpeg-600, 1 g L-malic acid, 1 g D-aspartic acid, 1 g yeast extract per liter ASW with vitamins and trace elements, pH7.2–7.5), NS (0.5 g sodium nitrate, 0.5 g sodium sulfate, 0.1 g yeast extract per liter artificial seawater with vitamins and trace elements, pH 7.2–7.5), YTSS (2.0 g yeast extract, 1.25 g tryptone, 20 g sea salt (Sigma) per liter distilled water and autoclave, pH 7.0). All agar plates were incubated at 28◦C or under anaerobic conditions at 24◦C. Bacterial colonies were picked from the plates and purified further on 2216E plates. The bacterial 16S rRNA gene sequencing and sequence analysis were performed using a previous method (Embley, 1991). The sequence data have been submitted to the NCBI database under accession numbers AM988866 through AM989325. Genomic DNA was extracted from 200 µL of culture using a TIANamp Bacteria DNA Kit (Tiangen, China). The primer pair pufLF (5<sup>0</sup> -CTKTTCGACTTCTGGGTSGG-3<sup>0</sup> ) and pufMR (5<sup>0</sup> -CCATSGTCCAGCGCCAGAA-3<sup>0</sup> ) were designed to amplify pufL and pufM through PCR to allow detection of the photosynthetic reaction center genes in the strains (Cho et al., 2007).

## Genome Sequencing and Analysis

Whole genome sequencing of strain P5 was accomplished using a hybrid approach, combining Illumina short read data with PacBio long read data (Koren et al., 2012). For PacBio sequencing, 5 µg of sample DNA was sheared to 10 Kb by a Covaris <sup>R</sup> g-TUBE <sup>R</sup> (Covaris, United States). A PacBio <sup>R</sup> SMRTbellTM Template Prep Kit (PacBio, United States) was used to construct the library. The sequencing primers were annealed using a PacBio DNA/Polymerase Kit (PacBio, United States) and polymerase combined with the SMRTbell templates. We obtained long read data from PacBio RS II PacBio RS II (PacBio, United States). The raw Illumina data were filtered by the FASTX-Toolkit

to remove the adapters, N bases, and low-quality reads. The clean data were assembled using Velvet v1.2.03 with default parameters. The PacBio long reads were assembled by RS HGAP assembly 3. The complete genome was finally gap closed by Sanger sequencing. The final assembled genomes of P5 were automatically annotated and analyzed through the IMG/ER<sup>2</sup> . The comparison and visualization of multiple genomes was conducted with BRIG (Alikhan et al., 2011).

## Physiological and Biochemical Analysis of Strain P5

Strain P5 was isolated from surface seawater off Kueishantao Island, northeast Taiwan by 2216E medium. The strain was adaptively grown for 3 days (inoculated into a new bottle everyday) in ASW containing a 1 mM concentration of dissolved organic carbon (DOC). The DOC concentration was adjusted by ASW supplemented with full strength medium (5 g peptone and 1 g yeast extract per liter ASW and the final concentration was determined using a total organic carbon analyzer (Shimadzu, Japan). Bacteria (10<sup>5</sup> cells/mL) were then washed three times with autoclaved ASW and used to inoculate growth medium consisting of artificial seawater (ASW) base combined with substrates including 0.45 µM DOC, 2.5 mM NaHCO3, and 1 mM Na2S2O<sup>3</sup> in phosphate-buffered saline (pH 7.4) (autotrophic culture: NaHCO<sup>3</sup> and Na2S2O3; heterotrophic culture: DOC). Cultures were incubated at 22◦C at 160 rpm/min in the darkness or light (12 h)/dark (12 h) cycle for 4 days. To determine bacterial cell density, cultures were stained with SYBR Green I (1:100 dilution; Molecular Probes, United States) for 15 min, and measured via flow cytometry (BD Accuri C6, United States). All culture experiments were performed in triplicate.

The strain P5 was cultured in 2216E and freeze-dried by a Freeze Dry System (Labconco Corp., Czechia) for 48 h, a moderate amount of chloroform was added, and cells were broken under a Ultrasonicator SM-650D (Shunma, China). The Ultrasonicator was operated for a total of 10 min containing 9 s for running, and 5 s for stopping. The power was 22% (total power 650 W). The supernatant was measured using a UV-Vis spectrophotometer (Agilent 8453, United States) after being centrifuged at 7006 g/min for 10 min. Bacteriochlorophyll a was measured with a FIRe Fluorometer Induction and Relaxation System (Satlantic, Canada) at 800 nm.

#### Nucleotide Sequence Accession Numbers

The metagenomic datasets are publicly available in the MG-RAST system under project identifiers W\_0m (4671689.3 and 4668298.3), W\_5m (4671690.3 and 4668304.3), W\_surface (4671687.3 and 4668303.3), W\_outside (4671688.3 and 4668305.3), Y\_0m (4671691.3 and 4668302.3 for 16S rDNA and functional metagenome, respectively), Y\_5m (4671692.3 and 4668301.3), Y\_surface (4671693.3 and 4668300.3), Y\_outside (4671694.3 and 4668299.3), YS\_S (4647123.3 and 4644370.3), YS\_R (4647124.3 and 4644369.3), S\_S (4647121.3 and 4644371.3), S\_R (4647122.3 and 4644368.3), and DS\_R (4647120.3 and 4644372.3). They have also been deposited in the NCBI Short Read Archive: 16S rDNA reads, SRR5229862- SRR5229869 for W\_0m, W\_5m, W\_surface, W\_outside, Y\_0m, Y\_5m, Y\_surface, and Y\_outside, respectively, YS\_S (SRR5149598), YS\_R (SRR5149595), S\_S (SRR5149599), S\_R (SRR5149606), DS\_R (SRR5149594), metagenomic DNA reads, SRR5229878-SRR5229885 for W\_0m, W\_5m, W\_surface, W\_outside, Y\_0m, Y\_5m, Y\_surface, and Y\_outside, respectively, YS\_S (SRR5149596), YS\_R (SRR5149597), S\_S (SRR5149607), S\_R (SRR5149602), DS\_R (SRR5149604). The complete genome sequence of strain P5 has been deposited into GenBank under the accession numbers CP015039 (chromosome) and CP015040-CP015043 (plasmids).

## RESULTS AND DISCUSSION

#### Environmental Parameters

The fluids in the shallow-sea hydrothermal system were slightly acidic (approximately pH 6) and oxic (Supplementary Table S1). Salinity, dissolved inorganic carbon, nitrate, nitrite, and phosphate concentrations were nearly oceanic (Supplementary Table S1). Elemental sulfur (S<sup>0</sup> ) is naturally enriched in the shallow-sea hydrothermal fluids near Kueishantao Island. Compared to deep-sea vents, the sulfide concentration was lower in this shallow vent system (Chen et al., 2005; Chan et al., 2016). Compared to the common marine environment, the concentartion of DIC was higher, while the DOC centration was similar in the shallow-sea hydrothermal system (Supplementary Table S1). These physico-chemical properties of the hydrothermal system might support not only autotrophic bacteria, but also heterotrophic bacteria.

#### Cell Densities

The DAPI-based total cell counts were on average on the order of 10<sup>6</sup> cells mL−<sup>1</sup> (Supplementary Table S2). The general bacterial probe mixture EUB338 hybridized 90–94% of all DAPI-stained cells in the white vent area, while the percentage of bacterial probe in the yellow vent area was 57–87% (Supplementary Table S2). Compared to the white vent (temperature up to 58◦C inside the vent), the relatively high temperature (up to 116◦C) near the yellow vent might result in lower bacterial abundance and their proportions of total microbial cells at the Y\_0m site (0 m above the yellow vent) (only 57%).

#### Bacterial Community Composition

The investigation of the microbial communities using 16S rRNA gene amplicon sequencing revealed that a significant proportion of the sequences in the water samples (14–89%) and sediment samples (21–47%) could either not be unassigned or were assigned to unclassified groups, indicating the presence of so far uncharacterized bacteria within this geothermal ecosystem. Campylobacteria were apparently present in relatively high abundance in the sediment samples, and represented the main bacterial groups at the Y\_0m site and the W\_0m 318 site (water samples, 0 m above the vents) (**Figure 2**). However, the

<sup>2</sup>http://img.jgi.doe.gov

relative abundance of 16S rRNA sequences of Campylobacteria decreased in the water column above the vents (**Figure 2**). Few sequences of Campylobacteria were classified into those from sulfur-reducing chemolithoautotrophs Nautiliales-like organisms that were previously known to be a dominant group in the white vent (Zhang et al., 2012; Tang et al., 2013). The shallowsea hydrothermal vents varied greatly in chemical composition of gasses discharging such as CO<sup>2</sup> and H2S (Chen et al., 2016). In the previous one, the DIC concentrations was about 6 mM in 0 m above the white vent (Zhang et al., 2012), however, there was only a little higher (∼2.7 mM) than that at nearby seawater in this study. On the other hand, the H2S concentrations have declined since 2014 (Chen et al., 2016). In addition, physical conditions such as tide could influence the hydrothermal ecosystem (Chen et al., 2005). These might influence on the microbial community composition. Gammaproteobacteria were found abundantly at the surface waters and outside the white vent (W\_surface site and W\_outside site) (**Figure 2**). However, the very few sequences of sulfuroxidizing Gammaproteobacteria Thiomicrospira-like organisms were detected. Members of the family Alteromonadaceae and Oceanospirillaceae within Gammaproteobacteria are the most abundant organisms in W\_surface and W\_outside, respectively. Y\_5m (5 m above the yellow vent) and Y\_surface (the surface water immediately above the yellow vent) have abundant members of Alphaproteobacteria (**Figure 2**). Flavobacteriia were present at all sites (**Figure 2**). There are some relatively rare taxa found in the hydrothermal system including Deltaproteobacteria, Beatproteobacteria, Actinobacteria, Cyanobacteria, Aquificae, Chorobi, Spirochaetes, and Tenericutes. Deltaproteobacteria, and Firmicutes are mainly found in sediments samples (**Figure 2**).

Overall, homology-based taxonomic assignments of shotgun metagenomics data predicted by the MG-RAST showed that Gammaproteobacteria and Alphaproteobacteria were abundant classes in the white and yellow vent water samples, making up approximately 60% of the microbial assemblage, whereas Campylobacteria were abundant class in the vent sites and sediment samples. The most dominant proteobacterial orders overall were Rhodobacterales and Rhizobiales within the class Alphaproteobacteria; Alteromonadales, Pseudomonadales, and Oceanospirillales within the class Gammaproteobacteria. However, many shotgun metagenomic sequences (mostly > 50%

of the total sequences) still remained functionally unannotated and taxonomically unassigned due to the limitations of the available tools and the paucity of reference genomes. These could possibly lead to a discrepancy between shotgun metagenomics data taxonomic assignment and 16S rRNA-based classification.

## Metabolic Potentials and Functional Biomarkers of Microbial Communities

Based on the relative abundance of SEED subsystems, multidimensional scaling plots show that the hydrothermal water samples in this study clustered together at the functional level, and were separate from the hydrothermal sediments samples (**Figure 3** and Supplementary Table S3); permutational multivariate analysis of variance (PERMANOVA) indicates that this difference is significant (P = 0.009). We observed a set of SEED functions "Carbohydrates" and "Amino Acids and Derivatives" that contributed strongly to the difference between the communities in the sediments and water samples; that set of SEED functions is more abundant in the hydrothermal water samples. Compared to a metagenome from the previous investigation in the white vent area, "Carbohydrates" were more abundant in the metagenomic datasets in this study, but there were fewer "Protein metabolism" results. Compared to the bacterial community compositions from the previous investigation, Flavobacteriia were more abundant in this study. The Flavobacteriia genomes contain diverse and abundant glycoside hydrolases (GHs) that are involved in carbohydrate metabolisms (Tang et al., 2017). Two categories, "Carbohydrates" and "Protein metabolism," together contributed to approximately 26% of the differences between communities in the two metagenomes. However, "Sulfur metabolism" and "Photosynthesis" contribute less to their differences. The hydrothermal water samples were functionally distinct from those in the costal and open seawater samples (PERMANOVA, P < 0.01 in all cases). The genes associated with the cell wall and capsule category, and virulence, disease, and defense category were more abundant in the hydrothermal water samples than those in the costal and open seawater samples metagenomes, which are main contributors to the dissimilarity between metagenomes (Supplementary Table S3).

The functional metagenomes suggested that oxidation of reduced sulfur compounds in the hydrothermal system could occur through the Sox multienzyme system (**Figure 4**), which catalyzes the complete oxidation of reduced sulfur compounds to sulfate (Yamamoto and Takai, 2011). Several enzymes have been proposed to as possible oxidizers of inorganic sulfur compounds, including sulfide quinone oxidoreductase and sulfite oxidase, which oxidize sulfide to elemental sulfur and sulfite to sulfate, respectively (**Figure 4**) (Yamamoto and Takai, 2011). Genes encoding polysulfide reductase (Psr) are present in the sediments metagenome, resulting in the reduction of polysulfide derived from elemental sulfur to sulfide (Yamamoto and Takai, 2011). Previous metagenome analyses have suggested that Campylobacteria in the vent can gain energy from sulfurreduction catalyzed by Psr to fix CO<sup>2</sup> by the rTCA cycle (Tang et al., 2013). However, the relative gene abundance of Psr

in the water samples was lower than those in the sediments (**Figure 4**). The Deltaproteobacteria contributed to Psr gene sequences in the sediment metagenomes. In addition, few key genes encoding ATP-dependent citrate lyase for the rTCA cycle (Hügler and Sievert, 2011) were found in any of the metagenome datasets in this study, whereas genes encoding ribulose-1,5-bisphosphate carboxylase (RuBisCO) were found; those genes mediated the CBB cycle for carbon fixation. The relative abundances of RuBisCO (normalized to recA genes) in the hydrothermal water samples, on average, were 0.12 and 0.14 in the yellow vent and white vent areas, respectively, which were both lower than found (0.98) in a previous investigation (Tang et al., 2013). In contrast, heterotrophic metabolism might be predominant both in the water column and the sediment, as indicated by a set of transporters, peptidases, and GHs genes in all the metagenomic datasets (**Figure 4**). Although the relative abundances of genes encoding GH and peptidase between the sediment samples and water samples display no statistical difference (t-test, P > 0.01), they are more diverse in the waters than in the sediment of hydrothermal system, indicating a wider spectrum of substrate utilization for bacteria in the waters of this shallow-sea system. Similarly, more ATP-binding cassette systems (ABC transporters), symporters, and TonB-dependent receptors (TBDTs) were found in the hydrothermal waters, which allowed them to import organic matter efficiently (**Figure 4**). All metagenomes were found to possess exporters involved in the efflux of heavy metals and metabolites (Supplementary Table S4). Predicted substrates for the transport systems in the metagenomes include a variety of carbohydrates, carboxylic acids, amino acids, peptides, metals, and other nutrients (Supplementary Table S4).

All metagenomes were found to contain genes encoding sulfate adenylyltransferase, adenylsulfate kinase, and adenylsulfate reductase, which are required for assimilatory sulfate reduction to supply sulfur for biosynthesis in aerobic marine bacteria (**Figure 4**). Functional metagenomic analyses indicated that the relative abundance of genes encoding Psr, Ni-Fe hydrogenase, and periplasmic nitrate reductase exhibited statistically significant differences between hydrothermal sediments and water samples (t-test, P < 0.05) and it was more abundant in sediment samples. The genes encoding Ni–Fe hydrogenase in the metagenome enabled bacteria to use H<sup>2</sup> as an energy source (Petersen et al., 2011). There is some possibility that nitrate could be used as alternative electron acceptor in the presence of periplasmic nitrate reductase (Canfield et al., 2010). Thus, chemolithotrophs might contribute to the chemical transformations of elements in the sediment.

Overall, the Rubisco gene sequences were mainly homologous to those in the order Rhodobacterales, Rhizobiales, Methylococcales, and Thiotrichales, in contrast to previous metagenomes where most of the Rubisco gene sequences homologs were affiliated with Thiomicrospira-like (Tang et al., 2013). Cyanobacteria also contribute to Rubisco gene pool at the sea surface of the hydrothermal system. The order Desulfovibrionales within Deltaproteobacteria in the vent sites and sediment samples contributed to genes encoding NiFe hydrogenase. The majority of genes encoding TBDTs were closely

FIGURE 4 | Heat map of functional composition among the samples. The value of functional gene relative to recA (single copy control gene) is assigned with a color relative to the maximum value among all comparisons of each gene. The colors represent the minimum (blue), middle (white, 0.5), and maximum (red) values listed on the right. The number in the box represented the components of each functional category.

related to Flavobacteriia and Gammaproteobacteria (primarily members of the Alteromonadales and the Pseudomonadales) in the water samples, while TBDTs sequences from the Campylobacteria and Deltaproteobacteria class were found in sediment samples. The members of the order Rhodobacterale contributed greatly to the diversity of ABC transporters in the metagenomes, and the members of Flavobacteriia contribute greatly to the diversity of peptidases and glycoside hydrolases in the metagenomes. The sox sequences in the metagenomic data were mainly homologous to those found in the orders Rhodobacterale and the class Campylobacteria. Genes encoding periplasmic nitrate reductase and dissimilatory nitrite reductase were more prevalent in the sediment dataset than in the water sample dataset, and they were mainly homologous to sequences from the members of Campylobacteria and Deltaproteobacteria.

#### Taxonomic Assignment of the Isolated Strains

A total of 408 isolates were identified by near full-length 16S rRNA gene sequence analysis (Supplementary Table S5). Of these isolates, 187 were from the yellow-vent area, 111 were from the white-vent area, and 65 were from the surface water above the vents. It was noted that 42 isolates were obtained from the waters above the dead vent. A previous study showed that heterotrophic bacteria dominated inactive deep-sea hydrothermal system (Sylvan et al., 2012). The isolates were distributed into seven different bacterial classes from four different phyla, namely Proteobacteria, Bacteroidetes, Actinobacteria, and Firmicutes. However, the failure to isolate Epsilonprobacteria strains may be because the optimal conditions for their culture have yet to be established. The majority of the characterized strains belonged to the Alphaproteobacteria (174 strains), Gammaproteobacteria (145 strains), and Actinobacteria (46 strains). The rest of

the strains were assigned to Betaproteobacteria, Flavobacteriia, Cytophaga, and Bacillus. The spectrum of different genera was greatest within the classes Alphaproteobacteria (31 genera), followed by Gammaproteobacteria (15 genera), Flavobacteriia (14 genera), and Actinobacteria (12 genera) (**Figure 5**). For the classes Alphaproteobacteria, Erythrobacter (81 strains) and Paracoccus (31 strains) were the predominant genus (**Figure 5**). For the classes Gammaproteobacteria, Vibrio (41 strains) and Pseudoalteromonas (36 strains) constituted a high proportion (**Figure 5**). For the classes Actinobacteria, Microbacterium (16 strains) was dominant. These taxonomic groups are commonly detected in ocean environments (Haggerty and Dinsdale, 2017).

These isolates may possess novel functions and exhibit a broad range of ecological attributes and life-history strategies. For example, some strains belonged to oligotrophic taxa (including members of Erythrobacteraceae and Sphingomonadaceae groups); some strains belonged to the copiotrophic taxa (including members of Vibrio and Alteromonadaceae phyla) and may be able to rapidly consume labile DOC (Church, 2009; Kirchman, 2015). The members of Flavobacteriaceae have the ability to degrade complex high-weight molecule organic compounds (Buchan et al., 2014). Those in the genera Erythrobacter and Novosphingobium can to metabolize nutrient-poor and recalcitrant carbon substrates (Church, 2009; Kirchman, 2015). Furthermore, pufL and pufM genes, which encode photoreaction center L and M polypeptides, respectively, in Erythrobacter were identified as aerobic anoxygenic phototrophic bacterial gene biomarkers. Previous studies have suggested that they have the capacity to undergo photoheterotrophy in marine environments (Jiao et al., 2007).

represent bootstrap values (based on 100 resamplings). Bootstrap values more than 50% are noted. The GenBank accession numbers for 16S rRNA gene

On the basis of the taxonomic assignment by the EzTaxon classifier using annotated 16S rRNA gene sequences (Chun et al., 2007), a total of 58 strains from the classes Alphaproteobacteria, Gammaproteobacteria, Flavobacteriia and Actinobacteria might represent new bacterial species not yet validly described with less than 98% maximum identity with their closest BLAST hits (Supplementary Table S5). Among them, the information on the first complete genome of a cultivated actinomycete strain JLT9 isolated from the shallow-sea hydrothermal system was reported, which contained various sulfur oxidation genes (**Figure 6A**) (Han et al., 2017). The complete genome of the Maribacter sp. T28 harbored the xylanolytic, alginolytic and pectinolytic enzymes responsible for polysaccharide degradation (**Figure 6B**) (Genbank accession, CP018760) (Zhan et al., 2017). The genome data suggested that Marivivens sp. JLT3646 within Alphaproteobacteria has the potential to degrade aromatic monomers (**Figure 6C**) (Genbank accession, CP018572 and CP018573 for chromosome and plasmid, respectively) (Chen et al., 2017).

sequences are shown in parentheses. Bar, 1 nt substitution per 100 nt.

The gene sequences homologs affiliated with Rhodobacterales were abundant, accounting up to 20.57% of the total sequences in the water sample. Strain P5 (**Figure 6D**) exhibited a 97.48% 16S rRNA sequence similarity with Rhodovulum adriaticum DSM 2781 within the order Rhodobacterales (**Figure 7**). Total of eight strains of Rhodovulum were isolated form the white and yellow vents area (Supplementary Table S5). The members of Rhodovulum were also frequently found near the shallow-sea submarine vents of Panarea Island (Maugeri et al., 2013a). The newly discovered strain P5 of great metabolic versatility could be considered to be representative of heterotrophic bacteria isolate from the shallow-sea hydrothermal system.

## Biochemical and Physiological Characteristics of Strain P5

Strain P5 is a Gram-negative, spindle-shaped, purple bacterium (1.5–2.0 µm in length and 0.9–1.0 µm in width) (**Figure 6D**) that can grow at the temperature range of 20–40◦C (optimum, 26–34◦C), in the pH range of 5–9 (optimum, 6–8), in the presence of 0–6.0% (w/v) NaCl (optimum, 3.0%), and in

aerobic or microanaerobic conditions (Supplementary Table S6). Electron microscopy of ultrathin sections revealed the presence of a vesicular type internal photosynthetic membrane that is a common feature in other Rhodovulum species (**Figure 8A**) (Masuda et al., 1999; Srinivas et al., 2007; Kompantseva et al., 2010; Maugeri et al., 2013a). The in vivo absorption spectrum of intact P5 cells exhibited four major peaks at 508, 588, and 804 and 850 nm, thus confirming the presence of bacteriochlorophyll a and carotenoids (**Figure 8B**). The functionality of the photosynthetic apparatus was investigated by infrared kinetic fluorescence measurements. P5 displayed clear induction of bacteriochlorophyll a with FV/F<sup>M</sup> 0.749 ± 0.003 (mean ± SD; n = 3), which confirmed that its fully functional photosynthetic reaction centers are connected to an efficient electron transfer chain (**Figure 8C**). These results suggested that P5 had phototrophy capability similar to other Rhodovulum species (Masuda et al., 1999; Srinivas et al., 2007; Kompantseva et al., 2010; Maugeri et al., 2013a). Strain P5 is capable of heterotrophic, autotrophic and photoheterotrophic growth (**Figure 8D**).

#### Genomic Features of Strain P5

Strain P5 exhibited a genome of 4,137,334 bp (one chromosome and four plasmids) with a G+C content of 64.64% that coded 4,243 protein-coding and 68 RNA genes (including nine rRNA operons). Bidirectional BLAST analyses showed that approximately 3,000 genes exhibited > 50% sequence identity between strain P5 and two sequenced genomes of Rhodovulum species [Rhodovulum sulfidophilum DSM 1374 (Nagao et al., 2015b) and Rhodovulum sulfidophilum DSM 2351 (Nagao et al., 2015a)] (**Figure 9**), indicating identical or equivalent function.

Strain P5 was predicted to possess complete central carbon metabolic pathways, including glycolysis, the pentose phosphate pathway, the Entner–Doudoroff pathway, and the tricarboxylic acid cycle. The P5 genome harbored genes encoding 16 GHs and 30 peptidases as well as a urease gene cluster, indicating the potential for the degradation of carbohydrates, proteins, peptides, and urea (Supplementary Table S7). The most abundant transporter systems in the P5 genome were ABC transporter components, followed by TRAP-type transport components, and the exporters and antiporters, accounting for 8.1% of the total protein-coding genes of the chromosome in P5. Predicted uptaking substrates for the complete transporter systems comprised a variety of carbohydrates, carboxylic acids, amino acids, peptides, alkanesulfonate, metals, and other nutrients. The genome harbored two genes sets encoding type IV secretion machinery systems that are used to deliver proteins and DNA into the extracellular environment, whereas the other two Rhodovulum species lack these genes (**Figure 9**). This organism has genes for the complete CBB required for autotrophic carbon fixation. Well-known electron donors utilized for microbial autotrophic growth are hydrogen and reduced forms of sulfur (sulfide, S<sup>0</sup> , and thiosulfate). Strain P5

possessed the complete repertoire of genes for the oxidation of reduced sulfur compounds, which increased niche-specialization competitiveness in the shallow-sea environments. These genes encoded enzymes for the oxidation of reduced sulfur compounds including the Sox enzyme complex for oxidation of reduced sulfur to sulfate (SO<sup>4</sup> <sup>2</sup>−), sulfide quinone oxidoreductase, mediating the oxidation of sulfide to elemental sulfur, rhodanese sulfurtranferase for oxidation of thiosulfate (S2O<sup>3</sup> <sup>2</sup>−) to sulfite (SO<sup>3</sup> <sup>2</sup>−), and reverse dissimilatory sulfite reductase for oxidation of elemental sulfur to sulfite, adenosine 5-phosphosulfate reductase and sulfate adenylyltransferase for oxidation of sulfite to sulfate (SO<sup>4</sup> <sup>2</sup>−). Strain P5 has two gene clusters encoding different Ni–Fe hydrogenases, which is predicted to catalyze the reversible oxidation of hydrogen gas and enables bacteria to use molecular hydrogen as a source of energy. The heterodisulfide reductase-related proteins (Hdr) are only identified in the P5 genome (**Figure 9**), and are likely candidates to be involved in energy coupling through electron bifurcation from diverse electron donors such as formate or H<sup>2</sup> via formate dehydrogenase or Hdr-associated hydrogenase (Kaster et al., 2011). Similar to other Rhodovulum species, the genome of strain P5 also contains all of the genes required for photosynthesis gene clusters (PGCs), including those involved in biosynthesis of bacteriochlorophyll a, carotenoids, light-harvesting systems, and reaction center components (Supplementary Table S7).

The genomes of P5 contain one large gene set that encodes a flagellum system and genes that encode methyl-accepting chemotaxis proteins, indicating that P5 may use them to facilitate their movement toward nutrient rich zones. Genes encoding gas vesicle proteins have been identified, providing buoyancy to cells as flotation devices in order to obtain the optimum amount of light and nutrients at a suitable depth in the environment. Several chemotaxis proteins were located near the gas vesicle proteins (gvp) gene cluster. In contrast, two Rhodovulum species lack the gvp gene cluster (**Figure 9**).

Strain P5 carries several prophage-like elements, in which a Mu-like head group phage has been induced successfully (Lin et al., 2016). The presence of clustered regularly interspaced palindromic repeat (CRISPR) arrays and their associated Cas genes in P5 and two other Rhodovulum species form a system

that is possibly involved in bacterial defense against phages or plasmids. However, CRISPR sequences revealed no similarity between strains with regards to the numbers of repeats and spacer sequences, indicating that the histories of phage infection are different. Both bacteria have a predicted gene transfer agent (Supplementary Table S7).

With respect to morphological and biochemical traits as well as phylogenetic relationships and genomic analysis, isolate P5 is clearly a strain belonging to the genus Rhodovulum.

#### Summary

In this study, few 16S rRNA gene sequences were related to Thiomicrospira- and Nautiliales-like chemolithoautotrophic bacteria, both of which were previously reported to be abundant in the investigated Kueishantao shallow-sea hydrothermal system (Zhang et al., 2012; Tang et al., 2013). In contrast, heterotrophic bacteria in the investigated hydrothermal system were abundant. This hints at substantial spatial and/or temporal variability in the composition of the microbial communities of the shallow-sea hydrothermal system could be enormous. The genetic potential of the microbial community was analyzed using marker genes for the carbon, nitrogen, and sulfur metabolism. In the present investigation, we detected the potential for chemotrophic CO<sup>2</sup> fixation mainly through the CBB cycle. Sulfur oxidation and reduction marker genes were present. Genes encoding enzymes involved in the denitrification pathway and H<sup>2</sup> utilization were also detected. The metagenomes contained abundant genes responsible for heterotrophic utilization of organic substrates including transporters, glycoside hydrolases, and peptidases genes. Cultivation attempts targeting heterotrophs resulted in the isolation of 408 heterotrophic strains that are typically considered as oligotrophic, copiotrophic, or phototrophic bacteria, in which a novel species Rhodovulum sp. P5 was isolated. Physiological analysis indicated that P5 is capable of heterotrophy, autotrophy, and phototrophy. The P5 genome possessed the complete CBB cycle, PGCs, Ni-Fe hydrogenases, and the complete repertoire of genes involved in the oxidation of reduced sulfur compounds that are putative metabolic potentials of heterotrophs associated with the adaptions to the shallowsea hydrothermal system. The genome harbored two gene sets encoding type IV secretion machinery systems, the gvp gene cluster, and the genes encoding heterodisulfide reductase that were not present in two Rhodovulum isolates from other

#### REFERENCES


habitats, which increased niche-specialization competitiveness in the shallow-sea environments. This study unveiled that the shallow-sea hydrothermal system harbored diverse microbial communities and their potential functions, enabling us to conduct more focused studies on heterotrophic activity in situ and ecophysiological features of isolated bacteria to ultimately get a complete picture on the tight coupling between microbes and biogeochemical cycling in the shallow-sea hydrothermal ecosystems in the future.

## AUTHOR CONTRIBUTIONS

KT and NJ conceived and designed the experiments. YZ, DL, YH, C-TC, DW, Y-SL, JS, and QZ conducted the experiments. KT and YZ analyzed the data. All of the authors assisted in writing the manuscript, discussed the results, and commented on the manuscript.

## FUNDING

This study was supported by the National Key Research and Development Program of China (2016YFA0601100), the National Programme on Global Change and Air-Sea Interaction (GASI-03-01-02-05), the National Natural Science Foundation of China project (41276131, 41676070, 41776167, and 91751207), and the Fundamental Research Funds for the Central Universities (20720150078). This study was a contribution to the international IMBeR project.

## ACKNOWLEDGMENTS

We thank the Marine Research Station, Institute of Cellular and Organismic Biology, Academia Sinica for providing lab space for sample processing.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.00279/full#supplementary-material



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

Copyright © 2018 Tang, Zhang, Lin, Han, Chen, Wang, Lin, Sun, Zheng and Jiao. 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.

Frontiers in Microbiology | www.frontiersin.org

fmicb-09-00279 February 20, 2018 Time: 17:10 # 15

# Microbial Diversity and Community Structure of Sulfate-Reducing and Sulfur-Oxidizing Bacteria in Sediment Cores from the East China Sea

#### Yu Zhang1,2,3, Xungong Wang1,2,3, Yu Zhen1,2,3 \*, Tiezhu Mi1,2,3, Hui He2,3,4 and Zhigang Yu3,5

<sup>1</sup> College of Environmental Science and Engineering, Ocean University of China, Qingdao, China, <sup>2</sup> Key Laboratory of Marine Environment and Ecology, Ministry of Education, Qingdao, China, <sup>3</sup> Laboratory for Marine Ecology and Environmental Science, Qingdao National Laboratory for Marine Science and Technology, Qingdao, China, <sup>4</sup> College of Marine Life Science, Ocean University of China, Qingdao, China, <sup>5</sup> Key Laboratory of Marine Chemical Theory and Technology, Ministry of Education, Qingdao, China

#### Edited by:

Hongyue Dang, Xiamen University, China

#### Reviewed by:

Wei Xie, Tongji University, China Jake Bailey, University of Minnesota, United States

> \*Correspondence: Yu Zhen zhenyu@ouc.edu.cn

#### Specialty section:

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

Received: 07 May 2017 Accepted: 18 October 2017 Published: 07 November 2017

#### Citation:

Zhang Y, Wang X, Zhen Y, Mi T, He H and Yu Z (2017) Microbial Diversity and Community Structure of Sulfate-Reducing and Sulfur-Oxidizing Bacteria in Sediment Cores from the East China Sea. Front. Microbiol. 8:2133. doi: 10.3389/fmicb.2017.02133 Sulfate-reducing bacteria (SRB) and sulfur-oxidizing bacteria (SOB) have been studied extensively in marine sediments because of their vital roles in both sulfur and carbon cycles, but the available information regarding the highly diverse SRB and SOB communities is not comprehensive. High-throughput sequencing of functional gene amplicons provides tremendous insight into the structure and functional potential of complex microbial communities. Here, we explored the community structure, diversity, and abundance of SRB and SOB simultaneously through 16S rRNA, dsrB and soxB gene high-throughput sequencing and quantitative PCR analyses of core samples from the East China Sea. Overall, high-throughput sequencing of the dsrB and soxB genes achieved almost complete coverage (>99%) and revealed the high diversity, richness, and operational taxonomic unit (OTU) numbers of the SRB and SOB communities, which suggest the existence of an active sulfur cycle in the study area. Further analysis demonstrated that rare species make vital contributions to the high richness, diversity, and OTU numbers obtained. Depth-based distributions of the dsrB, soxB, and 16S rRNA gene abundances indicated that the SRB abundance might be more sensitive to the sedimentary dynamic environment than those of total bacteria and SOB. In addition, the results of unweighted pair group method with arithmetic mean (UPGMA) clustering analysis and redundancy analysis revealed that environmental parameters, such as depth and dissolved inorganic nitrogen concentrations, and the sedimentary dynamic environment, which differed between the two sampling stations, can significantly influence the community structures of total bacteria, SRB, and SOB. This study provided further comprehensive information regarding the characteristics of SRB and SOB communities.

Keywords: sulfate-reducing bacteria, sulfur-oxidizing bacteria, microbial community, high-throughput sequencing, East China Sea

## INTRODUCTION

fmicb-08-02133 November 7, 2017 Time: 12:31 # 2

Sulfur cycling, one of the key biological processes in marine sediments, is dominated by sulfate-reducing bacteria (SRB) and sulfur-oxidizing bacteria (SOB). Dissimilatory sulfate reduction, which is mediated by SRB, is considered the main process in the biomineralization of organic matter in marine sediments and might account for up to 50% of organic matter mineralization in most continental shelf sediments (Jørgensen, 1982). In addition, as much as 12–29% of the organic carbon flux on the ocean seafloor is channeled through sulfate reduction (Bowles et al., 2014). Interestingly, 80–95% of the massive amount of hydrogen sulfide formed through sulfate reduction is recycled within sediments and gradually oxidized back to sulfate (Jørgensen and Nelson, 2004). Thus, SRB and SOB control the key processes of organic matter degradation and the biogeochemical cycling of sulfur and carbon. Studying the community structure and diversity of SRB and SOB is important for revealing the roles of these bacteria in the biogeochemical cycles of carbon and sulfur and for providing insight into the biological factors driving the marine sulfur cycle.

SRB and SOB show high diversity, including both phylogenetic and metabolic diversity (Ghosh and Dam, 2009; Müller et al., 2014). To explore the environmental abundance and diversity of SRB and SOB, genes encoding key enzymes in the sulfate reduction and sulfur oxidation biochemical pathways have been used as molecular markers. For example, the dissimilatory sulfite reductase gene dsrAB, which encodes a key enzyme that catalyzes the last, essential step in the dissimilatory sulfate reduction pathway, has been frequently employed as a functional gene in studies of SRB in various environments (Foti et al., 2007; Pester et al., 2012; He et al., 2015). The known SOB have been demonstrated to use different enzymes, pathways, and electron transport and energy conservation mechanisms for the oxidation of sulfide. The sulfur-oxidizing (Sox) (Meyer et al., 2007), reverse dissimilatory sulfite reductase (Loy et al., 2009), adenosine-5-phosphosulfate reductase (Meyer and Kuever, 2007), and sulfide quinone oxidoreductase enzyme systems (Pham et al., 2008) play vital roles in sulfide oxidation, and among these, the Sox multi-enzyme system is considered a fundamental and primordial molecular mechanism for sulfur oxidation (Ghosh and Dam, 2009) that is widespread among the known SOB (Meyer et al., 2007). Moreover, the soxB gene, which encodes the SoxB subunit of the Sox enzyme system, has been widely employed to characterize the abundance and diversity of SOB in various environments (Kojima et al., 2014; Thomas et al., 2014; Tourna et al., 2014).

The diversity and community structure of SRB and SOB have, to date, been studied using denaturing gradient gel electrophoresis (DGGE) (Varon-Lopez et al., 2013; Yang et al., 2015), restriction fragment length polymorphism (Luo et al., 2011; Reed and Martiny, 2013), and clone library approaches (Yang et al., 2013; Purcell et al., 2014). However, obtaining comprehensive information about the highly diverse SRB and SOB communities, particularly SRB and SOB from the prime habitats of marine sediments, using these methods is challenging (Aoki et al., 2015; Zhang et al., 2016). High-throughput sequencing, which constitutes a powerful approach for achieving complete coverage of microbial communities, has been recently applied in the analysis of microbial community diversity and composition. For instance, high-throughput sequencing of the 16S rRNA gene revealed the presence of a number of novel bacterial groups (Zhu et al., 2013; Mahmoudi et al., 2015) and a high diversity of bacteria (Wang et al., 2015). Moreover, the community structure and diversity of functional microbes, particularly nitrogen-cycling microbes, such as ammoniaoxidizing bacteria, ammonia-oxidizing archaea and denitrifying microbes, have been investigated through high-throughput sequencing analyses of functional genes (Tago et al., 2014; Saarenheimo et al., 2015). However, only a few studies have evaluated the community structure and diversity of SRB and SOB using high-throughput sequencing (Meyer et al., 2016; Cui et al., 2017).

The East China Sea (ECS) is the largest marginal sea of the Northwest Pacific Ocean, with a vast continental area of 0.5 × 10<sup>12</sup> m<sup>2</sup> . The Changjiang River (Yangtze River), which transports great quantities of freshwater, nutrients, organic matter, and other chemical elements into the ECS, has a marked effect on the sea (Milliman and Meade, 1983; Beardsley et al., 1985). The ECS is also influenced by the warm and oligotrophic Taiwan Warm Currents in the south and the Kuroshio Current (from the Western Equatorial Pacific) in the east (Jiao et al., 2005). More interestingly, the ECS has also experienced serious environmental problems, including eutrophication, harmful algal blooms, and chemical pollution (Shen et al., 2011; Cheng et al., 2012; Yu et al., 2013). Because of the strong influence of its riverine inputs, currents and environmental problems, the ECS exhibits complex physical and chemical conditions and has become an ideal study area for ecological investigations of the temporal and spatial dynamics of biota (Dang et al., 2008; Xiong et al., 2014; Zhang et al., 2014). Increasing research efforts have focused on microbial ecology in the ECS, particularly the microbial community structure and diversity of organisms associated with the sulfur cycle, such as SRB and SOB (Nie et al., 2009; Wu et al., 2009; He et al., 2015); however, because of the limitations of the applied methods (DGGE or clone libraries), only a few dominant species have been identified. In the present study, the 16S rRNA gene and two functional genes (dsrB and soxB) were combined in a high-throughput sequencing analysis, and this combination allowed for an in-depth analysis of the community structure and diversity of SRB and SOB in sediment cores from the ECS. In addition, quantitative PCR (qPCR) was performed to reveal the vertical distribution of SRB and SOB populations and provide insight into the SRB and SOB community characteristics.

#### MATERIALS AND METHODS

#### Site and Sampling Description

Two core samples from ECS sediments were collected from station S31 (length: 21 cm) and station S33 (length: 50 cm) during a research cruise in July 2011 on the R/V Run-Jiang (Supplementary Figure S1). A QuAAtro nutrient auto analyzer

(Seal Analytical Ltd., United Kingdom) was used to measure the pore-water dissolved N, P, and Si concentrations of the following: nitrate (NO3-N), nitrite (NO2-N), ammonium (NH4- N), phosphate (PO4-P), and silicate (SiO3-Si) (Supplementary Table S1). The sampling methods as well as the methods used for the determination of environmental parameters [pH and the concentrations of Fe(II), Mn(II), sulfate and excess <sup>210</sup>Pb] were described in detail by He et al. (2015) (Supplementary Table S1 and Figure S2). After collection, the samples from stations S31 and S33 were divided into three depth sections (0–4, 8–12, and 16–20 cm) and five depth sections (0–4, 8–12, 16–20, 32–36, and 46–50 cm), respectively, and subsequently frozen at −80◦C for nucleic acid extraction.

#### DNA Extraction

Genomic DNA was extracted from each sediment sample using the PowerSoil DNA Isolation Kit (Mo Bio Laboratories Inc., Carlsbad, CA, United States) according to the manufacturer's recommended protocol. The extracted DNA was stored at −80◦C prior to further analyses.

## High-Throughput Sequencing and Data Processing

Illumina HiSeq2500 (PE250) Sequencing and Analysis The community structure, richness, and diversity of total bacteria and SRB were studied through Illumina HiSeq2500 (PE250) sequencing based on the 16S rRNA and dsrB genes. Fragments of the 16S rRNA gene (∼466 bp, V3–V4 region) and dsrB gene (∼350 bp) were amplified using the primer pairs 341F/806R and DSRp2060F/DSR4R, respectively (Supplementary Table S2); these primer pairs have been widely employed in previous studies of bacteria from various environments (Geets et al., 2006; Michelsen et al., 2014). PCR amplification was performed in a 30 µL reaction volume containing 15 µL of Phusion <sup>R</sup> High-Fidelity PCR Master Mix (New England Biolabs), 0.2 µM forward and reverse primers, and approximately 10 ng of template DNA. The amplification reactions were performed in a thermal cycler (Bio-Rad T100, United States) and consisted of an initial denaturation step at 98◦C for 1 min followed by 30 cycles of denaturation at 98◦C for 10 s, annealing at 50◦C for 30 s, elongation at 72◦C for 30 s, and a final step at 72◦C for 5 min. The PCR products were analyzed via 2% agarose gel electrophoresis to assess the quality and size of the resultant amplicons. The PCR products were then mixed at equidensity ratios and purified using the GeneJET Gel Extraction Kit (Thermo Fisher Scientific). 16S rRNA and dsrB gene libraries were subsequently prepared using the NEB Next <sup>R</sup> UltraTM DNA Library Prep Kit (New England Biolabs), and the libraries were sequenced on the Illumina HiSeq2500 (PE250) platform following the manufacturer's recommendations.

Paired-end reads (16S rRNA and dsrB) were assigned to the samples based on their unique barcode, truncated by removing the barcode and primer sequence, and then merged using FLASH (Version 1.2.7) (Magoè and Salzberg, 2011). The merged reads were subsequently subjected to quality-based filtering with QIIME (Caporaso et al., 2010). Briefly, the reads were truncated at any site containing more than three sequential bases with a Phred quality score (Q) below 20 and any read containing ambiguous base calls and reads with less than 75% (with respect to the total read length) consecutive high-quality base calls (Wang et al., 2016) were discarded.

#### Pyrosequencing and Sequence Analysis

The soxB gene (∼750 bp) was amplified with the soxB693F/soxB1446B primer set, which was previously shown to yield the most successful and reliable amplification results (Supplementary Table S2) (Meyer et al., 2007). The 25-µL amplification reaction for the soxB gene included 4 µL of 5 × Q5 reaction buffer (New England Biolabs), 2 µL of 2.5 mM deoxynucleoside triphosphate (dNTP) mix, 5 µL of 5 × Q5 High GC Enhancer, 1 µL of each primer (10 µM), 0.25 µL of Q5 High-Fidelity DNA Polymerase (New England Biolabs), 2.5 µL of template DNA, and 8.25 µL of double-distilled H2O. The reactions were maintained at 98◦C for 5 min for DNA denaturation; they were then subjected to 32 cycles of 98◦C for 30 s, 55◦C for 40 s, and 72◦C for 1 min and then to a final extension at 72◦C for 7 min to ensure complete amplification. The obtained PCR product was cut from a 1.5% agarose gel and purified using the AxyPrep DNA Gel Extraction Kit (Axygen, AP-GX-250). Pyrosequencing of the soxB gene was subsequently conducted using the 454 FLX+ platform according to the manufacturer's recommended protocol.

The raw data were processed using the QIIME pipeline (Caporaso et al., 2010). For quality filtering, low-quality reads shorter than 150 bp, reads with an average quality score below 19, sequences with ambiguous base calls and sequences with homopolymer runs exceeding 6 bp were removed (El-Chakhtoura et al., 2015). Moreover, the barcode and primer regions were trimmed from the sequences.

#### Operational Taxonomic Unit Clusters and Taxonomic Assignment

After quality control, UPARSE (Edgar, 2013) was employed to cluster all of the clean reads into operational taxonomic units (OTUs) using 97 and 90% similarity cutoffs (Petri et al., 2001; Müller et al., 2014) for the 16S rRNA gene and functional genes (dsrB and soxB), respectively. In addition, the most abundant sequence from each OTU was selected as the representative sequence. Taxonomic information for each representative 16S rRNA and functional gene sequence was obtained by searching the Greengenes and the NCBI (National Center for Biotechnology Information) databases, respectively. Reads that did not match any sequences in the database were clustered into the unclassified group.

#### Quantification of Gene Copies in Sediments

The total bacterial 16S rRNA gene, the SRB dsrB gene, and the SOB soxB gene in the sediment samples were amplified using the primers 341F/518R, DSRp2060F/DSR4R, and soxB693F/soxB1164BK145, respectively (Supplementary Table S2); these primers were demonstrated to work effectively in previous studies (Geets et al., 2006; Dang et al., 2010;

Krishnani et al., 2010). The method used for quantification of the dsrB and 16S rRNA genes in the investigated samples has been described previously (He et al., 2015). All qPCR assays targeting the soxB gene were performed in triplicate using an ABI PRISM <sup>R</sup> 7500 Sequence Detection System (Applied Biosystems, Foster City, CA, United States). Each 20-µL qPCR mixture contained the following reagents: 10 µL of FastStart Universal SYBR Green Master Mix (Rox) (Roche Diagnostics, Mannheim, Germany), 0.2 µg µL <sup>−</sup><sup>1</sup> of bovine serum albumin, 0.4 µM of each primer, and 2.0 µL of template DNA. The qPCR amplification conditions were as follows: 95◦C for 10 min followed by 40 cycles of 15 s at 95◦C, 45 s at 55◦C, and 1 min at 72◦C. To evaluate the specificity of the qPCR amplification, the PCR products were sequenced by the Beijing Genomics Institute. In addition, the specificity of the amplification products was verified by melting curve analysis and visualized in agarose gels. Melting curves were obtained at 60–95◦C; a read was obtained after every 1◦C increase in temperature, and the temperature was maintained for 1 s between reads. The resultant qPCR data were analyzed using ABI PRISM 7500 SDS software.

Plasmids containing the target gene fragments were extracted from Escherichia coli hosts using a FastPlasmid Mini Kit (CWBIO, Beijing, China) and quantified using a Picodrop microliter spectrophotometer (Picodrop, Saffron Walden, Essex, United Kingdom). Standard curves for the qPCR assays were obtained with serial 10-fold dilutions of reference plasmids. Standard curves with efficiencies ranging from 90 to 110% and a corresponding R-value greater than 0.99 were considered credible.

#### Statistical Analyses

After randomly reducing the number of reads to the lowest number of reads in any individual sample, the richness indices (Chao 1 estimates), diversity indices (Shannon index), and Good's coverage were obtained with QIIME (Caporaso et al., 2010). Furthermore, unweighted pair group method with arithmetic mean (UPGMA) clustering based on a weighted UniFrac distance matrix was conducted using QIIME to examine the differences in the bacterial communities between the sediment samples. The relationships between community composition and sediment characteristics were examined using BIOENV, detrended correspondence analysis (DCA), and redundancy analysis (RDA) with the R package vegan (Oksanen et al., 2011). The DCA analysis indicated that a linearmodel-based RDA analysis was more suitable than unimodal CCA for our data (Hernández-Landa et al., 2015). BIOENV provides the subset of environmental variables that best explain the community variation among the sampling stations using Spearman rank correlation, and RDA determines the percentage of the community composition variation explained by this subset of environmental variables. The significance of the environmental variables was tested by a Monte Carlo permutation test (999 unrestricted permutations, p < 0.05). A Pearson's correlation analysis between microbial abundance and environmental parameters was performed using SPSS statistics software.

## Nucleotide Sequence Accession Numbers

The sequence data generated in this study were deposited in the NCBI Short Read Archive database under accession numbers SRP077048 (16S rRNA), SRP077077 (dsrB), and SRP077086 (soxB).

## RESULTS

## Bacterial 16S rRNA Gene Analysis

A total of 371,534 high-quality 16S rRNA gene sequences with an average length of 418 bp were obtained from an Illumina HiSeq2500 (PE250) sequencing analysis of the eight sediment samples. The high-quality sequences were clustered into OTUs with 97% sequence similarity. Accordingly, bacterial richness (Chao 1 estimator) and diversity (Shannon index) estimates were calculated for each sample and are shown in **Table 1**. Good's coverage estimate for each sample ranged from 98.2 to 99.2%, indicating that the sampling was sufficient to cover almost all bacterial communities. A total of 1,607–2,196 OTUs with 97% similarity were identified in the eight sediment samples. The samples from S33 exhibited more OTUs than those from S31, and the surface sediment (0–4 cm) samples presented the lowest number of OTUs at each station. The diversity indices (Shannon indices) and species richness indices (Chao 1) indicated that the samples from S33 displayed greater richness and diversity than those from S31. These results revealed higher microbial diversity at S33. Furthermore, the diversity indices did not change markedly with increasing depth. In addition, the lowest and the highest richness indices were obtained for the 0–4 and 32–36 cm sections, respectively.

In total, 52 different bacterial phyla were detected across all sediment samples, and the dominant phyla (exhibiting a relative abundance >1% in at least one sample), which accounted for more than 94.57% of the total sequences in each sample, are shown in **Figure 1**. Proteobacteria was the most abundant phylum in all samples and accounted for 36.9–62.4% of all sequences. After the relative abundance of Proteobacteria reached its peak value at a depth of 8–12 cm, it exhibited a tendency to significantly decline with increasing depth at each station. Within Proteobacteria, the majority of sequences were assigned to Deltaproteobacteria (25.8–70.0%), Gammaproteobacteria (12.7– 46.5%), and Alphaproteobacteria (11.1–34.3%) (**Figure 2**). Furthermore, Deltaproteobacteria and Alphaproteobacteria were more abundant at S33, whereas Gammaproteobacteria was more abundant at S31.

In the shallow sediment layers (0–20 cm) from each station, Actinobacteria and Acidobacteria were the second and third most dominant communities, accounting for 5.8–27.6 and 6.5– 9.4% of all bacteria amplicons, respectively. Furthermore, the relative abundance of Actinobacteria decreased with increasing depth at each station, whereas the relative abundance of Acidobacteria increased with increasing depth at S31 and decreased with increasing depth at S33. In the deeper sediment layers (32–36 and 46–50 cm) from S33, Chloroflexi and


TABLE 1 | Vertical distribution of sediment bacterial community diversity and richness estimators based on the 16S rRNA gene.

Crenarchaeota were the second and third most dominant communities, accounting for 13.1 and 11.4% of bacterial amplicons, respectively.

Above all, the bacterial communities at the two investigated stations showed similarities in their dominant groups of phyla, particularly in shallow sediment layers. However, certain variations, such as differences in class groups, were observed between the two stations. The similarities and dissimilarities between the bacterial communities in the different sediment samples were further quantified through UPGMA clustering analysis based on the weighted UniFrac distance metric (**Figure 3A**). Overall, the bacterial community structures in the shallow sediment layers (0–20 cm) from S31 and S33 were similar, and those in the deeper sediment layers (32–36 and 46–50 cm) from S33 clustered separately. These findings, which showed high jackknife support, suggest that sampling depth plays a relatively important role in bacterial community structure. However, analysis of the cluster containing the shallow sediment samples revealed that the samples from S31 formed a separate group and clustered away from the samples from S33, revealing the existence of differences in bacterial community structure between the two sampling sites.

#### Functional Gene Analysis Functional Gene Diversity

The total numbers of high-quality sequences from the dsrB and soxB genes obtained from all samples were 615,062 and 107,688, respectively. Using a similarity cutoff of 90%, these sequences were classified into 829 OTUs for the dsrB gene and 380 OTUs for the soxB gene. The corresponding species richness (Chao 1 estimator) and diversity estimates (Shannon index) were calculated for each sample and are listed in **Tables 2A,B**. Good's coverage values of the functional genes were all above 99%, indicating that the obtained sequences could adequately reflect the diversity of SRB and SOB in the sediments. Overall, the soxB gene diversity was substantially higher than that of the dsrB gene, whereas the dsrB gene richness was markedly higher than that of the soxB gene.

The diversity indices of the dsrB gene did not change markedly with increasing depth at S31 and increased with increasing depth

at S33. Moreover, the diversity indices of the dsrB gene in the shallow sediment layers (0–20 cm) from S31 were greater than those from S33 and lower than those obtained for the deeper sediment layers (32–36 and 40–46 cm) from S33. The richness of the dsrB gene was lowest in the 8–12 cm sediment depth range at station S31 and reached its peak value in the depth range of 32– 36 cm at station S33. For the soxB gene, the diversity and richness indices decreased with increasing depth at S31, whereas at S33, the diversity and richness indices decreased with increasing depth until suddenly increasing and peaking in the depth range of 46–50 cm.

#### Functional Gene Community Composition **dsrB gene**

The community structure of the dsrB gene in the samples was analyzed based on the 109 dominant (showing >1% abundance in at least one sample) and core (common to all samples) OTUs (representing 85.3–98.2% of the total sequences). Overall, 103 OTUs, accounting for 81.1–94.7% of the total dsrB sequences, were affiliated with Deltaproteobacteria, and six OTUs, representing 0.45–6.6% of the total dsrB sequences, were classified as belonging to Firmicutes. As shown in **Figure 4**, among the dominant Deltaproteobacteria, the majority of sequences (15 OTUs with relative abundances ranging from 40.91 to 73.36%) could not be assigned at the class or family levels. The 42 OTUs belonging to the Desulfobacteraceae family (ranging from 9.60 to 34.05%) represented a significant fraction and tended to increase with increasing depth at each station. A total of 42 OTUs with relative abundances ranging from 4.23 to 7.49% were affiliated with Syntrophaceae. The Desulfobulbaceae family (four OTUs) reached an abundance of 4.06% on average at S31 but only accounted for 0.40% of the sequences at S33.

The relative abundance of all dominant OTUs (total of 28 OTUs) from each sample exceeded 80%, indicating that the community composition of the dsrB gene in sediments was well reflected by the dominant OTUs. Thus, the distribution of dominant OTUs was analyzed to better understand the composition and structure of the dsrB gene in the samples (**Figure 5**). OTU1, showing relative abundances ranging from 38.8 to 63.8%, was the predominant SRB group in all sediment layers. Moreover, the relative abundance of OTU1 decreased with increasing depth at S33, whereas at S31, no significant difference was observed between the depth ranges of 0–4, 16– 20, and 8–12 cm. In addition, the relative abundance of OTU1 in the shallow sediment layers (0–20 cm) from S33 was greater than that in the samples from S31. OTU2 reached an average abundance of 3.98% at S31 but only accounted for 0.07–0.5% of the sequences at S33. A similar trend of differences was observed for OTU5, OTU6, OTU7, OTU8, and OTU9, whereas OTU3, OTU4, OTU10, OTU11, and OTU12 were more abundant at S33 than at S31.

Taken together, the results regarding the distribution of dominant OTUs at S31 and S33 indicated certain differences. Interestingly, similar results were obtained via UPGMA clustering analysis (**Figure 3B**). The UPGMA analysis showed that the SRB communities at S31 and S33 clustered separately from each other, revealing the dissimilarity in the SRB community structure between the two sampling stations and among different sediment depths.

#### **soxB gene**

Taxonomic classification revealed that Proteobacteria accounted for the most abundant clades at the phylum level, representing 46.91–77.17% (131 OTUs) of the sequences. Chlorobi and

Spirochaetes were the second and third most dominant phyla, accounting for 5.32 (two OTUs) and 2.31% (eight OTUs) of the sequences, respectively (**Figure 6**). In addition, 238 OTUs representing 20.05–45.82% of the sequences were unclassified (**Figure 6**). The dominant Proteobacteria phylum was classified into three classes: Alphaproteobacteria (varying from 31.37 to 53.00%), Betaproteobacteria (varying from 1.00 to 14.64%), and Gammaproteobacteria (varying from 4.28 to 12.50%).

The composition and structure of the SOB community in the various samples were compared at the family level (**Figure 7**). Bradyrhizobiaceae was the dominant family in all samples and was more abundant (34.72%) in S31 than in S33 (ranging from 15.79 to 29.86%). Four other minor families (Rhodobacteraceae, Hyphomicrobiaceae, Rhodospirillaceae, and an unclassified group) within Alphaproteobacteria were more abundant in S33 than in S31. Among the Betaproteobacteria, Burkholderiaceae,



#### TABLE 2B | Similarity-based OTUs and species richness and diversity estimates based on the soxB gene.


with a relative abundance of 4.31% (on average), was the dominant family in S31, whereas an unclassified group with a relative abundance of 5.91% (on average) was the dominant family in S33. Furthermore, within Gammaproteobacteria, Ectothiorhodospiraceae was the dominant family in almost all samples from S33, but its abundance was only 0.11–0.71% at S31. The UPGMA clustering analysis (**Figure 3C**) based on the weighted UniFrac distance metric showed that the SOB communities from S31 and S33 clustered separately from each other, further demonstrating the differences in the SOB community structure between the two sampling stations and among different sediment depths.

#### Influence of Environmental Variables on Total Bacterial, SRB, and SOB Communities

BIOENV identified that the environmental factors most strongly correlated with the total bacteria, SRB, and SOB communities were depth, pH, dissolved inorganic nitrogen (DIN), phosphate

(PO4-P), and silicate (SiO3-Si) (correlation >0.6). Together, these variables explained 76.58, 90.10, and 88.25% of the variation in the bacteria, SRB, and SOB communities, respectively, as determined through RDA (**Figure 8**). The environmental variables that were found to contribute significantly to the microbial community–environment relationship (900 Monte Carlo permutations) were depth and DIN for all of the examined genes (16S rRNA, dsrB and soxB gene).

#### Abundance of Total Bacterial 16S rRNA and Functional Genes

The abundance of total bacterial 16S rRNA and functional genes (dsrB and soxB) in the 21- and 50-cm-long core samples from S31 and S33, respectively, were determined via qPCR (**Figure 9**). In the core sample from S31, the vertical abundance profile of the dsrB, soxB, and 16S rRNA genes showed marked fluctuations, and the range of fluctuation in the abundances of the 16S rRNA and soxB genes was smaller than that in the abundance of the dsrB gene. However, analysis of the core sample from S33 showed that the abundances of the dsrB, soxB, and 16S rRNA genes gradually decreased with increasing depth.

## DISCUSSION

In this study, a comparative analysis of the sulfur cycle-related microbial communities (SRB and SOB) in sediment cores was

performed using an approach combining 16S rRNA, dsrB, and soxB gene high-throughput sequencing with qPCR analyses. The combination of these molecular techniques provided a detailed description of the SRB and SOB communities. The traditional approaches (e.g., DGGE and clone library), which have been widely applied for the detection of SRB and SOB in various environments, might notably underestimate their diversity. High-throughput sequencing based on functional genes (dsrB and soxB) can result in a greater number of sequences and almost complete coverage (Good's coverage values were all greater than 99%), providing more comprehensive information regarding the diversity and community composition of the SRB and SOB communities. Thus, the diversity, richness and OTU numbers observed in our high-throughput sequencing analyses were notably greater than those obtained in previous studies that employed other molecular methods (Jiang et al., 2009; Yang et al., 2013; Yousuf et al., 2014; He et al., 2015). Further analysis indicated that rare species make important contributions to the greater richness and diversity of the SRB and SOB communities. For instance, as shown in **Table 3A**, the number of dominant OTUs in each sample was similar, ranging from 8 to 13 and representing 3.08% of OTUs (on average); however, these OTUs comprised more than 77% of the sequence abundance for the dsrB gene. Although the rare OTUs (relative abundance <0.01%) (Galand et al., 2009) accounted for 54.03% of OTUs (on average), they only comprised 0.84% (on average) of the

sequence abundance. Similar results were obtained for the soxB gene (**Table 3B**).

The 16S rRNA sequencing analysis identified 235 and 191 OTUs as belonging to potential SRB and SOB, respectively, based on family species classification. Furthermore, these SRB OTUs were affiliated with 10 families, accounting for 7.63– 19.62% (14.12% on average) of the total 16S rRNA gene sequences (**Table 4A**), and the SOB OTUs were affiliated with 12 families, accounting for 5.99–18.78% (12.53% on average) of the total 16S rRNA sequences (**Table 4B**). This finding is consistent with the results of a previous study in this region that found that numerous sequences were affiliated with potential SRB and SOB (Ye et al., 2016). Overall, these results confirm the high abundance of potential SOB and SRB groups in the ECS sediments and suggest the existence of an active sulfur cycle in this area, and these sulfur cycle-related bacterial communities have considerable potential for further exploration.

The dsrB gene sequencing analysis performed in this study identified two phyla (Proteobacteria and Firmicutes), and Deltaproteobacteria (within Proteobacteria) represented the dominant SRB. Our data were consistent with previous observations obtained from various environments, such as marsh sediments (Jemaneh et al., 2013), the Great Salt Lake (Kjeldsen et al., 2007), mangrove sediments (Varon-Lopez et al., 2013), wastewater treatment plants (Biswas et al., 2014), and marine and estuary sediments (Jiang


TABLE 3A | Rare and dominant OTUs based on the dsrB gene.

fmicb-08-02133 November 7, 2017 Time: 12:31 # 12

TABLE 3B | Rare and dominant OTUs based on the soxB gene.


et al., 2009; Wu et al., 2009; He et al., 2015), suggesting that Deltaproteobacteria have a wide adaptation range and play major roles in global sulfur cycling. The soxB gene sequencing results revealed that the SOB community was dominated by Alphaproteobacteria and unclassified groups, followed by Gammaproteobacteria and Betaproteobacteria. For comparison, Gammaproteobacteria and Epsilonproteobacteria are considered dominant Sox organisms in marine sediments (Sievert et al., 2008; Lenk et al., 2011; Akerman et al., 2013; Dyksma et al., 2016); however, no Epsilonproteobacteria groups were observed in the present study. This discrepancy could be due to primer bias; the primers used for soxB gene amplification amplify the soxB genes of Epsilonproteobacteria and Chloroflexi very poorly (Meyer and Kuever, 2007). In addition, Bradyrhizobiaceae within Alphaproteobacteria and Burkholderiaceae within Betaproteobacteria were the dominant families in our study area; however, previous research revealed that these families might adapt to oligotrophic sulfur environments, such as soil (Jung et al., 2005; Masuda et al., 2010) and rhizosphere soils (Anandham et al., 2008; Masuda et al., 2016). Therefore, further study is required to determine whether the families described here are the result of contamination or actually exist. More interestingly, most of the domain and core OTUs obtained through dsrB and soxB gene high-throughput sequencing were difficult to clearly assign at the family or genus level, partially due to the limited number of reference sequences. For instance, 15 OTUs for the dsrB gene, with a relative abundance ranging from 40.91 to 73.36%, and 238 OTUs for the soxB gene, with a relative abundance ranging from 20.05 to 45.82%, could not be assigned at the family or genus level. Overall, these findings suggest that the ECS environment contains a high abundance of yet unknown SRB and SOB lineages, which should be studied in more detail. Indeed, this phenomenon is consistent with the following recent findings: 62% of dsrB sequences in surface sediments from the ECS exhibit low similarity with previously cultured SRB (Zhang et al., 2016); 57.6% of dsrAB clones from polluted harbor sediments belong to unclassified groups of SRB (Zhang et al., 2008); and 60% of sequences obtained through a dsrA gene analysis show no clear affiliation with a known SRB (Quillet et al., 2012). Furthermore, similar results have been reported by Yousuf et al. (2014) for soxB. In addition, we noticed that these unclassified OTUs exhibit high similarity with environmental dsrB/soxB sequences, as demonstrated through BLAST identity searches. For example, most of the unclassified dsrB OTUs displayed high similarities (>90%) with uncultured SRB recovered from Japanese fish farm sediments (Kondo et al., 2012) and ECS sediments (He et al., 2015; Zhang et al., 2016). Similarly, most of the unclassified soxB sequences exhibited high similarity (>80%) with sequences recovered from coastal soil ecosystems (Yousuf et al., 2014). These observations suggest that these unclassified sequences are widely distributed in marine and terrestrial environments and play a vital role in the biogeochemical cycling of sulfur and carbon.

A comparison of the SRB community composition identified based on the 16S rRNA and dsrB genes showed that 16S rRNA gene sequencing detected 10 SRB families, whereas dsrB gene sequencing only detected four SRB families (**Tables 4A,B**). Moreover, it was difficult to obtain more detailed taxonomic information for the SRB community based on the 16S rRNA and dsrB genes. Overall, the dsrB gene proved to be less efficient for the proper annotation of taxonomic information for SRB groups. This phenomenon might be attributed to the limited number of reference sequences used to annotate taxonomic information for

Zhang et al. SRB and SOB in the ECS

the dsrB gene. In contrast, the results obtained from the highthroughput sequencing of the soxB gene revealed 15 SOB families, whereas the high-throughput sequencing of the 16S rRNA gene revealed 12 SOB families. In addition, the soxB gene can provide more detailed taxonomic information (at the genus or even the species level) for the SOB community (data not shown) than the 16S rRNA gene. Taken together, the results indicate that the soxB gene might better define SOB groups.

The 16S rRNA, dsrB and soxB gene high-throughput sequencing data obtained in the present study revealed a high diversity of SRB and SOB in sediments from the ECS, and the classification of SRB and SOB communities based on 16S rRNA

TABLE 4A | SRB community composition based on taxonomic information of the dsrB and 16S rRNA genes.


–, Not detected.

TABLE 4B | SOB community composition based on taxonomic information of the soxB and 16S rRNA genes.


–, Not detected.

gene and functional gene was analyzed and compared according to the results of a previous study (Meyer et al., 2016). Overall, the present study provided more comprehensive information regarding the SRB and SOB community characteristics and suggested the existence of an active sulfur cycle in the study area. Significantly, the sulfate amount never decreases below approximately 24 mM in the core samples, which appears to contradict the conclusion that SRB and SOB communities are responsible for active sulfur cycling. In fact, this phenomenon will be observed based on the fact that the sulfate consumed by sulfate reduction is recycled within sediments by oxidation of reduced inorganic sulfur compounds to sulfate. As previously reported, 80–95% of the massive amount of hydrogen sulfide formed through sulfate reduction is recycled within sediments and gradually oxidized back to sulfate (Jørgensen et al., 1990; Jørgensen and Nelson, 2004). Jørgensen et al. (1990) reported that sulfate reduction rates increased in the top 10 cm of core in the Belt Sea; however, the concentration of sulfate varied little (approximately 25 mM). Similarly, Leloup et al. (2009) revealed that substantial sulfate reduction rates were measured in the top 25 cm of core, and the sulfate concentration profile was approximately 20 mM. However, the diversity of sulfur cyclerelated bacteria and the potential sulfur cycle activity might still be overestimated or underestimated because certain named SRB/SOB might did not actually take part in the sulfur cycle in sediments, such as Pelotomaculum and Sporotomaculum, which are not able to grow with sulfite and/or sulfate as electron acceptors even though they are closely affiliated with SRB (Brauman et al., 1998; Imachi et al., 2006). Instead, certain unnamed SRB/SOB, such as archaeal anaerobic methanotrophs, might be involved in the sulfur cycle in sediments (Treude et al., 2014). Fortunately, these groups are only found in special habitats and account for a small portion of the identified communities (Brauman et al., 1998; Imachi et al., 2006; Miyashita et al., 2009; Treude et al., 2014). Thus, the use of 16S rRNA and functional genes (dsrB and soxB) as proxies for determining the microbial groups that play active roles in sulfur cycling and for studying the diversity of SOB and SRB is a powerful approach.

The sediment environments of the ECS are highly heterogeneous (Dang et al., 2008; Fang et al., 2013); this characteristic has a strong influence on the spatial heterogeneity of the sediment microbial communities in the ECS, as observed in previous studies (Dang et al., 2008; Feng et al., 2009; Chen et al., 2014; Ye et al., 2016). Based on the sediment grain size combined with the overlying water masses, the ECS can be divided into three domains: the inner shelf mud area, the outer shelf sand area, and the slope plus Okinawa Trough mud area (Liu et al., 2006; Xu et al., 2016). S31 is located within the inner shelf mud deposits along the Zhejiang coast, whereas S33 is located at the periphery of the Zhejiang coastal mud area (Supplementary Figure S1). Thus, most of the environmental parameters showed substantial variation among the different samples and among the various sediment depths (Supplementary Table S1 and Figure S2). The concentration of DIN, for instance, ranged from 91.44 to 449.21 µM in station S31 and from 18.81 to 120.73 µM in station S33 (Supplementary Table S2). <sup>210</sup>Pb has been proven to serve as an effective proxy for sedimentary

dynamics processes, such as erosion–transportation–deposition, boundary scavenging, and resuspension process, in coastal and shelf regions (Huh and Su, 1999; Su and Huh, 2002). The excess <sup>210</sup>Pb activity varied markedly between the two sediments (Supplementary Figure S2). At station S31, <sup>210</sup>Pbex activity was stable from the surface sediment to a depth of 8 cm; however, at a depth below 12 cm, <sup>210</sup>Pbex activity increased gradually and displayed characteristics of reverse accumulation, indicating that this sedimentary layer has experienced sedimentary dynamic events. At station S33, <sup>210</sup>Pbex activity remained stable until a depth of 5 cm and then gradually declined. This finding indicates that the sedimentary dynamic environment at S33 was more stable than that at S31, as observed in our previous study (He et al., 2015). Our results showed that the abundances of 16S rRNA, dsrB and soxB genes show substantial fluctuations at S31 (**Figure 9**), and none of the gene abundance values were significantly correlated with sediment depth. In contrast, all of the genes (16S rRNA, dsrB and soxB) gradually decreased with increasing depth at S33 (**Figure 9**) and were significantly negatively correlated with sediment depth. These results suggest that the sedimentary dynamic environment is likely an important factor in controlling the vertical distribution of the abundances of total bacterial 16S rRNA gene and functional genes (dsrB and soxB). Further analysis revealed similar vertical fluctuations in the abundance of the 16S rRNA, dsrB and soxB genes at S31, but the range of fluctuation obtained for the abundance of the 16S rRNA and soxB genes was smaller than that observed for the dsrB gene. In addition, the ratios of the functional genes (dsrB and soxB) to the total bacterial 16S rRNA gene were calculated: for the dsrB gene, the ratio ranged from 0.22 to 40.56% in S31 and from 0.06 to 4.2% in S33, and the ratios calculated for the soxB gene were less than 1% at both stations. These results indicate that the abundance of SRB might be more sensitive to the sedimentary dynamic environment than that of SOB and total bacteria.

The UPGMA clustering analysis results demonstrated different community structures of total bacteria, SRB, and SOB at the two sampling sites and among the different sediment depths, which might be due to the differences in the sediment depths and sedimentary dynamic environments between the two sampling stations. Overall, the sedimentary dynamic environment is most likely related to unmeasured variables, such as salinity, organic matter, dissolved oxygen, temperature, and H2S, which are widely accepted as important factors in controlling the community

#### REFERENCES


structures of total bacteria and sulfur-cycling bacteria (SRB and SOB) (Headd and Engel, 2013; Ye et al., 2016; Zhang et al., 2016). In addition, an RDA analysis revealed that sediment depth and DIN have a significant influence on the community structure of total bacteria, SRB, and SOB. In summary, our results suggest that environmental parameters, such as sediment depth and DIN, and the sedimentary dynamic environment, which showed differences between the two sampling stations, can significantly influence the community structure of total bacteria, SRB, and SOB.

#### AUTHOR CONTRIBUTIONS

YZhe, TM, and ZY conceived and designed the experiments; YZha, XW, and HH performed the experiments and analyzed the data; YZha and YZhe wrote the paper.

#### FUNDING

This work was supported by the National Natural Science Foundation of China (Nos. 41620104001 and 41521064), the Scientific and Technological Innovation Project of the Qingdao National Laboratory for Marine Science and Technology (2016ASKJ02) and the Open Fund of Key Laboratory of Marine Ecology and Environmental Sciences, Institute of Oceanology, Chinese Academy of Sciences, and Laboratory of Marine Ecology and Environmental Science, Qingdao National Laboratory for Marine Science and Technology (KLMEES201601).

#### ACKNOWLEDGMENT

We are grateful to all staff at Run-Jiang for the assistance provided in the collection of samples and geochemical data during the cruise.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2017.02133/full#supplementary-material




East China Sea (ECS) from 2000 to 2010. J. Environ. Prot. 2, 1285–1294. doi: 10.4236/jep.2011.210148



**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 Zhang, Wang, Zhen, Mi, He and Yu. 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.

# Sediment Depth-Dependent Spatial Variations of Bacterial Communities in Mud Deposits of the Eastern China Marginal Seas

Yanlu Qiao1,2, Jiwen Liu1,2 \*, Meixun Zhao2,3 and Xiao-Hua Zhang1,2

<sup>1</sup> Laboratory of Marine Microbiology, College of Marine Life Sciences, Ocean University of China, Qingdao, China, <sup>2</sup> Laboratory for Marine Ecology and Environmental Science, Qingdao National Laboratory for Marine Science and Technology, Qingdao, China, <sup>3</sup> Key Laboratory of Marine Chemistry Theory and Technology, Ministry of Education, Ocean University of China, Qingdao, China

The mud sediments of the eastern China marginal seas (ECMS) are deposited under different hydrodynamic conditions with different organic matter sources. These events have been demonstrated to exert significant influences on microbial communities and biogeochemical processes in surface sediments. However, the extent to which such effects occur in subsurface microbial communities remains unclear. In this study, both horizontal and vertical (five sites, each for eight layers) distributions of bacterial abundance and community composition in mud deposits of the South Yellow Sea (SYS) and East China Sea (ECS) were investigated by quantitative PCR and Illumina sequencing of the 16S rRNA gene. Both bacterial abundance and diversity were higher in the ECS than in the SYS, and tended to be higher in up than in deep layers. Proteobacteria (JTB255 marine benthic group), Acidobacteria and Bacteroidetes were dominant in the upper layers, whereas Lactococcus, Pseudomonas, and Dehalococcoidia were enriched in the deep layers. The bacterial communities in surface and subsurface sediments showed different inter-taxa relationships, indicating contrasting co-occurrence patterns. The bacterial communities in the upper layer samples clustered in accordance with mud zones, whereas those in the deep layer samples of all sites tended to cluster together. TOC δ <sup>13</sup>C and TON δ <sup>15</sup>N significantly affected the bacterial community composition, suggesting that the abundance and composition of organic matter played critical roles in shaping of sedimentary bacterial communities. This study provides novel insights into the distribution of subsurface bacterial communities in mud deposits of the ECMS, and provides clues for understanding the biogeochemical cycles in this area.

Keywords: bacterial communities, eastern China marginal seas, spatial distribution, diversity, mud deposits

## INTRODUCTION

Marginal seas are the transitional zones between the coastal and open oceans and occupy about 10% of the global ocean. These shallow, narrow and fast-deposition areas are reported to be major reservoirs of organic carbon burial in the marine system (Hedges and Keil, 1995), and have significant impacts on global biogeochemical cycles and even global climate changes to a

#### Edited by:

Stefan M. Sievert, Woods Hole Oceanographic Institution, United States

#### Reviewed by:

Maria Pachiadaki, Bigelow Laboratory for Ocean Sciences, United States Meng Li, Shenzhen University, China

> \*Correspondence: Jiwen Liu liujiwen@ouc.edu.cn

#### Specialty section:

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

Received: 08 February 2017 Accepted: 14 May 2018 Published: 31 May 2018

#### Citation:

Qiao Y, Liu J, Zhao M and Zhang X-H (2018) Sediment Depth-Dependent Spatial Variations of Bacterial Communities in Mud Deposits of the Eastern China Marginal Seas. Front. Microbiol. 9:1128. doi: 10.3389/fmicb.2018.01128

considerable degree (Chen C.T.A. et al., 2004; Coelho et al., 2013). Correspondingly, sediments in these areas harbor a higher abundance of microbes than those in open oceans (Kallmeyer et al., 2012). These abundant microorganisms contribute significantly to the cycling of marine biogenic elements (Azam and Malfatti, 2007; Falkowski et al., 2008), especially carbon (Dyksma et al., 2016).

Bacterial communities were found to vary in different marginal sea sediments (Bowman and McCuaig, 2003; Bertics and Ziebis, 2009; Harrison et al., 2009; Zinger et al., 2011; Wang et al., 2012; Learman et al., 2016; Probandt et al., 2017). These can be explained by environmental heterogeneity (including sediment sources and hydrodynamic conditions) that can significantly influence the distribution of microbial communities and related biogeochemical processes. For example, organic matter has been demonstrated to be a driver of benthic microbial community structure across the Antarctic surface sediment (Learman et al., 2016). Meanwhile, shifts in bacterial community were observed in oil-contaminated and nitrogen-polluted sediments in the Antarctic near shore, Mediterranean Sea and East China Sea (ECS) (Powell et al., 2003; Polymenakou et al., 2006; Xiong et al., 2014), suggesting composition of organics and nutrients to be important factors as well. In addition, dissolved oxygen (DO) of the bottom water, temperature and sediment median grain size, have been detected to be vital factors shaping benthic bacterial communities in the Chinese marginal sea (Bohai Sea and Pearl Estuary), Arctic and North Sea (Wang et al., 2013; Liu et al., 2014; Zheng et al., 2014; Nguyen and Landfald, 2015; Probandt et al., 2017). Most studies mentioned above were based on surface sediments, but relatively few have focused on the vertical profile of bacterial communities in typical marginal sediments (Franco et al., 2007; Böer et al., 2009; Harrison et al., 2009; Liu et al., 2014). Depth related shifts in bacterial community in marginal sediments have been reported and were attributed to different factors such as contents of organic carbon, chlorophyll a and inorganic nutrients (Bowman and McCuaig, 2003; Böer et al., 2009; Harrison et al., 2009). None of these studies included a highly resolved vertical profile of sedimentary bacterial communities. Thus, the extent to which the benthic surface environmental heterogeneity affect subsurface microbial communities needs further investigation.

The eastern China marginal seas (ECMS) are typical eutrophic seas with different mud areas formed by sediments derived mainly from the Yellow River and Yangtze River. These mud areas are characterized by different sediment sources and hydrodynamic conditions resulted from complex water masses and ocean currents; therefore, they provide different environmental niches for microorganisms to survive. Accordingly, previous studies have shown distinct distribution patterns of functional microorganisms in different ECMS mud sediments (Yu et al., 2016; Gao et al., 2017). However, compositional distributions of total bacterial community in different mud sediments are currently unknown. We hypothesized that the total bacterial communities varied in surface sediments but converged in subsurface sediments in different mud areas of the ECMS. In this study, a high

resolution vertical profile of bacterial abundance and community composition from five sites, each for eight layers, of the ECMS was provided. In addition, the bacterial co-occurrence patterns, which can help uncover potential inter-taxa relationships, in both surface and subsurface sediments were explored by using correlation based network analysis.

## MATERIALS AND METHODS

## Study Site and Sampling

To compare sedimentary bacterial communities in different mud zones of the ECMS, five sites (SYS01, SYS02, ECS01, ECS02, and ECS03) distributed in four typical mud zones of the South Yellow Sea (SYS) and ECS were chosen. Locations of these samples have been reported by Yu et al. (2016). SYS01 and SYS02 are located in the SYS mud zone, where the deposits are mainly from sinking of the modern and old Yellow River-derived sedimentary organic matter (Hu et al., 2013). In addition, mud deposits in this area are considered as a result of the presence of cold water mass in summer, accompanied by seasonal weaken of the Yellow Sea Warm Current (Hu, 1984). Situated in the Yangtze River Estuary mud zone, ECS01 is mainly influenced by freshwater flowed out of the Yangtze River, which makes Yangtze River to be the dominant sediment source of ECS01 (Liu et al., 2007). ECS02 is located in the Zhe-Min mud zone. This area is influenced by a couple of alternatively predominant reversed currents (the Zhe-Min Coastal Current and Taiwan Warm Current); its sediments are mainly transported from the Yangtze River and the estuary mud zone along the Zhe-Min coast (Liu et al., 2007). ECS03 belongs to the distal Cheju Island mud zone, and its sediments are derived from both the Yangtze River and the old Yellow River, transported by the Yellow Sea Warm Current and river runoff from the Yangtze River (Liu et al., 2003).

The sediment samples were collected by a box corer during a cruise of R/V Dong Fang Hong 2 from 12 July to 2 August, 2013. Two PVC tubes were used to subsample the collected sediments at each site. One PVC core was immediately sliced at a 1-cm interval with a stainless-steel cutter and the sliced sediments were stored at −20◦C (onboard) or −80◦C (in laboratory) before organic matter measurement and DNA extraction. An aliquot of sediments at depths of 0–1, 12–13, and 32–33 cm from sites SYS01, SYS02, ECS02, and ECS03 was fixed with paraformaldehyde (2% final) in sterile plastic vessels and conserved in 1:1 PBS-ethanol at −20◦C for 4<sup>0</sup> , 6-diamidino-2-phenylindole (DAPI) counting. The parallel core was prepared for pore water extraction. Pore water samples were collected by the Rhizon samplers at the cm-scale, poisoned by HgCl<sup>2</sup> and stored at 4◦C before dissolved inorganic nutrient measurement. For each core, eight sediment layers that were 0–1, 1–2, 2–3, 3–5, 7–8, 12–13, 22–23, and 32–33 cm (written as −0, −1, −2, −3, −5, −10, −20, and −30 cm, respectively) were chosen for microbiological analyses. Total organic carbon (TOC), total nitrogen (TN), stable carbon (TOC δ <sup>13</sup>C) and nitrogen isotopes (TON δ <sup>15</sup>N) in sediments, dissolved inorganic nutrients (NO<sup>3</sup> <sup>−</sup>, NO<sup>2</sup> <sup>−</sup>, NH<sup>4</sup> <sup>+</sup>, PO<sup>4</sup> <sup>3</sup>−, SiO<sup>3</sup> <sup>2</sup><sup>−</sup> and SO<sup>4</sup> <sup>2</sup>−) in pore water, salinity, DO and Chl a in bottom water were determined as previously described (Yu et al., 2016).

#### DNA Extraction

fmicb-09-01128 May 30, 2018 Time: 8:52 # 3

Genomic DNA was extracted from 0.25 g of sediment (wet weight) using the Power Soil DNA Isolation Kit (Mo Bio Laboratories, Inc., Carlsbad, CA, United States) and a FastPrep-24 cell disrupter (MP Biomedicals, Irvine, CA, United States) according to the manufacturer's instructions. Quality and quantity of the extracted DNA were measured by a Nanodrop spectrophotometer ND-2000 (Thermo Fisher Scientific, United States). DNA was then subpackaged and stored at −80◦C.

#### Quantification Analysis

Paraformaldehyde fixed sediment samples were diluted and homogenized with low-power ultrasonic wave at 20 W for 30 s. A volume of 50 µl sonicated sample was mixed with 10 mL PBS, collected on the 0.2-µm pore size filter (Isopore GTTP, Millipore), and stained with DAPI. For each sample, cell numbers were counted in ten random views under the fluorescence microscope. To exclude eukaryotic cells, only cells that are 0.5–5 µm in size were counted.

Quantitative PCR was performed to quantify the abundance of total bacteria and sulfate-reducing bacteria (SRB) in the samples using primers of the 16S rRNA gene and dissimilatory sulfite reductase β-subunit (dsrB) gene, respectively (Varon-Lopez et al., 2013; Yin et al., 2013). A 20 µl mixture contained 10 µl of SYBR Premix ExTaq II (2×), 0.4 µl of ROX Reference Dye II (50×) (TaKaRa, Tokyo, Japan), 0.2 µl of primers for each gene (10 µM), and 2 µl of template. Primers and thermal cycling steps are shown in **Table 1**. All assays were conducted in triplicate with negative controls using an ABI 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, United States).

Standard curves were constructed by PCR amplifying a 10-fold serial dilution of plasmids containing target gene fragments. The amplification curves showed well linear relationships (R <sup>2</sup> > 0.999) and the amplification efficiencies were 91.65 and 95.28% for the bacterial 16S rRNA gene and dsrB gene, respectively. The single-peak melting curves and the only bond in gel electrophoresis guaranteed specificity of the qPCR analysis.

#### High Throughput Sequencing and Reads Processing

The PCR and sequencing were performed as previously described (Liang et al., 2015) with minor modifications. Primers Eub515F/Eub907R (**Table 1**) were used for bacterial 16S rRNA gene amplification. The forward and reverse primers were tagged with adapter, pad and linker sequences, and the reverse primer was linked with barcode sequences for pooling of multiple samples in one run of MiSeq sequencing. PCR was run on an ABI GeneAmp <sup>R</sup> 9700 cycler and thermal cycling steps are shown in **Table 1**. The PCR products of each sample were pooled, purified by an AxyPrepTM DNA Gel Extraction Kit (Axygen, Hangzhou, China) and quantified using a QuantiFluorTM-ST Solid Standard (Promega, Madison, WI, United States). Sequencing was conducted on a MiSeq Desktop Sequencer at Majorbio Bio-Pharm Technology Co., Ltd., Shanghai, China.

The raw data were filtered according to the pipeline of Quantitative Insights into Microbial Ecology (QIIME<sup>1</sup> , Caporaso et al., 2010). Reads were assigned to samples according to their barcodes with no mismatch. The raw reads that had a quality score higher than 20 over a 5 bp window size and a minimum length of 100 bp (Kong, 2011) were retained. The pair-end reads were joined with at least a 50 bp overlap and less than 5% mismatches using FLASH (Magoc and Salzberg, 2011). A perl script daisychopper.pl (Gilbert et al., 2009) was used to randomly subsample sequences from each sample according to the least read numbers for equalizing sampling efforts. Operational taxonomic units (OTUs) clustering and taxonomy assignment were also performed in QIIME. Specifically, OTUs were defined at a 97% sequence similarity level, and then chimera sequences were detected and removed with UCHIME (Edgar et al., 2011) as recommended by QIIME tutorials. Taxonomy was assigned using the RDP Classifier v2.2 (Wang et al., 2007) against the SILVA v115 16S rRNA gene reference database<sup>2</sup> with a minimum support threshold of 70%.

<sup>1</sup>http://qiime.org <sup>2</sup>http://www.arb-silva.de



The Illumina sequences were deposited in the National Center for Biotechnology Information Short Read Archive database under SRP076973.

#### Statistical Analysis

fmicb-09-01128 May 30, 2018 Time: 8:52 # 4

The diversity indices, including Good's coverage, Chao1 and Shannon index, were calculated for alpha diversity analysis. Molecular ecological network analyses were conducted by the package 'Hmisc,' 'igraph,' and 'qvalue' in R software (RDC TEAM, 2008) using bacterial groups at the family level with read numbers > 50 across all samples to simplify the network. Co-occurrence pairs with a Spearman's correlation coefficient > 0.7 or < −0.7 and a P-value < 0.01 (Benjamini and Hochberg adjusted) were considered as a valid co-occurrence event. The R script was provided in Supplementary Material. Gephi (version 0.8.2 beta, Bastian et al., 2009) was used for network visualization. Linear discriminate analysis (LDA) effect size (LEfSe) (Segata et al., 2011) was used to identify taxa with significant differences between mud zones and depths at various taxonomic levels. For beta diversity, classification of bacterial communities was performed by principal coordinate analysis (PCoA) using Fast UniFrac (Lozupone et al., 2011). Pairwise analyses of similarities (ANOSIM) was performed in PRIMER 5 (Plymouth Marine Laboratory, West Hoe, Plymouth, United Kingdom). The relationships between phylotypes and environmental factors were evaluated by redundancy analysis (RDA) in CANOCO (Version 5.0, Microcomputer Power) with 9999 Monte Carlo permutation tests using square root-transformed data. Pearson correlation test was also used to evaluate correlations between percentage composition of taxa and environmental factors. In order to accurately estimate the correlations, only the top 20 phyla, top 30 classes, top 50 orders, top 50 families, top 50 genera were tested.

## RESULTS

## Environmental Characterization

Detailed sediment and pore water environmental parameters have been described by Yu et al. (2016). In brief, at all sites, the NO<sup>3</sup> <sup>−</sup> concentration in pore water, TOC and TN content in sediment had a tendency to decrease with depth. Higher sedimentary C/N ratio and NO<sup>3</sup> <sup>−</sup> concentration were detected in SYS than in ECS samples. The pore water concentration of NH<sup>4</sup> + at ECS01 was higher than that at other sites. Much depleted values of TON δ <sup>15</sup>N and TOC δ <sup>13</sup>C were observed in sediment samples of ECS01.

#### Direct Cell Counting and Quantitative PCR

The result of direct cell counting showed that microbial cell numbers in the ECMS varied from 3.17 × 10<sup>8</sup> to 4.19 × 10<sup>9</sup> cells g−<sup>1</sup> (**Figure 1A**). The microbial cell numbers decreased with depth, and significant difference was observed between 0–1 cm samples and 32–33 cm samples (P < 0.05). Meanwhile, sediments from the ECS contained more microbial cells than those from the SYS (P < 0.01).

Quantitative PCR was also used to detect the abundance of the bacterial 16S rRNA gene, which showed a range of abundance between 1.5 × 10<sup>8</sup> (SYS01-30) to 9.0 × 10<sup>9</sup> copies g−<sup>1</sup> (ECS02-1) (**Figure 1B**). Similar to the result of direct cell counting, samples from sites located on the south tended to have higher copy numbers of the bacterial 16S rRNA gene. Vertically, the abundance fluctuated at a high level within the top 5 cm, and then declined gradually with depth. The 16S rRNA gene abundance was significantly correlated with PO<sup>4</sup> <sup>3</sup><sup>−</sup> (r = −0.466, P = 0.007), TN (r = 0.463, P = 0.003), TON δ <sup>15</sup>N (r = −0.439, P = 0.005), and C/N (r = −0.321, P = 0.043). Copy number of dsrB gene varied in the range of 2.7 × 10<sup>6</sup> to 9.6 × 10<sup>7</sup> copies g −1 , and no significant difference was observed among studied sites or depths.

#### Bacterial Diversity and Richness

A total of 1,472,593 overlapped reads across the 40 samples were generated through Illumina sequencing, and 1,190,856 reads were left after quality control (Supplementary Table S1). Read numbers in each sample were limited to 24,048 after rarefaction for further analyses. All sequences yielded 10,746 OTUs at a 97% sequence similarity level (Supplementary Data Sheet S2). The Good's coverage values ranged from 91.99 to 98.17% across samples, indicating that sequences generated from these samples could represent most of the bacterial community in the studied sites. Sites located in the ECS had higher Chao1 and Shannon indices than those in the SYS (P < 0.05). Site ECS03 located in the Distal Cheju Island mud area had the highest bacterial diversity in terms of Shannon index (P < 0.05). As for depth, Shannon diversity of the surface samples (0–2 cm) was higher than that of the 30 cm samples (P < 0.05). A negative correlation was found between Shannon diversity and C/N (r = −0.544, P = 0.0003).

#### Taxonomic Description

In total, 47 phyla were observed in the 40 ECMS sediment samples. The most dominant phylum was Proteobacteria occupying 45.62% of all sequences. This was followed by Planctomycetes, Chloroflexi, Acidobacteria, Bacteroidetes, Firmicutes, Nitrospirae, candidate division WS3, Gemmatimonadetes and Actinobacteria, which jointly accounted for 44.43% of all sequences. Within Proteobacteria, Deltaproteobacteria (21.93%) and Gammaproteobacteria (18.79%) were the most abundant classes. In addition, eight minor phyla (SM2F11, WCHB1-60, OC31, CKC4, candidate division KB1, GOUTA4, Thermotogae, and GAL08) were represented each by less than 50 sequences.

The bacterial community compositions varied among depths and sites (**Figure 2**). The proportions of Proteobacteria, Acidobacteria, and Bacteroidetes were high in surface sediments, whereas Chloroflexi and Firmicutes tended to be enriched in deep layers. The sample similarity analysis based on the Bray–Curtis dissimilarity at the genus level showed that samples were clustered into two groups (Supplementary Figure S1). The boundary between these two groups was plotted as a dotted line in **Figure 2**, illustrating the separation of surface and deep bacterial communities. To discover bacterial groups with significant differences between these two sediment types,

LEfSe was conducted from the phylum to genus levels with a LDA threshold of 3.5. The results of LEfSe confirmed the tendency shown at the phylum and genus levels and revealed that Lactococcus (genus of Firmicutes, dominated by Lactococcus piscium), Pseudomonas (genus of Pseudomonadales, dominated by Pseudomonas azotoformans and P. fragi) and Dehalococcoidia (class of Chloroflexi) were significantly abundant in deep layers, whereas Acidobacteria (class) and JTB255 marine benthic group (JTB255-MBG, a family of Xanthomonadales) preferred surface layers (**Figure 3**). In addition, LEfSe with a LDA value of 3.5 was also used to predict effect differences in bacterial groups among different sites (Supplementary Figure S2). A total of 28 bacterial groups, including five phyla, five classes, six orders, seven families, and five genera, were highlighted to be the specialized taxa for each site. Syntrophobacterales from the family to genus levels and Desulfobacteraceae at the family level were enriched at SYS01. The abundance of Firmicutes, Chloroflexi, and Pseudomonadales were significantly higher at SYS02. JTB255-MBG was enriched at ECS01, making the proportion of Proteobacteria reached to its top at this site. Bacteroidetes (from phylum to class levels) was enriched at ECS02. ECS03 had a higher abundance of Planctomycetes and Deltaproteobacteria than other sites.

Inter-taxa relationship network was structured based on bacterial families whose read numbers were more than 50 across all samples. A total of 242 nodes and 2289 edges were presented

in the network with a threshold of ±0.7 for Spearman's coefficient and 0.01 for P-value (**Figure 4**). The bacterial families from Proteobacteria, Acidobacteria, and Planctomycetes displayed wide correlations with others. Proteobacteria made up more than 1/3 of the nodes in the network. The families tended to networking into two modules, and hubs of the two modules belonged to Proteobacteria, Bacteroidetes, Lentisphaerae, and Chloroflexi, and Proteobacteria, Firmicutes, and Planctomycetes, respectively. Twenty-six negative correlations were observed and distributed mainly between Proteobacteria and Chloroflexi.

A total of 36 pairs of bacterial groups and environmental factors were observed to have significant correlations (P < 0.01 and | r| > 0.6) (Supplementary Table S2). Sixteen taxa belonging to Chloroflexi, Planctomycetes, Spirochaetae, candidate division OP8 and Deltaproteobacteria were significantly correlated with sediment depth. Nine taxa belonging to Deltaproteobacteria, Gammaproteobacteria and BD2-11 terrestrial group (order of Gemmatimonadetes) and seven taxa belonging to Bacteroidetes, Nitrospirae, and Planctomyces showed significant correlations with TON δ <sup>15</sup>N and TOC, respectively.

## Community Comparison at the OTU Level and Environmental Factors Explaining Community Variations

The samples clustered basically according to different mud zones as shown in the PCoA (**Figure 5A**), which considered both the topology of evolutionary trees and abundance of OTUs. The upper layer samples clustered in accordance with mud zones, whereas the deep layer samples of all sites tended to cluster together. To be specific, the upper layer samples (0, 1, 2, 3, and 5 cm) of each site clustered tightly with the exception that those at SYS02 were slightly scattered. Contrastingly, the deep layer samples (20 and 30 cm) of each site were more similar with each other. Interestingly, sediments at the 10-cm layer displayed different clustering relationships in different sites. At ECS02 and ECS03, the 10-cm layer resembled more closely the upper layer samples of the same site, whereas at SYS01, SYS02, and ECS01, they tended to group with the deep layer samples. The two-way ANOSIM revealed that sediment depth (global R = 0.835, p < 0.001) could explain more variances than mud zones (global R = 0.690, p < 0.001). RDA analysis was performed and uncovered that seven environmental factors had significantly influences, which jointly accounted for 70.2% of the total variation. TOC δ <sup>13</sup>C (F = 5.4, P = 0.001) contributed the most with 29.0%, followed by TON δ <sup>15</sup>N (F = 5.2, P = 0.001), TOC (F = 4.3, P = 0.001), PO<sup>4</sup> <sup>3</sup><sup>−</sup> (F = 3.1, P = 0.002), NH<sup>4</sup> <sup>+</sup> (F = 2.3, P = 0.01), TN (F = 2.3, P = 0.017), and C/N (F = 2.3, P = 0.023). No significant correlation was observed between NO<sup>3</sup> <sup>−</sup>, NO<sup>2</sup> −, or SiO<sup>3</sup> <sup>2</sup><sup>−</sup> and the communities. Influences of the top seven environmental factors on bacterial communities were shown in **Figure 5B**. NH<sup>4</sup> <sup>+</sup> seemed to exert significant impacts on structuring bacterial communities of ECS01. The up sediments of SYS01 appeared to be separated from other samples by TOC.

FIGURE 4 | Inter-taxa relation network with thresholds of ±0.7 for Spearman's coefficient and 0.01 for P-value. Correlations between families were represented by colored lines between nodes (red for positive, blue for negative). Size of node depended on the number of connections. A, Alteromonadaceae; D, Dehalococcoidia; γ, Gammaproteobacteria; J, JTB255; O, Oceanospirillaceae; P, Pseudomonadaceae; S, Streptococcaceae.

## DISCUSSION

#### Bacterial Abundance in Sediments of the ECMS and Potential Environmental Drivers

In this study, direct cell counting was implemented to evaluate the microbial abundance in sediments of the ECMS. According to Liu et al. (2015), the bacterial abundance was two to three orders of magnitude higher than the archaeal abundance in sediment of the ECMS. Thus, the counted cell numbers are approximately equal to the bacterial cell numbers. The cell counts were consistent with the result of 16S rRNA gene quantification. Sedimentary bacterial abundance in the ECMS was in the same range with that reported in the SYS (Liu et al., 2015), and was slightly lower than that in bioturbated coastal sediments from the Catalina Island (Bertics and Ziebis, 2009; Plotieau et al., 2013). By contrast, this abundance was higher than that in the Eastern Mediterranean Sea (Polymenakou et al., 2006) and in ODP sites from the Okinawa Trough and Peru margin (Mauclaire et al., 2004; Jean et al., 2005). Different environmental features may explain some of this variance in bacterial abundance across different areas. The bacterial abundance in sediments from the southern mud areas especially at site ECS01 and ECS02 was significantly higher (**Figure 1B**). These two sites are located just outside of the Yangtze River Estuary and Zhe-Min coasts and are readily influenced by terrigenous nutrient input from land, which could lead to the observed high bacterial abundance.

The abundance of bacteria steadily decreased with sediment depth, in well agreement with the global distribution pattern of benthic microbial abundance (Kallmeyer et al., 2012). Generally, aerobic bacterial respiration consumed DO rapidly in upper layers of the eutrophic sediment, which would result in reduced rate of microbial carbon oxidation in deep sediments and subsequent decrease in bacterial abundance (Røy et al., 2012). Availability of organic matter, as the main electron donor in marine sediments, may also affect bacterial abundance. In this study, the 16S rRNA gene abundance was positively correlated with TN (P < 0.01), negatively correlated with C/N (P < 0.01) and TON δ <sup>15</sup>N (P < 0.05). These results could partly contribute to the negative correlation between Shannon diversity and C/N, and revealed that the content, source and composition of organic matter are important in determining the abundance of benthic bacteria in the ECMS, with fresh and marine organic matter (higher TN and lower C/N ratio) supporting higher bacterial abundance.

The abundance of dsrB examined in this study was similar to that in the Pearl River estuary (Jiang et al., 2009), Bohai Sea and Yellow Sea (Liu et al., 2015), but lower than that in the Baltic Sea (Leloup et al., 2009) and Blake Sea (Leloup et al., 2007). As sulfate-reducing prokaryotes are anaerobic, the high level of DO in overlaying water of the ECMS sediments, compared with that in the Baltic Sea and Black Sea, may contribute to these variations. No significant differences in copy numbers of dsrB gene were observed among studied sites or depths. This was consistent with the invariable SO<sup>4</sup> <sup>2</sup><sup>−</sup> contents observed in the pore waters (Yu et al., 2016).

#### Distribution Patterns of Bacterial Community in Sediments of the ECMS and Potential Environmental Drivers

Limited studies of benthic microbial community in the ECMS were focused only on surface sediments. They showed that sediment sources, hydrodynamic conditions and concentration of nutrients might be the crucial factors in shaping bacterial communities (Liu et al., 2014; Xiong et al., 2014). To uncover the extent of influence of such effects on subsurface bacterial communities, the vertical profile of bacterial communities in mud sediments of the ECMS was sampled in this study. We found that bacterial communities of the up and deep sediment layers in the ECMS exhibited contrasting distribution patterns across sites.

Bacterial communities of the upper layers in each site were clearly separated (**Figure 5A**) and were found to be influenced by different environmental factors, such as TOC, NH<sup>4</sup> <sup>+</sup> and PO<sup>4</sup> <sup>3</sup><sup>−</sup> (**Figure 5B**). TOC separated the upper layer sediment of SYS01 from others, confirming the role of TOC as an important factor in shaping relative abundance of benthic bacterial groups (Jorgensen et al., 2012; Liu et al., 2014). NH<sup>4</sup> <sup>+</sup> distinguished the bacterial communities in ECS01 sediments especially the upper layer samples from others, confirming that sedimentary bacterial

community structures can be effected by nitrogen pollution (Xiong et al., 2014).

These distribution patterns are reflected by differential correlations between taxonomic groups and environment factors, in particular organic matter and nutrients (Supplementary Table S2). For example, Flavobacteriaceae, a major group of Bacteroidetes, showed a significantly positive relationship with TOC (r > 0.6 and P < 0.01), and this agreed with their chemoorganotrophic lifestyle functioning especially in degrading high molecular weight dissolved organic matter, such as polysaccharides (Bennke et al., 2016; Teeling et al., 2016). Planctomyces also preferred high concentration of nitrogen and organic carbon substrates. However, Nitrospirae preferred relatively oligotrophic environments as evidenced by the significantly negative relationship with TOC. Nitrospirae are nitrite oxidizing bacteria functioning in aerobic nitrite oxidization. However, they have been found to widely distribute in anaerobic marine sediments (Liu et al., 2014; Nunoura et al., 2016; Chen et al., 2017). Whether sedimentary Nitrospirae are inactive or have other uncharacterized physiologies needs further investigation. JTB255-MBG, Acidiferrobacter and BD2-11 terrestrial group (belonging to Gemmatimonadetes) preferred substrates with a low content of TON δ <sup>15</sup>N, while Syntrophobacterales was opposite. These results confirmed that varied sediment sources could provide different environmental niches for the growth of different bacterial communities.

Different from the scattered distribution of upper layer samples, the deep layer (20 and 30-cm layer) samples showed a closer clustering relationship regardless of studied sites (**Figure 5A**). TN and TOC seemed to play roles in converging the deep layer samples of different mud zones (**Figure 5B**). However, it was noteworthy that these environmental factors fluctuated more widely among sites than among depths of one site. Thus, they might be not the direct driving force that clustered the deep layer samples of studied sites. As mentioned above, these factors mainly influenced the relative abundance of up sediment-dominant Flavobacteriaceae, JTB255-MBG and Acidiferrobacter. These impacts would disappear in deep sediment layers with a lower abundance of up sediment-dominant taxa. Subsequently, the deep layer samples clustered closely. Indeed, DO under the top 1 µm sediments decreased sharply from ∼120–250 µM to an undetectable level in the study area (Yu et al., 2016). Thus, it might be inferred that DO, or redox state, was the crucial factor contributing to the differences between up and deep layer communities.

It was interesting to note that the 10 and 20-cm layer samples displayed different clustering patterns in each site. We speculated that this discrepancy might be attributed to site-specific hydrodynamic conditions, although no environment factor detected here could explain this difference. For example, ECS02 was influenced by the Taiwan Warm Current, resulting in a higher summer flow velocity in this site (Lim et al., 2007; Liu et al., 2007), while SYS02 was located in the bottom of the Yellow Sea Trough covered by the Yellow Sea Cold Water Mass, causing a lower flow velocity and deposition rate than other sites detected (Huh and Su, 1999; Yang et al., 2003; Chen Z. et al., 2004). These differences in fluid dynamics lead to varied particle size and redox profiles across different sites, thus influencing the cluster of 10 and 20-cm layer samples.

## Bacterial Community Compositions in Sediments of the ECMS

The sedimentary bacterial community composition in the ECMS was in high accordance with previous studies of the same area (Xiong et al., 2014; Liu et al., 2015) and other marginal seas (Zinger et al., 2011; Wang et al., 2012; Sun et al., 2013). In comparison with those from the deep sea or coastal areas adjacent to open oceans (Schauer et al., 2010; Dyksma et al., 2016; Walsh et al., 2016), a higher ratio of Deltaproteobacteria to Gammaproteobacteria was observed in all sediment depths of this study. Deltaproteobacteria and Gammaproteobacteria (average

48.1 and 41.2%, respectively) were the major predominant classes of Proteobacteria, and they have been demonstrated to play important roles in organic matter mineralization and dark carbon fixation, respectively, in coastal sediments (Thamdrup and Canfield, 1996; Dyksma et al., 2016). The high abundance of Deltaproteobacteria relative to Gammaproteobacteria may reflect specific response of marginal bacterial communities to terrigenous organic inputs, and indicate a higher potential of organic matter mineralization than carbon fixation in the ECMS sediments.

There were significant differences in diversity and composition of bacterial communities between up and deep sediment samples (Supplementary Figure S1 and **Figure 5A**). The upper layers owned a more diverse community than the deep layers. Species capable of thriving under aerobic and anaerobic environments can coexist at the shallow sediment, thus resulting in the high diversity. LEfSe analysis showed that Lactococcus and Pseudomonas of Gammaproteobacteria, and members of Dehalococcoidia (class of Chloroflexi) were enriched in the deep sediment samples (**Figure 3**). The former two genera were usually found in non-marine environments and played significant roles in spoilage of meat, dairy and fish (Williams et al., 1990; Champagne et al., 1994; Sakala et al., 2002). The proportions of these two genera were higher in several deep layer samples, which might be related to the general decrease in bacterial abundance with depth combined with their presence as contaminants in the extraction kit (Salter et al., 2014). However, sampling and DNA extraction methods in the present study were in accordance with methods used in previous studies on different samples (Nguyen and Landfald, 2015; Chen et al., 2017), in which case these potential contaminant genera were not detected. Thus, it is also possible that they might exist in deep layer sediments and feed on organics, such as remnants of marine animals. Dehalococcoidia were widely distributed in deep marine sediments (Durbin and Teske, 2011; Jorgensen et al., 2012). They exhibited significant correlations among each other and formed the deep-abundant module in the network analysis illustrating inter-taxa relationships (**Figure 4**). Different subgroups of Dehalococcoidia can inhabit different ecological niches (Bowman and McCuaig, 2003; Tas et al., 2010; Durbin and Teske, 2011; Wasmund et al., 2015), and they potentially own versatile ecological functions such as CO<sup>2</sup> fixation, dimethyl sulfoxide utilization, aromatics and fatty acids oxidization, and acetate production (Hug et al., 2013; Wasmund et al., 2014). The observed positive co-occurrence patterns within members of this class may suggest that these variable physiological features can be highly dependent and integrated, or help relieve interspecific competitions. The deep-abundant module was also involved in correlations among Dehalococcoidia and several other bacterial clades, including strictly anaerobic Spirochaetaceae, Sva0485 and low-oxygen-adapted candidate phylum OP8 (Bowman and McCuaig, 2003; Farag et al., 2014), Pseudomonadaceae and Streptococcaceae, implicating biogeochemical complexity in the deep marine sediments.

Comparatively, correlations among Acidobacteria, Bacteroidetes, Proteobacteria, and Lentisphaerae constituted the up-abundant module in the network, and the former three taxa were shown to have significantly higher proportions in the upper layers by LEfSe. JTB255-MBG, a member of order Xanthomonadales belonging to Gammaproteobacteria, was the most abundant clades (average 35%) in the upper layer sediments examined in this study. In the network, JTB255-MBG showed a high degree of connectivity with other surface-abundant taxa, indicating that growth of JTB255-MBG may be highly dependent on other bacterial clades, which may provide potential insights in developing new cultivating strategies for obtaining a pure isolate of this clade. Strains of Lentisphaerae were detected to produce transparent exopolymers (TEP) (Cho et al., 2004), a key factor of biofilm initiation and outgrowth (Berman and Passow, 2007). The involvement of Lentisphaerae and other marine biofilm residents, including Rhodopirellula, Oceanospirillaceae, Alteromonadaceae, Acidobacteria, Planctomyces, OM190, and Bacteroidetes (Bengtsson and Øvreås, 2010; Eichorst et al., 2011; Ruvindy et al., 2015; Lawes et al., 2016) in the up-abundant module indicated that biofilm may regulate bacterial interactions in the upper layer sediment.

## CONCLUSION

This study presented a detailed description of spatial and depth-related distribution patterns of bacterial communities in sediments of the ECMS. Abundance, diversity, and community structure varied significantly with sediment depth. The up and deep bacterial communities displayed different distribution patterns. The upper layer samples clustered in accordance with mud zones, whereas the deep layer samples of all sites tended to cluster together. TOC δ <sup>13</sup>C and TON δ <sup>15</sup>N significantly affected the bacterial community composition, suggesting that abundance and composition of organic matter played critical roles in shaping bacterial communities. Moreover, bacterial communities in the shallow and deep sediments showed different intertaxa relationships, indicating different co-occurrence patterns in surface and subsurface sediments. This study provided a detailed outline of subsurface bacterial communities in mud deposits of the ECMS for the first time, and provided clues for uncovering biogeochemical cycles in this area.

## AUTHOR CONTRIBUTIONS

YQ carried out sample collecting, laboratory work, data analysis, and drafted the manuscript. JL conceived the study, revised and finalized the manuscript. MZ and X-HZ participated in the design of the study and helped to draft the manuscript. All authors read and approved the final manuscript.

## FUNDING

This work was supported by the National Key Research and Development Program of China (No. 2016YFA0601303) and the National Natural Science Foundation of China through grants 41730530, 41521064, 41476112, and 41506154.

#### ACKNOWLEDGMENTS

fmicb-09-01128 May 30, 2018 Time: 8:52 # 10

We thank all of the scientists and crew members on the R/V Dong Fang Hong 2 during the expedition for their great efforts and help in sample collection.

#### REFERENCES


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.01128/full#supplementary-material


surface sediments. Environ. Microbiol. 19, 1584–1599. doi: 10.1111/1462-2920. 13676


fmicb-09-01128 May 30, 2018 Time: 8:52 # 11


**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 Qiao, Liu, Zhao and Zhang. 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.

# Sulfur-Oxidizing Bacteria Mediate Microbial Community Succession and Element Cycling in Launched Marine Sediment

Hideyuki Ihara<sup>1</sup> , Tomoyuki Hori <sup>2</sup> \*, Tomo Aoyagi <sup>2</sup> , Mitsuru Takasaki <sup>3</sup> and Yoko Katayama<sup>4</sup> \*

*<sup>1</sup> United Graduate School of Agricultural Science, Tokyo University of Agriculture and Technology, Fuchu, Japan, <sup>2</sup> Environmental Management Research Institute, National Institute of Advanced Industrial Science and Technology, Tsukuba, Japan, <sup>3</sup> Department of Food and Environmental Sciences, Faculty of Science and Engineering, Ishinomaki Senshu University, Ishinomaki, Japan, <sup>4</sup> Institute of Agriculture, Tokyo University of Agriculture and Technology, Fuchu, Japan*

A large amount of marine sediment was launched on land by the Great East Japan earthquake. Here, we employed both on-site and laboratory studies on the launched marine sediment to investigate the succession of microbial communities and its effects on geochemical properties of the sediment. Twenty-two-month on-site survey showed that microbial communities at the uppermost layer (0–2 mm depth) of the sediment changed significantly with time, whereas those at the deeper layer (20–40 mm depth) remained nearly unchanged and kept anaerobic microbial communities. Nine months after the incidence, various sulfur-oxidizing bacteria (SOB) prevailed in the uppermost layer, in which afterwards diverse chemoorganotrophic bacteria predominated. Geochemical analyses indicated that the concentration of metals other than Fe was lower in the uppermost layer than that in the deeper layer. Laboratory study was carried out by incubating the sediment for 57 days, and clearly indicated the dynamic transition of microbial communities in the uppermost layer exposed to atmosphere. SOB affiliated in the class Epsilonproteobacteria rapidly proliferated and dominated at the uppermost layer during the first 3 days, after that Fe(II)-oxidizing bacteria and chemoorganotrophic bacteria were sequentially dominant. Furthermore, the concentration of sulfate ion increased and the pH decreased. Consequently, SOB may have influenced the mobilization of heavy metals in the sediment by metal-bound sulfide oxidation and/or sediment acidification. These results demonstrate that SOB initiated the dynamic shift from the anaerobic to aerobic microbial communities, thereby playing a critical role in element cycling in the marine sediment.

Keywords: sulfur-oxidizing bacteria, launched marine sediment, microbial community, high-throughput sequencing, Epsilonproteobacteria

## INTRODUCTION

Coastal marine sediment governs the biogeochemical cycling of elements in the ocean, for instance as reservoirs of organic substances synthesized at the ocean surface (Middelburg et al., 1993) and of heavy metals (Morse and Luther, 1999). Marine sediment has diverse characteristics depending on both geographic features and human activities. On seafloor, depletion of dissolved oxygen

#### Edited by:

*Hongyue Dang, Xiamen University, China*

#### Reviewed by:

*Satoshi Ishii, University of Minnesota, USA Cuong Tu Ho, Institute of Environmental Technology, Vietnam*

> \*Correspondence: *Tomoyuki Hori hori-tomo@aist.go.jp Yoko Katayama katayama@cc.tuat.ac.jp*

#### Specialty section:

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

Received: *05 December 2016* Accepted: *20 January 2017* Published: *03 February 2017*

#### Citation:

*Ihara H, Hori T, Aoyagi T, Takasaki M and Katayama Y (2017) Sulfur-Oxidizing Bacteria Mediate Microbial Community Succession and Element Cycling in Launched Marine Sediment. Front. Microbiol. 8:152. doi: 10.3389/fmicb.2017.00152*

**228**

was induced by aerobic degradation of the accumulated organic matters (Holmer and Kristensen, 1992; Cloern, 2001). Subsequently, residual organic matters are degraded by anaerobes such as sulfate-reducing bacteria (SRB). The resultant hydrogen sulfide reacts with metals including heavy metals. The reduced sulfur compounds are then preserved in the sediment (Jørgensen and Fenchel, 1974; Jørgensen, 1977). On the other hand, sulfur-oxidizing bacteria (SOB) are phylogenetically diverse and prevail in the sulfide-rich environments (Lenk et al., 2011, 2012; Dyksma et al., 2016). The accumulated reduced metal sulfides can be oxidized in the presence of oxidants by SOB, leading the heavy metal mobilization that has a critical impact on marine ecosystems. Seitaj et al. (2015) reported the seasonal change of two types of filamentous SOB affiliated to the family Beggiatoaceae in the class Gammaproteobacteria and the family Desulfobulbaceae in the class Deltaproteobacteria. However, characteristics of SOB colonizing in the sediment and their diversity are still poorly understood.

The Great East Japan Earthquake, which was the most severe earthquake recorded in Japan, occurred in the Tohoku region on 11 March 2011; the accompanying huge tsunami caused serious damage in coastal areas (Mimura et al., 2011). In addition to giving high salt stress to soil environments (Asano et al., 2013), the tsunami transported a large amount of marine sediment onto land. Up to now, the sediment has been intensively investigated to reveal its relationship with the surrounding coastal marine sediment (Tanaka et al., 2012). In addition, the risk of the sediment contaminated with heavy metals has been addressed (Kawabe et al., 2012; Tsuchiya et al., 2012; Sera et al., 2014; Nakamura et al., 2016). Concerning geochemical properties of the sediment, it has been reported that ignition loss (IL), an indicator of organic matter content, ranged from 1.2 to 16.3% and the pH range was 1.1–9.6 (Ministry of the Environment, 2011). For heavy metals, the content of As has been found to account for 1.4–32.1 mg/kg-sediment (Sera et al., 2011). However, most of the studies involved only transient data, and described the spatial and geochemical differences in the sediment. Time-course changes in the microbial and geochemical properties of the sediment after the exposure to terrestrial environments, therefore, remain to be elucidated.

Although our recent study on the launched marine sediment incubated under anaerobic conditions showed nearly unchanged microbial communities in the presence of sulfate, ferric iron and CO<sup>2</sup> (Hori et al., 2014), and only amendment with nitrate facilitated the metabolic activities of anaerobic SOB in the classes Epsilonproteobacteria and Gammaproteobacteria (Aoyagi et al., 2015). Aerobic microbial activities are of considerable importance in the transformation of elements in the sediment because the surface of the sediment is always exposed to oxygen that is the highest energy-producing substrate for microbes. Nevertheless, very little is known about the structure and function of aerobic microbial communities in the sediment. In particular, SOB that use oxygen as electron donor are expected to play critical roles in the sediment because oxidation of sulfur compounds was the start of element cycle in the nitrate-supplemented incubation, while information on their physiological activities under oxic conditions is limited.

The objective of this study was to clarify microbial community succession in the launched marine sediment resulting from the exposure of the sediment to oxic conditions. To this end, we herein conducted deep sequencing of 16S rRNA genes that has provided the detailed characteristics of microbial communities (Caporaso et al., 2011; Itoh et al., 2014; Mahmoudi et al., 2015; Navarro et al., 2015). In addition to the on-site survey for 22 months, laboratory incubation of the sediment was employed to examine the short-term microbial community dynamics and the changes in geochemical properties of the sediment. Monitoring at short intervals with a special focus on the exposure to atmosphere clarified the remarkable microbial community succession mediated by SOB and their involvement in the succeeding element cycling.

## MATERIALS AND METHODS

#### Sampling of Launched Marine Sediment

Launched marine sediments by the Great East Japan Earthquake were collected at Higashi-matsushima, Miyagi, Japan (Table S1 and Figures S1–S3; 38◦ 25′ 49′′N, 141◦ 14′ 39′′E). Color and texture of the uppermost layer (0–2 mm depth) were reddish brown and slightly dried, whereas those of the deeper layer were black and moist. Due to the visual appearances under the environmental conditions, it was assumed that the uppermost layer was oxic and the deep layer was anoxic. The sediments were sampled from 0 to 200–300 mm depth using a spade or a core sampler, transported to the laboratory under cool, and then separated vertically into the uppermost (0–2 mm depth) and deep (20–40 mm depth) layers. These sediment samples were stored at −80◦C. To select sampling date for the main examination, prior analysis of the 16S rRNA gene deep sequencing as mentioned below was performed in singlicate. Details of the prior analysis are shown in the supporting information (Figure S3). Based on the results of the analysis, we decided to use the sediments collected in December 2011, March 2012 and October 2013 for the examination of on-site changes in microbial communities and geophysical characteristics. The sediment collected in November 2012 was kept for around half a year at 4◦C before conducting laboratory incubation. No significant change in microbial communities during the storage period was checked by the deep sequencing of 16S rRNA genes.

## Laboratory Incubation of the Sediment

The brock-state sediment was stored in a thick polyethylene bag to keep the humidity. After removing the air-exposed surface of the sediment, the inside part of the sediment remained under anoxic conditions was obtained for laboratory incubation. After thorough mixing, the sediment (approximately 200 g) was placed in a polyethylene terephthalate container (reverse truncated cone: top, 12.9 cm dia.; bottom 9 cm dia.; height 6.5 cm) to a depth of around 30 mm. These procedures were conducted in a 100- L polyvinyl fluoride bag that was filled with nitrogen gas to minimize exposure of the sediment to air. The containers with the sediment were then placed in a 27-L chamber containing water-soaked cotton to maintain 70–100% relative humidity, and incubated in the atmosphere in dark at 20–25◦C for 57 days. Humidity and temperature were monitored with a data logger (Ondotori TR-72U; T&D, Nagano, Japan). Two containers were sampled destructively from on days 1, 3, 7, 14, 28, and 57. Thus, a total of 12 containers (6 dates and 2 replication) were prepared in this experiment. The 0–2 mm depth (uppermost) and 12– 16 mm depth (deep) layers of the sediments were collected in duplicate from each container. Consequently, the quadruplicate samples were used for subsequent analyses. The day 0 samples were collected in quadruplicate at the beginning of the experiment.

#### Geochemical Analysis of the Sediment

Geochemical properties measured in the uppermost layer of the on-site sediment were pH, concentrations of sulfate and metals (Na, Mg, Al, K, Ca, Fe, Cr, Cu, As, Se, Cd, and Pb). While the properties measured in the deep layer were ignition loss (IL), concentrations of ions (Na+, K+, Mg2+, Ca2+, Cl−, and SO2<sup>−</sup> 4 ) and metals. For pH analysis, the sediment was suspended in ultrapure water at a ratio of 1:2.5 (w/w) and then the suspension was vortexed. Following the centrifugation at 21,500 × g for 1 min at 4◦C, pH of the supernatant was measured with a pH electrode (pH Meter M-12; Horiba, Kyoto, Japan). To determine IL, the sediment was dried at 100◦C until the weight became constant, and then heated at 600◦C for 2 h. For measurement of ion concentrations, 0.03–1.3 g of sediment was suspended with 10 mL of ultra-pure water and shaken for 30 min at 4◦C. After centrifugation at 250 × g for 5 min at 4◦C, the supernatant was diluted with ultra-pure water, and filtered through a cellulose acetate filter (0.2 µm pore size). The resultant samples were analyzed by an ion chromatograph (883 Basic IC Plus; Metrohm Japan Ltd., Tokyo, Japan) equipped with a Metrosep A Supp 4 column (250 × 4 mm) and a Metrosep A Supp 4/5 guard column (Metrohm Japan Ltd.) for anions, and an ion chromatograph (861 Advanced Compact IC; Metrohm Japan Ltd.) equipped with an IC YS-50 column (4.6 × 125 mm) and an IC YS-G guard column (Showa Denko, Tokyo, Japan) for cations. Detailed method for metal analysis is shown in the Supplementary Information.

During the laboratory incubation of the sediment, IL, sulfate ion concentration, pH, total carbon (TC), total nitrogen (TN), dissolved organic carbon (DOC), and dissolved nitrogen (DN) were determined. The sediment was suspended in ultra-pure water at a ratio of 1:10 (w/w) for measurement of sulfate ion concentration and at a ratio of 1:2.5 (w/w) for measurement of pH. After shaking for 30 min and centrifugation at 250 × g for 5 min at 4◦C, sulfate ion concentration was measured with the ion chromatograph as described above. pH of the supernatant was measured with the pH electrode.

TC and TN of the dry sediment were measured with a carbon-nitrogen analyzer (MT-700; Yanako, Kyoto, Japan). For measurement of DOC and DN, the sediment was suspended in ultra-pure water at a ratio of 1:50 (w/w) and the solution was shaken at 4◦C for 1 h. After centrifugation at 250 × g for 5 min at 4◦C, the supernatant was filtered through a cellulose acetate filter (0.2 µm pore size) and DOC and DN of the supernatant were measured using a total organic carbon analyzer (TOC-VE; Shimadzu, Kyoto, Japan) connected to a total nitrogen measuring unit (TNM-1; Shimadzu).

#### Extraction of DNA from the Sediment, Polymerase Chain Reaction (PCR) Amplification, and Deep Sequencing of 16S rRNA Genes

DNA was extracted from the sediment in triplicate according to the bead-beating method described by Noll et al. (2005) with some modifications: 10–20 mg of autoclaved skim milk was added to 100–500 mg of the sediment before bead beating to improve the DNA recovery, and isopropyl alcohol was used instead of ethanol for precipitation of DNA (crude extract/isopropyl alcohol/3 M sodium acetate was 10:9:1 [v/v/v]). Subsequently, RNA in the crude DNA extract was removed with RNase A (Nippon Gene, Tokyo, Japan). DNA concentration was determined spectrophotometrically (NanoDrop 2000; Thermo Scientific, Kanagawa, Japan). Eighteen and 39 DNA extracts from the on-site and incubated sediments, respectively, were utilized for the construction of deep sequencing libraries.

PCR targeting on the V4 region of 16S rRNA genes was conducted with the primer set 515F (5′ - GTGCCAGCMGCCGCGGTAA-3′ )/806R(5′ -GGACTACHVG GGTWTCTAA-T-3′ ) attached to sequences for the adapter region. The reverse primer was encoded with 12-bp barcodes for multiplex sequencing (Caporaso et al., 2012). The PCR mixture included 10 µl of 5 × Q5 buffer, 1 µl of 2.5 mM dNTP, 2 µl of each of 10 pM 515F and 806R primers, 0.5 µl of DNA polymerase (Q5; NEB, Tokyo, Japan), 10 ng of template DNA, and sterile ultra-pure water for a final volume of 50 µl. PCR amplification was performed as follows: initial denaturation at 98◦C for 1.5 min, and then 20 or 30 cycles consisting of denaturation at 98◦C for 10 s, annealing at 55◦C for 30 s, and extension at 72◦C for 30 s, followed by final extension at 72◦C for 2 min. The accuracy of amplification was confirmed by electrophoresis on 1.2% agarose gel. The concentration of PCR products was similar in spite of the different cycle numbers applied, which imply no or little influence of the cycle numbers on the results of the subsequent deep sequencing.

Purification of PCR products prior to deep sequencing of 16S rRNA genes was performed as described by Hori et al. (2014). The prepared DNA segments were subjected to pairedend sequencing with a 300-cycle MiSeq reagent kit (Illumina, Tokyo, Japan), and then a MiSeq sequencer (Illumina). The obtained sequences were aligned using a Burrows-Wheeler Aligner ver 0.5.9. and filtered by quality value 30 (Q30) by command lines in the software QIIME ver 1.7.0. (Caporaso et al., 2010). Chimeric sequences were removed by using the Mothur software (Schloss et al., 2009). The software QIIME was used for phylogenetic classification of operational taxonomic units (OTUs) with a cut-off value of 97% similarity. Using this software, the α-diversity indices and the weighted UniFrac distances for principal coordinate analysis (PCoA) were calculated. Some of the predominant OTUs were compared to sequences deposited in the database of the DNA Data Bank of Japan (DDBJ) using the Basic Local Alignment Search Tool (BLAST) to determine their closest relatives. The sequence data obtained in this study have been deposited in the DDBJ database under accession numbers DRA004739 and DRA004740.

## Quantitative PCR (qPCR) of the Incubated Sediment

To measure the number of copies of 16S rRNA genes in the sediment incubated in the laboratory, qPCR was conducted in duplicate using Premix Ex Taq II (Takara Bio Inc., Shiga, Japan) and a Thermal Cycler Dice Real Time System II (Takara Bio Inc.). The 515F/806R primer set was used and the mixture was prepared according to the manufacturer's instructions. PCR amplification was performed as follows: initial denaturation at 95◦C for 30 s, and then 45 cycles consisting of 95◦C for 5 s, and 61◦C for 30 s. A standard curve was constructed using PCR products from the 16S rRNA gene from Escherichia coli with the primer set 27F (5′ -AGAGTTTGATCCTGGCTCAG-3′ )/1525R (5′ -AAAGGAGGTGATCCAGCC-3′ ).

## RESULTS

#### Microbial Communities in the On-Site Sediment

Deep sequencing of 16S rRNA genes was carried out to investigate microbial communities in the on-site sediment. The total number of sequences obtained from 18 sediment samples was around 7.2 hundred thousand, corresponding to an average of 39,775 sequences per library (Table S2). The α-diversity indices (i.e., Chao1, Shannon, and Simpson reciprocal) were calculated by using an equal number of sequences (30,789) subsampled 10 times from original libraries. These values were lower in the uppermost layer than in the deep layer, indicating that the uppermost layer had more specified and less diverse microbial communities than those in the deep layer.

PCoA illustrated that microbial communities in the uppermost layers of the sediments changed drastically during the monitored period (Figure S4). Phylogenetic information of the entire structures and predominant OTUs is shown in **Figure 1** and Table S3. **Figure 1** shows that the phylum Proteobacteria dominated in both the uppermost and deep layers, which accounted for 42.0–72.4% and 29.9–42.2% of the relative abundance, respectively. The class Gammaproteobacteria was predominant in the uppermost layer (relative abundances: 10.9– 42.0%), and analysis at the major order showed the clear bacterial succession depending on the sampling date. More specifically, the order Thiotrichales was predominant in December 2011 (10.5%), whereas the order Xanthomonadales became dominant in October 2013 (37.7%) (**Figure 1B**). It is worth noting that the dominant constituent of Thiotrichales detected in the sediment was only SOB belonging in the genus Thiomicrospira (Table S3). With respect to other SOB, the genus Sulfurimonas in the class Epsilonproteobacteria was dominant in December 2011 (**Figure 1C**). Also, Pandoraea thiooxydans (OTU 1598) in the class Betaproteobacteria accounted for 12.4% in the same time (Table S3). These results indicate that SOB was present and may have performed sulfur oxidation in the uppermost layer of the sediment. In October 2013, chemoorganotrophic bacteria in the order Xanthomonadales and the phylum Actinobacteria became dominant in the uppermost layer (**Figures 1A,B**). Organic compounds including carbon products of SOB would serve as substrate for the chemoorganotrophs.

In contrast, PCoA and phylogenetic analysis showed microbial communities in the deep layers remained nearly unchanged over 22 months (Figure S4 and **Figure 1**). The class Deltaproteobacteria was dominant (19.0–25.9%) and mainly comprised the three orders (i.e., Desulfobacterales, Desulfuromonadales and Syntrophobacterales) (**Figure 1D**). These taxa are known to include obligate anaerobic SRB, implying that the sulfate reduction was retained under the presumably anoxic conditions of the deep layer, which is in accordance with findings obtained in our previous studies (Hori et al., 2014; Aoyagi et al., 2015).

#### Geochemical Properties of the On-Site Sediment

Geochemical analyses were conducted to characterize chemical components of the sediment and their time-dependent changes under oxic conditions. IL and ion concentrations of the deep layer were consistently high, indicating that the sediment exhibited the high accumulation of organic matters and the salinity, and these levels were kept for at least the period monitored around 22 months (**Table 1**). The most abundant metal in the deep layer was Al, followed by Fe (**Table 2**). Metals, such as Na, Mg, K, and Ca that are common in natural environments, were also found in the deep layer. Concentrations of metals other than Fe were apparently lower in the uppermost layer than those in the deep layer. While pH of the uppermost layer in March 2012 was neutral (pH 7.1), that in December 2011 was acidic (pH 4.3) (Table S1). The acidification of the uppermost layer may have facilitated the metal mobilization, resulting in the low concentration of the metals. High concentrations of sulfate in October 2013 implied that sulfur oxidation occurred in the uppermost layer. Although concentrations of heavy metals such as Cu, As, Cd and Pb were high in the sediment compared to those in soils (Iimura, 1981), their concentrations were sufficiently below the environmental standard values in the Soil Contamination Countermeasure Act of Japan (http://www.env. go.jp/en/water/soil/contami\_cm.pdf).

#### Succession of the Sediment Microbial Communities during Laboratory Incubation

The on-site sediment was influenced by various environmental factors such as air exposure, insolation and rainfall, and it makes difficult to evaluate the relationship between the succession of microbial communities and the surrounding environment conditions. Thus, laboratory incubation was conducted to monitor the microbial responses to environmental changes more concretely under controlled conditions. We focused on the exposure to atmosphere because the microbial metabolism, such as chemoorganotrophic and chemolithotrophic transformation, are highly affected by redox conditions.

Results from qPCR showed that there was no significant difference in the copy number of 16S rRNA genes between the uppermost and deep layers (Table S4). The total number of sequences obtained from 39 sediment samples was around 2.2

FIGURE 1 | Microbial community structures in the uppermost (0–2 mm depth) and deep (20–40 mm depth) layers of the on-site sediments based on the 16S rRNA gene analysis (n = 3). The bars indicate average values of three replications. Sediment samples were collected in December 2011, March 2012 and October 2013. (A) Microbial communities are categorized by phylum except for Proteobacteria that is shown by class. The fraction of the dominant phylotypes (>3% of each library) in the classes Gammproteobacteria (B), Epsilonproteobacteria (C), and Deltaproteobacteria (D) are shown in the histograms.



*<sup>a</sup>The deep (20–40 mm depth) layer sediments were used for the analysis. The symbol "*±*" means the standard deviation of three replications. There was the significant difference in Cl*<sup>−</sup> *between the sediment in March 2012 and October 2013 (p* < *0.05).*


TABLE 2 | Changes of metal concentrations in the on-site sediment.

*<sup>a</sup>Sediment sample was collected at 0–2 mm depth from the surface. The measurement was conducted in singlicate because the quantity of obtained sample was small. <sup>b</sup>Sediment samples were collected at 20–40 mm depth from the surface (n* = *2) and the symbol "*±*" means the variation between two replications.*

million, corresponding to an average of 56,403 sequences per library. The α-diversity indices were calculated by using an equal number of sequences (31,950) subsampled 10 times from original libraries. The values were lower in the uppermost layer than in the deep layer, corresponding with the on-site survey that indicated low microbial diversity in the uppermost layer (Table S2). PCoA showed notable shifts in the uppermost-layer microbial communities (**Figure 2**), strongly suggesting that the exposure to the atmosphere immediately altered the physiological properties of microbes in the uppermost layer.

**Figure 3** shows the succession of microbial communities of the sediment during the incubation, and the most predominant 7 OTUs in the uppermost layer at each sampling date are summarized in **Table 3** and Table S5. Microbial communities in the uppermost layer changed considerably with incubation time. The relative abundance of the class Epsilonproteobacteria increased dramatically from 7.5% at day 0 to 61.5% at day 3 (**Figure 3B**). The family Helicobacteraceae was the most dominant taxon found in this class, and comprised the genera Sulfuricurvum and Sulfurimonas that are known as important SOB in marine sediment (Kodama and Watanabe, 2004). These dramatic succession from anaerobic chemoorganotrophic bacteria to SOB in the microbial communities strongly suggest the importance of sulfur oxidation processes in the launched marine sediment under oxic conditions. Growth of some SOB (e.g., Sulfurovum lithotrophicum) that did not prevail under nitrate-reducing conditions in the previous study (Aoyagi et al., 2015) was enhanced under oxic conditions in this study. The rapid proliferation of the class Epsilonproteobacteria was followed by increases in the classes Zetaproteobacteria and Betaproteobacteria (**Figure 3A**). Phylogenetic analysis at the OTU level showed the predominance of OTUs closely related to Mariprofundus ferrooxydans (EF493244) and Gallionella sp. (HQ117915), both of which exhibit Fe(II)-oxidizing activity (Hallbeck and Pedersen, 2005; Emerson et al., 2007). This implicates that ferrous iron oxidation occurred subsequent to, or in parallel with, the

FIGURE 2 | Comparison of microbial community structures in the uppermost (0–2 mm depth, red) and deep (12–16 mm depth, blue) layers of the sediments incubated in laboratory based on principal coordinate analysis (PCoA) (n = 3). These plots were calculated from an equal number of sequences (31 950) by weighted UniFrac analysis. , before incubation (Day 0); ◦, Day 1; 1, Day 3; , Day 7; ×, Day 14; +, Day 28; –, Day 57. Arrows indicate the trajectory of the community structure change in the uppermost layer.

sulfur oxidation by SOB. At the end of the incubation, the class Gammaproteobacteria and the phylum Actinobacteria became dominant. Because the family Streptomycetaceae in the Actinobacteria and the orders Xanthomonadales and Methylococcales in the Gammaproteobacteria are known to exhibit chemoorganotrophy (Bowman, 2005; Saddler and Bradbury, 2005; Kämpfer, 2012), it is considered that these bacteria became metabolically active in the uppermost layer after the proliferation of chemolithotrophic bacteria (i.e., SOB and Fe(II)-oxidizing bacteria [FeOB]).

A variety of SOB became dominant according to the time of the incubation. Specifically, Sulfurovum lithotrophicum (OTU 43060), Sulfurovum aggregans related species (OTU

Proteobacteria that is shown by class. The fraction of the dominant phylotypes (>3% of each library) in the classes Epsilonproteobacteria (B) and

25387), Sulfurimonas denitrificans (OTU 30483), Thiomicrospira psychrophila (OTU 21731) and Thiomicrospira crunogena (OTU 36501) were predominant at days 1–3, whereas Thioalkalispira microaerophila (OTU 45161) and Thiohalophilus thiocyanatoxydans related species (OTUs 16111 and 45198) increased after day 14 (**Table 3**). The successive dominance of SOB suggests that sulfur oxidation have an advantage over chemoorganotrophy in the organic compounds- and sulfides-rich sediment during the incubation.

Gammaproteobacteria (C) are shown in the histograms.

Microbial community structures in the deep layers during the incubation were quite similar each other, therefore, the representative data at day 57 is presented in **Figure 3**. The microbial communities consisted mainly of the class Deltaproteobacteria and the phylum Chloroflexi, which is consistent with the microbial communities in the deep layers of the on-site survey (**Figure 1**).

## Change in Geochemical Properties of the Sediment during Laboratory Incubation

Water content of the sediment was in the range of 55.9–58.8% during the laboratory incubation. The concentration of sulfate ion in the uppermost layer increased considerably from 4075 to 9219 mg/kg dry weight (dw), and pH decreased from 7.2 to 4.7 (**Figure 4**), indicating the sulfate formed by SOB acidified the sediment. Rate of the sulfate accumulation can be divided into two stages: faster rates during the first week (about 377 mg/kg dw/day) and slower rates in the succeeding period (about 50 mg/kg dw/day). In particular, the sulfate-accumulating rate in the first 3 days reached a maximum value of 705 mg/kg dw/day (i.e., 300 mg/kg ww/day). The IL in the uppermost and deep layers did not differ between the beginning and end of the incubation (Figure S5A). TC content significantly increased in the uppermost layer and the value reached 20.3 g/kg dw at the



*<sup>a</sup>The closely related species were assigned on BLAST in the DDBJ.*

*<sup>b</sup>The OTUs were characterized phylogenetically by using the QIIME software.*

*<sup>c</sup>The symbol "*±*" means the standard deviation of three replications.*

*<sup>d</sup>p-values indicate whether the relative abundance of OTU was significantly high comparing with that in the deep layer:* \**p* < *0.05,* \*\**p* < *0.01.*

*<sup>e</sup>The putative function of closely related species (only sequence similarities* >*95%). SO, sulfur oxidation; SR, sulfate reduction; FeO, Fe(II) oxidation; ChemO, chemoorganotrophy; NR, nitrate reduction.*

end of the incubation, whereas TN content showed no significant change (Figures S5B,C). In contrast, the DOC concentrations in both the uppermost and deep layers and the DN concentration in the uppermost layer exhibited the significant decreases (Figures S5D,E). The decreasing rate of DOC was notably higher in the uppermost layer than in the deep layer. Especially, the significant decrease in DOC in the uppermost layer was possibly due to the enhanced activities of chemoorganotrophs by the exposure to atmosphere.

#### DISCUSSION

SOB in the class Epsilonproteobacteria were predominant during the early phase of the laboratory incubation (days 1–3), whereas SOB in the class Gammaproteobacteria were predominant during the latter phase (**Figure 3** and **Table 3**). These differences might be explained by the distinct metabolic strategies of sulfur oxidation in these SOB. The phylum Proteobacteria is known to have several pathways for sulfur oxidation. The Gammaproteobacteria has an energy-producing pathway that is kinetically advantageous if oxygen and reduced sulfur compounds are steadily supplied, while the Epsilonproteobacteria has versatile energy-producing pathways to adapt to transient environmental conditions (Yamamoto and Takai, 2011). Thus, it is plausible that the Epsilonproteobacteria dominated at the earlier stage of the incubation because of their flexibility to environmental changes.

Our previous study showed that Sulfurimonas denitrificans was the SOB dominated during the incubation of the launched marine sediment under nitrate-reducing conditions (Aoyagi et al., 2015). On the other hand, exposure of the sediment to the oxic conditions resulted in the proliferation of more diverse dominant SOB than those under nitrate-reducing conditions (**Table 3**). Sulfate accumulation rate in the previous study (1800 mg/kg wet weight (ww)/day) was almost 6 times faster than the present one (300 mg/kg ww/day), presumably due to the difference of the experimental conditions: in the previous study, sediment was suspended as slurry and anaerobically pre-incubated for 1-month before addition of nitrate. Out of the predominant OTUs found in the uppermost layer of the incubated sediment in this study, 8 OTUs (OTU 43060, 25387, 21731, 36501, 30483, 45161, 16111, and 45198) were phylogenetically related to SOB. In particular, Sulfurimonas autotrophica, Sulfurovum lithotrophicum, Sulfurovum aggregans and Thiomicrospira crunogena have been isolated from deepsea sediments and/or hydrothermal vents (Jannasch et al., 1985; Inagaki et al., 2003, 2004; Mino et al., 2014). Other two dominant OTUs 32337 and 6816 in the uppermost layer were phylogenetically related to FeOB. The related species M. ferrooxydans has been isolated from hydrothermal vents and they have been known as important players in ecological iron cycling (Emerson et al., 2007; Hoshino et al., 2016). Thus, it is likely that the launched marine sediment examined harbor SOB and FeOB, both of which have been found in these aquatic ecosystems.

Dramatic environmental changes of the launched marine sediment, particularly the exposure to atmosphere, may cause strong effects geochemically and biologically. For example, the on-site sediment surface colored reddish brown (Figure S2), resulting from the formation of iron precipitates. The on-site detection of SOB and FeOB suggested that metabolic activities of these bacteria were related to the direct and/or indirect formation of iron precipitate because of the close relationship between the iron and sulfur cycling (Jørgensen and Fenchel, 1974; Jørgensen, 1977; Hsieh and Yang, 1989; Schippers and Jørgensen, 2002). Indeed, the laboratory incubation of the sediment showed the rapid increase and decrease of FeOB-related OTUs (**Figure 3** and **Table 3**).

Heavy metals are generally preserved as metal sulfides in coastal marine sediments due to hydrogen sulfide produced by SRB (Jørgensen, 1977; Zhang et al., 2014). Relatively high metal concentrations of the launched sediment were comparable with those found in marine sediment in the Ishinomaki bay that is near the sampling site (**Table 2**, Imai et al., 2004). The uppermost layer exhibited apparently lower concentrations of the metals than the deep layer, which suggests that the heavy-metal mobilization was facilitated by the natural weathering of the sediment in the terrestrial environment. Both the on-site and laboratory studies demonstrated the dramatic proliferation of SOB under oxic conditions in the sediment (**Figures 1**, **3**). The sulfur oxidation could be directly linked to the release of heavy metals from metal sulfides, as reported previously (Gadd, 2000, 2004; Sand et al., 2001; Stephens et al., 2001). Moreover, the production of sulfate from sulfur oxidation resulted in the acidification of the sediment (**Figure 4**). The leaching of metals at low pH has been reported previously (Evans, 1989; Masscheleyn et al., 1991; Calmano et al., 1993; Bowell, 1994). Thus, SOB might cause the mobilization of heavy metals via the direct and indirect procedures in the

sediment. The time-dependent changes in sulfur compounds and heavy metals in the sediment will be necessary to clarify the involvement of SOB in these processes.

Although the launched marine sediment was rich in organic matters, chemoorganotrophic bacteria became dominant after the proliferation of chemolithotrophic SOB and FeOB (**Figure 3** and **Table 3**). Most of the organic substances in marine sediment have been considered as being relatively persistent (Kristensen et al., 1995; Kristensen, 2000), suggesting the organic substances available for chemoorganotrophic bacteria was limited in this study. In fact, no obvious decreases in IL, TC and TN were observed during the laboratory incubation (Figures S5A–C). The increase in TC in the uppermost layer over the time course of the incubation suggests that SOB and FeOB fixed CO<sup>2</sup> as a carbon source and biosynthesized organic substances (Figure S5B). SOB might facilitate the growth of chemoorganotrophic bacteria by supplementing the easily degradable organic substances for them. Because the accumulation of organic matters on the seafloor can adversely affect the ecosystem, it has long been a challenge to stimulate the degradation of organically enriched marine sediment by biological and chemical procedures (Vezzulli et al., 2004; Kunihiro et al., 2008; Wada et al., 2008; Yamamoto et al., 2008; Laverock et al., 2010).

We herein clarified the succession of microbial communities in the launched marine sediment by combining the longterm on-site survey and short-term laboratory incubation. Although the laboratory incubation was unable to recreate the on-site environment in its entirety, it provided the important information about the change of microbial communities due to the exposure to atmosphere. A variety of SOB, especially the class Epsilonproteobacteria, rapidly proliferated and induced the subsequent growth of FeOB and chemoorganotrophic bacteria. Furthermore, the metabolically activated SOB possibly contributed to the mobilization of heavy metals that bound to the sediment. Consequently, the epsilonproteobacterial SOB initiated the dynamic shift from the anaerobic to aerobic microbial communities, which play a pivotal role in element cycling in the marine sediment.

#### AUTHOR CONTRIBUTIONS

HI: Main worker in this paper. TH and TA: Contribution for DNA analysis and discussion. MT: Contribution for sampling and discussion. YK: Supervisor of the first author, and contribution for discussion.

#### REFERENCES


#### ACKNOWLEDGMENTS

We thank Izumi Watanabe in Tokyo University of Agriculture and Technology for support of metal analysis.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.00152/full#supplementary-material


isolated from an underground crude-oil storage cavity. Int. J. Syst. Evol. Microbiol. 54, 2297–2300. doi: 10.1099/ijs.0.63243-0


**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 © 2017 Ihara, Hori, Aoyagi, Takasaki and Katayama. 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.

# Novel Pelagic Iron-Oxidizing Zetaproteobacteria from the Chesapeake Bay Oxic–Anoxic Transition Zone

#### Beverly K. Chiu<sup>1</sup> , Shingo Kato<sup>2</sup> , Sean M. McAllister<sup>3</sup> , Erin K. Field<sup>4</sup> and Clara S. Chan1,3 \*

<sup>1</sup> Department of Geological Sciences, University of Delaware, Newark, DE, United States, <sup>2</sup> Project Team for Development of New-Generation Research Protocol for Submarine Resources, Japan Agency for Marine-Earth Science and Technology, Kanagawa, Japan, <sup>3</sup> School of Marine Science and Policy, University of Delaware, Newark, DE, United States, <sup>4</sup> Department of Biology, East Carolina University, Greenville, NC, United States

Chemolithotrophic iron-oxidizing bacteria (FeOB) could theoretically inhabit any environment where Fe(II) and O<sup>2</sup> (or nitrate) coexist. Until recently, marine Fe-oxidizing Zetaproteobacteria had primarily been observed in benthic and subsurface settings, but not redox-stratified water columns. This may be due to the challenges that a pelagic lifestyle would pose for Zetaproteobacteria, given low Fe(II) concentrations in modern marine waters and the possibility that Fe oxyhydroxide biominerals could cause cells to sink. However, we recently cultivated Zetaproteobacteria from the Chesapeake Bay oxic–anoxic transition zone, suggesting that they can survive and contribute to biogeochemical cycling in a stratified estuary. Here we describe the isolation, characterization, and genomes of two new species, Mariprofundus aestuarium CP-5 and Mariprofundus ferrinatatus CP-8, which are the first Zetaproteobacteria isolates from a pelagic environment. We looked for adaptations enabling strains CP-5 and CP-8 to overcome the challenges of living in a low Fe redoxcline with frequent O<sup>2</sup> fluctuations due to tidal mixing. We found that the CP strains produce distinctive dreadlock-like Fe oxyhydroxide structures that are easily shed, which would help cells maintain suspension in the water column. These oxides are by-products of Fe(II) oxidation, likely catalyzed by the putative Fe(II) oxidase encoded by the cyc2 gene, present in both CP-5 and CP-8 genomes; the consistent presence of cyc2 in all microaerophilic FeOB and other FeOB genomes supports its putative role in Fe(II) oxidation. The CP strains also have two gene clusters associated with biofilm formation (Wsp system and the Widespread Colonization Island) that are absent or rare in other Zetaproteobacteria. We propose that biofilm formation enables the CP strains to attach to FeS particles and form flocs, an advantageous strategy for scavenging Fe(II) and developing low [O2] microenvironments within more oxygenated waters. However, the CP strains appear to be adapted to somewhat higher concentrations of O2, as indicated by the presence of genes encoding aa3-type cytochrome c oxidases, but not the cbb3-type found in all other Zetaproteobacteria isolate genomes. Overall, our results reveal adaptations for life in a physically dynamic, low Fe(II) water column, suggesting that niche-specific strategies can enable Zetaproteobacteria to live in any environment with Fe(II).

#### Edited by:

Stefan M. Sievert, Woods Hole Oceanographic Institution, United States

#### Reviewed by:

Esther Singer, Joint Genome Institute, United States Beth Orcutt, Bigelow Laboratory for Ocean Sciences, United States

> \*Correspondence: Clara S. Chan cschan@udel.edu

#### Specialty section:

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

Received: 27 March 2017 Accepted: 26 June 2017 Published: 18 July 2017

#### Citation:

Chiu BK, Kato S, McAllister SM, Field EK and Chan CS (2017) Novel Pelagic Iron-Oxidizing Zetaproteobacteria from the Chesapeake Bay Oxic–Anoxic Transition Zone. Front. Microbiol. 8:1280. doi: 10.3389/fmicb.2017.01280

Keywords: iron-oxidizing bacteria, Zetaproteobacteria, biominerals, iron oxides, biofilm, pelagic bacteria

## INTRODUCTION

fmicb-08-01280 July 14, 2017 Time: 15:38 # 2

Chemolithotrophic Fe-oxidizing bacteria (FeOB) use Fe(II) oxidation for energy and growth, and are therefore thought to play important roles in Fe cycling. Fe is practically ubiquitous, raising the question of whether FeOB are active in every environment with Fe redox cycling, which would likely require a variety of niche-specific adaptations. Fe cycling is particularly important at coasts, where Fe transformations affect the chemistry of waters in coastal sediments and estuaries, and ultimately the concentrations of nutrients (e.g., Fe, P) and other metals (e.g., As) transported to the ocean (Charette et al., 2005; Jung et al., 2009). Significant redox activity occurs in stratified marine waters, such as the Chesapeake Bay, which experience seasonal anoxia in bottom waters (Officer et al., 1984). In our previous studies of the Chesapeake, water samples from the oxic–anoxic transition zone always yielded enrichments of chemolithotrophic FeOB (MacDonald et al., 2014; Field et al., 2016). From these enrichments, we isolated two FeOB, which represent the first known marine FeOB from the water column (isolate strain CP-8 previously reported in Field et al., 2016). The presence of FeOB was somewhat surprising given the relatively low (micromolar) concentrations of Fe, and the strong tidal mixing, which may frequently expose FeOB to higher O<sup>2</sup> concentrations, making it harder for them to compete with abiotic Fe(II) oxidation. Further study of these isolates may reveal their distinct adaptations to life in the estuarine water column, while also showing commonalities shared among all marine FeOB across different environments.

The Chesapeake FeOB isolates are members of the Zetaproteobacteria, all of which are marine neutrophilic chemolithotrophic FeOB. The other Zetaproteobacteria isolated to date primarily originate from deep sea hydrothermal microbial mats and sediments (Emerson et al., 2007; McAllister et al., 2011; Field et al., 2015; Makita et al., 2016), with some from coastal sediment (Laufer et al., 2016, 2017). Zetaproteobacteria sequences have also been found in coastal groundwater and worm burrows (16S rRNA gene analysis; McAllister et al., 2015) and briny terrestrial groundwater (metagenomics; Emerson et al., 2016). Steel coupon incubation experiments provide sequence and culture-based evidence that Zetaproteobacteria inhabit coastal waters (Dang et al., 2011; McBeth et al., 2011; Mumford et al., 2016), but the Chesapeake isolates are the first Zetaproteobacteria isolated directly from a coastal redoxstratified water column. In total, previous studies show that Zetaproteobacteria grow at oxic–anoxic interfaces where Fe(II) and O<sup>2</sup> are available, typically preferring lower O<sup>2</sup> concentrations (Chan et al., 2016), though Mariprofundus sp. DIS-1 is an exception in that it tolerates saturated O<sup>2</sup> conditions (Mumford et al., 2016). The molecular mechanism of neutrophilic Fe(II) oxidation is not well-known; comparative analysis of six existing Zetaproteobacteria isolate genomes with freshwater FeOB genomes has resulted in hypothesized pathways (Singer et al., 2011; Liu et al., 2012; Barco et al., 2015), but differences in single amplified genomes (SAGs) and metagenomes suggest that the pathway has some variants (Field et al., 2015; Fullerton et al., 2017). Fe(II) oxidation by the Zetaproteobacteria results in Fe(III) oxyhydroxides, typically in the form of twisted ribbon-like stalks, which form the framework of Fe microbial mats (Chan et al., 2016). Such large, dense stalk structures would make it difficult for a pelagic FeOB to maintain buoyancy. In sum, our knowledge of benthic Zetaproteobacteria may not necessarily be representative of FeOB in the water column.

Here we detail the isolation, physiological characterization, and genomic analysis of two new Fe-oxidizing Zetaproteobacteria from the Chesapeake Bay, Mariprofundus aestuarium CP-5 and Mariprofundus ferrinatatus CP-8. We compare the CP strains to the other Zetaproteobacteria and propose that physiological and genomic distinctions represent adaptive strategies for the Chesapeake Zetaproteobacteria to scavenge Fe in low Fe(II) waters and to withstand highly variable oxygen conditions associated with physically dynamic redoxclines.

#### MATERIALS AND METHODS

#### Sampling, Enrichments, and Isolation

The redox-stratified waters of the Chesapeake Bay at Station 858 (38◦ 58.600 N, 076◦ 22.080 W) were sampled aboard the R/V Hugh R Sharp in August, 2014. Details of sampling and the geochemical conditions can be found in Field et al. (2016). Water samples collected from the oxic–anoxic transition zone were used to inoculate FeOB enrichment cultures. Agarose-stabilized gradient tube cultures (Emerson and Floyd, 2005) were set up with a FeCO<sup>3</sup> plug (1% w/v high-melt agarose) and simulated estuary medium (0.15% w/v low-melt agarose), which is a 50:50 mixture of modified Wolfe's mineral medium (MWMM) and artificial seawater (ASW). Per liter, estuary medium contains 13.75 g NaCl, 2.69 g MgCl2-6H2O, 3.49 g MgSO4-7H2O, 0.36 g KCl, 0.75 g CaCl2-2H2O, 1 g NH4Cl, 0.05 g KH2PO4, 0.42 g NaHCO3. After autoclaving, estuary medium was amended with 1 µL/mL Wolfe's trace mineral solution and 1 µL/mL vitamin solution and adjusted to pH 6.2 with CO2. The headspace of all tube cultures contained a low O<sup>2</sup> gas mixture (N2/CO2/O2; 95:4:1).

Strains CP-5 and CP-8 were isolated by serial dilutionto-extinction from water samples CTD12-5 and IS8-11.3 respectively (water geochemistry in Supplementary Table ST1; further details in Field et al., 2016). Growth was confirmed by the development of colonies or distinct growth bands in agarose-stabilized tubes (**Figure 1**) and by microscopy. Purity was checked by microscopic observation, absence of heterotrophic growth on R2A-estuary medium agar plates, and sequencing of the 16S rRNA genes amplified with the bacterial-universal primer sets Bac27F and Uni1492R (Lane, 1991).

#### Physiological Characterization

To assess alternate substrate usage and optimal growth conditions of strains CP-5 and CP-8, growth tests were carried out in agarose-stabilized gradient tubes as described above, but buffered to pH 7.0 except for pH testing. To test if the strains could use non-Fe(II) substrates, we tested growth on 5 mM sodium thiosulfate, 5 mM sodium sulfide, 10 mM sodium pyruvate, 10 mM glucose, 10 mM sodium acetate, and 0.2% w/v yeast extract. The pH range of growth was determined using several

buffers: acetate-acetic acid (pH 5.0 and 5.2, 10 mM), MES (pH 5.5, 6.0, 6.4 and 6.9, 10 mM), and HEPES (pH 7.2, 7.4, 7.7, 8.0, 8.3 and 8.5, 10 mM). pH measurements taken before and after cultivation confirmed minimal (0.1–0.2) decreases during cultivation periods. Preferred growth temperature was determined by incubating cultures at 5, 10, 15, 20, 25, 30, 35, and 40◦C, and preferred salinity was determined using different ratios of MWMM (0h): ASW (35h): 0:10, 1:9, 2:8, 3:7, 4:6, 5:5, 6:4, 7:3, 8:2, 9:1, and 10:0. All cultures were assessed for growth after 2 weeks based on the development of growth bands and observation by fluorescent microscopy.

lines. Fluorescence micrographs of CP-5 (B) and CP-8 (C) cells.

To determine the preferred oxygen concentration for growth, the dissolved oxygen (DO) within strain CP-5 and CP-8 growth bands was measured 48 h after inoculation using a Firesting optical oxygen probe with a needle-type sensor (PyroScience, Aachen, Germany) mounted on a micromanipulator (Narishige International, Amityville, NY, United States). Attempts to test growth under anoxic conditions were also set up by preparing deoxygenated media (bubbling with N<sup>2</sup> and autoclaving in an N2-flushed vessel), setting up gradient tubes under a stream of N2, and using a 100% N<sup>2</sup> headspace. However, the Firesting optical oxygen probe detected trace O<sup>2</sup> in the gradient tube medium (∼250 nM) indicating that this procedure did not yield completely anoxic cultures.

To measure the growth rate of each strain, growth bands from triplicate gradient tube cultures (buffered with PIPES) were harvested daily over the course of the experiment (10 and 9 days for strain CP-5 and CP-8 respectively). Samples were stained with Syto 13 for cell counting using a Petroff-Hausser counting chamber. Aliquots of harvested growth bands were also used to measure total Fe concentrations in cultures over time. Fe concentrations in abiotic control gradient tubes were measured as well, using samples at the same height as biotic growth bands.

We used total Fe measurements to follow Fe(II) oxidation because nearly all Fe accumulated in developing growth bands was shown to be Fe(III) in the strain CP-5 growth experiment (data not shown). Samples for total Fe analysis were reduced with 200 mM hydroxylamine for 22–24 h and measured using the ferrozine method (modified from Stookey, 1970).

#### Microscopy

Phase contrast and fluorescent micrographs of cultures (stained with SYBR green I, in the case of fluorescence) were captured on an Olympus BH-2 microscope with 400x total magnification. For these analyses, we used liquid cultures (without agarose) grown for 24 h. Samples for scanning electron microscopy (SEM) were gently mounted on a 0.2-µm-pore-size polycarbonate filter, air dried, and coated with gold/palladium for observation, or with carbon for energy dispersive spectroscopy (EDS) analysis. Samples for transmission electron microscopy (TEM) were gently mounted on a Formvar-coated copper grid, air dried, and coated with gold/palladium. Electron microscopy was performed at the Delaware Biotechnology Institute Bioimaging Center, using a Hitachi S-4700 field emission SEM with an Oxford INCA EDS system and a Zeiss LIBRA 120 TEM.

#### Genome Sequencing and Analysis

For DNA extraction, strains CP-5 and CP-8 were grown using 25 mL FeCO<sup>3</sup> gradient plates (1 L total volume per strain) under microaerobic conditions (N2/CO2/O2; 95:4:1; Emerson and Floyd, 2005). Genomic DNA was isolated from these cultures using the FastDNA Spin Kit for Soil (MP Biomedicals, Santa Ana, CA, United States). We used the PowerClean Pro DNA kit (MO BIO Laboratories, Carlsbad, CA, United States) to remove remaining inhibitors. The purified DNA (2.5 and 0.5 µg of CP-5 and CP-8, respectively) was size-selected using electrophoresis (BluePippin, Sage Science, Beverly, MA, United States) to a minimum size of 6 kb, resulting in an average size of 12 kb. The genomes were sequenced using PacBio RSII technology at the University of Delaware Sequencing and Genotyping Center. Size-selected DNA was prepared for sequencing using the SMRTbell Template Prep Kit 1.0 (PacBio, Menlo Park, CA, United States) as per the manufacturer's instructions. One SMRT cell per genome was sequenced with P6-C4 chemistry and a 6-h movie. For strain CP-5, sequencing generated 1.37 Gbp of raw data (mean read length 15,263 bp; N<sup>50</sup> 26,034 bp); for strain

CP-8, sequencing generated 0.84 Gbp of raw data (mean read length 8,877 bp; N<sup>50</sup> 19,311 bp). Assembly was completed on the PacBio SMRT Portal. Subreads were filtered to a minimum length of 1 kb (CP-5) or 2 kb (CP-8) with a polymerase quality score minimum of 0.8. The hierarchical genome assembly process 3 (HGAP-3) was used to assemble a single high quality contig from each of the sequencing runs. The average coverage over the entire sequenced contigs was 382x for strain CP-5 and 300x for strain CP-8. Gepard (v.1.40; Krumsiek et al., 2007) was used to compare each genome against itself to check for inverted repeats and to close each contig into a complete circular genome.

The complete genomes of strains CP-5 and CP-8 were annotated using the pipeline of the Integrated Microbial Genome Expert Review (IMG/ER) system (Markowitz et al., 2012). Manual verification of predicted genes of interest was completed using MUSCLE alignments in MEGA (v.7.0.14) against reference gene sequences from UniProt or the RSCB Protein Data Bank (Edgar, 2004; Kumar et al., 2016). The Rapid Annotation using Subsystem Technology (RAST) platform (Aziz et al., 2008; Overbeek et al., 2014) was used to identify possible frameshifts (none were detected) and to help find genes unique to the CP strain genomes, with respect to the other Zetaproteobacteria. Average amino acid identities (AAIs) of bidirectional best hit proteins were calculated using a web-based calculator<sup>1</sup> . Reported AAI values are the average of the separate calculations run in both directions for each pair (standard deviation < 1.18%). Average nucleotide identities (ANIs) were calculated using OrthoANI (Yoon et al., 2017). An AAI heatmap was made using the R package gplots heatmap.2 (v 3.0.1); hierarchical clustering using complete agglomeration was used to calculate the dendrogram.

#### 16S rRNA Gene Analysis

The CP strain 16S rRNA genes were found in their completed genomes and aligned to the arb-SILVA database using the SINA online web tool (v.1.2.11; Pruesse et al., 2012). Aligned sequences were masked to unambiguously aligned base positions and a maximum-likelihood tree was constructed using RAxML with the GTR-gamma nucleotide substitution model (v.8.2.8; Stamatakis, 2014). Bootstrap values were estimated from 500 replicates. To calculate pairwise percent nucleotide identity, we calculated the Similarity score metric on the Ribosomal Database Project (RDP) website (Cole et al., 2014).

#### Genome Accession Numbers and Culture Availability

GenBank accession numbers for Mariprofundus aestuarium CP-5 and Mariprofundus ferrinatatus CP-8 are CP018799 and CP018800 respectively. IMG taxon IDs for strains CP-5 and CP-8 are 267118011 and 267180111 respectively. Both isolates are available on request from C.S. Chan (University of Delaware, United States) and at the Provasoli-Guillard National Center for Marine Algae and Microbiota (NCMA; Bigelow Laboratory for Ocean Sciences, United States).

## RESULTS AND DISCUSSION

#### Isolation and Physiological Characterization

Strains CP-5 and CP-8 were both successfully isolated using Fe(II)/O<sup>2</sup> gradient tubes after five transfers of the 10−<sup>5</sup> serial dilutions. Growth consistently appeared as a sharp orange band typical of microaerophilic FeOB (**Figure 1A**) and cells appeared as curved rods under fluorescent microscopy (**Figures 1B,C**). Strain CP-5 cells are 0.43 ± 0.05 µm × 1.01 ± 0.18 µm, and strain CP-8 cells are 0.45 ± 0.04 µm × 0.91 ± 0.08 µm. Purity was demonstrated by the lack of growth on R2A-estuary medium plates (no contaminant oligotrophs) and by a single unambiguous full length 16S rRNA gene sequence amplified from each culture.

Strains CP-5 and CP-8 have doubling times of 19.5 and 27 h, respectively. These generation times are slower than M. ferrooxydans PV-1 (12 h), but similar to the 24 h doubling time reported for the closely related Mariprofundus micogutta. During growth, strains CP-5 and CP-8 both accelerated Fe(II) oxidation, compared to uninoculated controls (**Figure 2**). The O<sup>2</sup> concentration in the growth bands of inoculated gradient tubes was <2 µM O<sup>2</sup> (Supplementary Figure S1), comparable to or lower than M. ferrooxydans PV-1 (Krepski et al., 2013). Strains CP-5 and CP-8 appear to be obligate Fe(II)-oxidizers as neither grew on reduced S or organic carbon substrates (**Table 1**). Overall, our experiments suggest that strains CP-5 and CP-8 are microaerophilic chemolithoautotrophic Fe(II)-oxidizers, consistent with all other Zetaproteobacteria isolates.

To optimize culturing of the CP strains, growth was tested over a range of salinity and pH. The preferred salinity was brackish, 14–17.5h, with no growth at 0h (freshwater) or <sup>35</sup>h (normal seawater). The preferred pH range was 6.9–7.2, and both strains grew at pH up to 8.3, unusually high for neutrophilic FeOB isolates, which typically prefer pH between 6.0 and 6.5 (e.g., M. ferrooxydans PV-1 and M. micogutta, **Table 1**; freshwater FeOB Gallionella capsiferriformans ES-2, Sideroxydans lithotrophicus ES-1, and Ferriphaselus amnicola OYT-1; Emerson and Moyer, 1997; Kato et al., 2014). One exception is Mariprofundus sp. DIS-1, which can grow at pH 8.0 (Mumford et al., 2016). The CP strain salinity and pH preferences reflect the brackish seawater from which they were sampled.

#### Phylogenetic Analyses

Strains CP-5 and CP-8 are representative of the Chesapeake Bay environment, as their 16S rRNA gene sequences match the dominant 16S rRNA sequences of the original FeOB enrichment cultures from which each strain was isolated (Supplementary Figure S2). Phylogenetic analysis of 16S rRNA gene sequences shows that strains CP-5 and CP-8 are Zetaproteobacteria within OTUs 18 and 37 respectively (as defined by McAllister et al., 2011 and determined using ZetaHunter<sup>2</sup> ) and cluster with nearly all other isolated Zetaproteobacteria (**Figure 3**). Among the Zetaproteobacteria isolates and SAGs, strains CP-5 and CP-8

<sup>1</sup>http://lycofs01.lycoming.edu/~newman/AAI/

<sup>2</sup>https://github.com/mooreryan/ZetaHunter

are most similar to each other based on ANI, average AAI, and 16S rRNA gene identity (**Table 2**). Because both strains share less than 97% 16S rRNA gene identity (Stackebrandt and Goebel, 1994) and have less than 95% ANI (Konstantinidis and Tiedje, 2005) with all other Zetaproteobacteria isolates and SAGs, including each other, strains CP-5 and CP-8 are two new species, with proposed names Mariprofundus aestuarium CP-5 and Mariprofundus ferrinatatus CP-8.

Comparisons of 16S rRNA sequences and %AAI among Zetaproteobacteria show that most of the isolates fall within a closely related group, i.e., the genus Mariprofundus (**Figure 3**). By 16S rRNA gene identity and %AAI, the CP strains are most closely related to Mariprofundus sp. DIS-1, isolated from a steel coupon incubated in a coastal bay (Mumford et al., 2016), and M. micogutta, isolated from marine hydrothermal sediment (**Table 2** and **Figure 3**; Makita et al., 2016). These close relationships show that Mariprofundus is a cosmopolitan genus that inhabits a variety of environments, coastal and deep sea, as well as planktonic, benthic, and subsurface.

## Iron Oxyhydroxide Biomineral Morphology

To investigate how suspended FeOB manage Fe oxyhydroxide precipitation to avoid sinking, we examined the Fe biominerals produced by strains CP-5 and CP-8. Both strains produce bundles of stubby rod-shaped extracellular structures (**Figure 4**), confirmed to be Fe-rich by SEM-EDX (Supplementary Figure S3) and morphologically distinct from abiotic mineral precipitates (Supplementary Figures S4, S5). This morphology has previously been identified in freshwater FeOB and referred to as dreadlocks (or dreads) given their resemblance to the dreadlock hairstyle (**Figure 5**; Kato et al., 2015). Dreads are somewhat similar to the fibrillar twisted Fe stalks produced by other microaerophilic FeOB (**Figure 5B**; Chan et al., 2011; Kato et al., 2014), in that they are bundles of elongated Fe oxyhydroxides (referred to as oxides from here on). However, dreads are short, never exceeding 10 µm in length, and many dreads can radiate from, and surround a single cell. In contrast, Fe oxide stalks range in length from 10's of µm to mm, extend from one side of the cell, and are used by mat-forming FeOB to anchor themselves to



<sup>∗</sup>Sulfide, thiosulfate; ∗∗glucose, acetate, pyruvate, yeast extract, ∗∗∗this study.

surfaces (Chan et al., 2011, 2016). Dreads were closely associated with CP cells observed by fluorescent microscopy (**Figure 4A**) while the radiating arrangement observed under SEM made it apparent that CP strain cells were once attached (**Figure 5A**). In fact, the lack of cell-attached dreads under SEM suggests they are easily shed. In total, these observations suggest that the CP strains produce short Fe oxide dreads as an adaptation to shed their biominerals to maintain suspension within the water column.

#### General Genome Features of Strains CP-5 and CP-8

The CP-5 and CP-8 strain genomes are both single circular chromosomes, which make them the first and only closed Zetaproteobacteria genomes. High consensus read coverage (382x for strain CP-5; 300x for strain CP-8) led to significant overlap of the ends of each CP strain genome assembly (15 and 9 kb, respectively), overall providing confidence in genome accuracy and completion. The CP-5 and CP-8 strain genomes are 2.54 and 2.30 Mbp, respectively; sizes, GC contents, and COG distributions are comparable to the other sequenced Zetaproteobacteria isolates (**Table 1** and Supplementary Tables ST2, ST3). The COG distribution of the two CP strains is highly similar (Supplementary Table ST3) and there are no obvious major metabolic or physiological differences apparent in the genes distinguishing the two strains from one another (Supplementary Tables ST4, ST5). The CP-5 and CP-8 strain genomes contain 258 and 211 genes without homologs in other Zetaproteobacteria isolates. As described below, the CP strains share several genes that are absent or rare in the other sequenced Zetaproteobacteria and may represent adaptations to life in the water column.

#### Electron Transport Chain Analysis

Based on the electron transport-related genes identified in the genomes (Supplementary Table ST6), strains CP-5 and CP-8 appear to have an electron transport system similar to other Zetaproteobacteria (**Figure 6**), with some key differences, described below. Like all microaerophilic FeOB, including Zetaproteobacteria, the CP strains have genes encoding the putative Fe oxidase, outer membrane cytochrome Cyc2 (e.g., Barco et al., 2015; Kato et al., 2015; Mumford et al., 2016), which has been proven to oxidize Fe(II) in Acidithiobacillus ferrooxidans (Castelle et al., 2008). The CP strain cyc2 gene sequences are homologous to characterized cyc2 sequences from PV-1 (e-values: 10−<sup>72</sup> to 10−73; Supplementary Table ST6; Barco et al., 2015) and each contain a predicted signal sequence, one CXXCH heme-binding motif, and an outer membrane beta barrel domain as with other cyc2 gene sequences (White et al., 2016). The CP strains both lack the putative outer membrane Fe oxidase MtoA (Liu et al., 2012), consistent with our observation that Cyc2 is common amongst microaerophilic and other FeOB, while MtoA is rare (Kato et al., 2015).


 of strains CP-5 and CP-8 to other

Zetaproteobacteria.

TABLE 2 | Source environment,

Frontiers in Microbiology | www.frontiersin.org

 16S rRNA gene identity, ANI, and AAI comparisons

fmicb-08-01280 July 14, 2017 Time: 15:38 # 8

numerous

Zetaproteobacteria

 genomes (strains CP-5, CP-8, ET2, DIS-1, EKF-M39, TAG-1, and SV108, and SAG C09), thereby

underestimating

 % completeness

 of these genomes.

One unusual feature in the CP strain genomes is the possession of aa3-type cytochrome c oxidases (Supplementary Table ST6) in place of the cbb3-type cytochrome c oxidases present in all other

(C) Scanning electron micrograph of bundles of iron oxide dreads produced

Zetaproteobacteria genomes to date. Several SAGs and isolate M. micogutta have both aa<sup>3</sup> and cbb3-type oxidases (Field et al., 2015), but no other Zetaproteobacteria has only the aa3-type. Between the aa<sup>3</sup> and cbb3-type oxidases, the cbb3-type oxidase is considered to be better adapted for low O<sup>2</sup> conditions given its higher affinity for oxygen (Arai et al., 2014), consistent with the association of Zetaproteobacteria with low O<sup>2</sup> habitats. Conversely, the lower oxygen affinity of the aa3-type oxidase suggests adaptation to somewhat higher O<sup>2</sup> conditions. Though the K<sup>m</sup> values of both oxidases would be considered low O<sup>2</sup> (Km,cbb<sup>3</sup> on the order of nanomolar and Km, aa<sup>3</sup> on the scale of micromolar O2; Arai et al., 2014), the difference suggests that the CP strains may have a higher O<sup>2</sup> niche. Curiously, the single Zetaproteobacteria isolate shown capable of growing in O2-saturated waters, DIS-1, possesses only the cbb3-type oxidase, suggesting other genetic adaptations contribute to its O<sup>2</sup> tolerance. Still, the uncommon possession of only aa3-type oxidases in the CP strains likely represents an adaptation to frequent exposure to high O<sup>2</sup> waters.

Periplasmic electron carriers are required for electron transport between Cyc2 and the terminal oxidase. Because of the high redox potential of Fe(II)/Fe(III)OOH (24 mV for ferrihydrite; Majzlan, 2012), these electron carriers are most likely cytochromes. In A. ferrooxidans, the cytochrome Cyc1 is one of these intermediate electron carriers (Malarte et al., 2005; Castelle et al., 2008). While there are homologs to cyc1 in several Zetaproteobacteria isolate genomes, the CP strain genomes lack homologs. However, Cyc1 is suggested to interact specifically with the cbb3-type oxidase in M. ferrooxydans PV-1 (Barco et al., 2015), making the lack of cyc1 homologs in the CP strains consistent with the absence of the cbb3-type oxidase. There is a different predicted periplasmic cytochrome found in the CP strains, which may transfer electrons between Cyc2 and the aa3-type terminal oxidase. This potential periplasmic cytochrome gene in both strains CP-5 and CP-8 codes for a 127aa protein, with a signal sequence and one CXXCH heme-binding motif (Supplementary Table ST6). In the strain CP-8 genome, this gene is located near the genes encoding the terminal aa3-type oxidase, but it is in a different genomic neighborhood in strain CP-5. Homologs of this periplasmic cytochrome are found in several Zetaproteobacteria isolates (PV-1, JV-1, M34, and EKF-M39; e-values 10−<sup>23</sup> to 10−21) and are also near terminal oxidases. The genomic context and presence in several Zetaproteobacteria (including seven SAGs) suggests that this predicted cytochrome plays a role in Fe(II) oxidation and energy conservation.

The high potential of Fe(II)/Fe(III)OOH requires FeOB to regenerate NADH using either reverse electron transport, or an alternate reductant. Like other Zetaproteobacteria, the CP strains have the components for reverse electron transport: a bc1 complex, ubiquinone synthesis genes, and NADH dehydrogenase (**Figure 6**). However, the CP strains are the only Zetaproteobacteria isolates that definitively lack an alternative complex III, indicating that it is not a necessary component for neutrophilic Fe(II) oxidation, despite its conservation in other FeOB (Singer et al., 2011; Kato et al., 2015). Both CP strains have a cytochrome b/diheme cytochrome c gene cluster

by strain CP-5.

FIGURE 5 | Scanning electron micrograph of extracellular iron oxide biomineral structures. (A) Dreads produced by strain CP-8, with likely location of missing cell denoted as a yellow oval. (B) Dreads surrounding a freshwater FeOB Ferriphaselus R-1 cell, highlighted in yellow (modified from Kato et al., 2015). To the right of the cell, a longer iron oxide stalk produced by R-1 is also visible.

(Supplementary Table ST6) that likely also plays an electron transport role. Present in all Zetaproteobacteria isolates and several SAGs, these two genes in each of the CP strains are also homologous to the fused cytbc gene in the Feoxidizing KS culture Gallionellaceae, which was proposed to pass electrons from periplasmic cytochromes to quinones and on toward denitrification (He et al., 2016). The CP strains lack a dissimilatory nitrate reductase, but this novel bc complex may still function to reduce quinones for reverse electron transport. Both CP strains have genes coding sulfide quinone oxidoreductases (Supplementary Table ST6), which would allow them to take advantage of the high sulfide concentrations in the Chesapeake Bay to reduce quinones. The CP strains also have hoxWHYUF genes, which could allow them to use H<sup>2</sup> to reduce NAD<sup>+</sup> to NADH (Tran-Betcke et al., 1990; Thiemermann et al., 1996), relieving at least some of the need for reverse electron transport. In sum, the CP strain genomes show multiple options for regenerating NADH for carbon fixation and biosynthetic pathways.

#### Carbon Metabolism Analysis

The CP strain genomes are consistent with autotrophy in these organisms. The CP strains each possess complete gene sets for the Calvin–Benson–Bassham (CBB) cycle, including form II ribulose 1,5-bisphosphate carboxylase (RuBisCO) for fixing inorganic carbon (Supplementary Table ST7). Also present are the genes to convert the chief product of the CBB cycle, glycerate 3P, to pyruvate, which can then be utilized in the predicted, complete tricarboxylic acid (TCA) cycle to generate energy and biosynthetic precursors (Supplementary Table ST7).

The CP strain genomes each contain form II RuBisCO and lack form I RuBisCO, as observed in several other Zetaproteobacteria (e.g., EKF-M39, SV108 M. micogutta, Zetaproteobacteria SAGs). CO<sup>2</sup> concentrations were ∼70–80 µM in the waters from which the CP strains were isolated (Cai et al., 2017). These concentrations are within the Km,CO<sup>2</sup> ranges for both Form I and Form II RuBisCO (Badger and Bek, 2008), so either should be functional in this environment. However, the absence of form I RuBisCO is somewhat unexpected in the CP strains given

that form I is considered to be better adapted to higher O<sup>2</sup> conditions than form II (Badger and Bek, 2008) and would provide a potential adaptation for more efficient carbon fixation during exposure to higher O<sup>2</sup> waters. Indeed, Mumford et al. (2016) suggest that the presence of both forms of RuBisCO in DIS-1 helps adapt this strain to a larger range of oxygenated environments. In any case, the form II RuBisCO genes in the CP strain genomes support Fe(II) oxidation chemolithoautotrophy, consistent with all other Zetaproteobacteria.

Support for strict autotrophy comes from the apparent lack of transporters for organic carbon substrates. Close analysis of a cluster of genes annotated as phosphotransferase (PTS) system genes in each CP strain genome suggests they do not make up a complete system for carbohydrate uptake, but may instead play a role in nitrogen regulation (Supplementary Table ST7). The CP strain genomes also lack complete ABC transport systems for sugars, peptides, and amino acids, making heterotrophy unlikely.

## Unusual Genomic Features for Biofilm Formation

We surveyed the CP genomes for genes that could represent adaptations to life in the Chesapeake Bay redoxcline, focusing on ones that were rare or absent in other Zetaproteobacteria. We found two gene clusters related to biofilm formation: the Wsp system, a chemosensory system that produces the biofilm-inducing signal molecule cyclic dimeric guanosine monophosphate (c-di-GMP), and the widespread colonization island (WCI), a pilus assembly system that enables surface attachment.

Each CP strain genome includes a complete wsp gene cluster (wspABCDEFR), which encodes the Wsp chemosensory system (**Figure 7** and Supplementary Table ST8). Genetic and protein functional studies have demonstrated the role of these genes in biofilm formation in Pseudomonas, the model organism for the Wsp system (D'Argenio et al., 2002; Hickman et al., 2005). The Wsp system is homologous to the Che chemotaxis system; both contain a methyl-accepting chemotaxis protein (MCP) chemoreceptor and a complex of signal transduction proteins (Bantinaki et al., 2007). However, the Wsp system regulates biofilm formation rather than flagellar motor switching as in the Che system. The major distinguishing feature of the Wsp system is subunit WspR, a diguanylate cyclase response regulator required for Wsp system-induced biofilm production. Phosphorylation stimulates WspR to synthesize the signal molecule cyclic di-GMP (c-di-GMP), which induces biofilm formation pathways, including the production of extracellular polymeric substances (EPSs; D'Argenio et al., 2002; Hickman et al., 2005; Malone et al., 2007). The signal activating the Wsp system MCP, WspA, remains unclear, but has been shown to be related to physical and/or chemical signals associated with growth on surfaces (Güvener and Harwood, 2007; O'Connor et al., 2012). This suggests that given a mechanism for initial surface attachment, the Wsp system could enable the CP strains to form biofilms to colonize particles in the water column.

Each CP strain genome contains two predicted copies of wspR that are homologs of the gene sequences of the functionally and structurally characterized WspR of Pseudomonas aeruginosa (e-values: 10−<sup>111</sup> to 10−93; Supplementary Table ST8; De et al., 2008). Like Pseudomonas wspR, all CP strain wspR sequences

contain the conserved C-terminal diguanylate cyclase domain GGEEF in the active site loop and the RxxD motif making up the conserved inhibitory site in GGEEF domain-containing proteins (De et al., 2008, 2009). The remaining Wsp system subunit genes (wspABCDEF) are also present in the CP strain genomes and homologous to the wsp counterparts in Pseudomonas species (e-values 0 to 10−35; Supplementary Table ST8). Other Zetaproteobacteria genomes either lack wsp gene homologs or only contain single subunits (TAG-1, wspR; SV108, wspR; M34, wspE). The exception is Zetaproteobacteria SAG C09, a Loihi Seamount Fe mat single cell genome (2.45 Mb; Field et al., 2015), which contains wspABCDEF, but clearly lacks wspR, with the gene cluster in the middle of a contig. Instead, the immediately adjacent features are a pseudouridine synthase and a tRNA, which are not obviously related to biofilm formation, though further downstream in the cluster (6 ORFs away from wspF), there is an adenylate cyclase gene. Adenylate cyclase forms the signal molecule cAMP, which is associated with many processes (Gancedo, 2013), one of which may be initial cell attachment to surfaces (O'Toole and Wong, 2016). Nevertheless, the lack of wspR in Zetaproteobacteria C09 genome suggests that the wspABCDEF homologs in C09 have a different role and output than in the CP strains. The absence of complete Wsp systems in Zetaproteobacteria genomes other than the CP strains suggests that Wsp-related biofilm formation may be an adaptation specific to pelagic Zetaproteobacteria for particle colonization.

The second genomic feature of the CP strains that is rare among Zetaproteobacteria is the WCI, a gene cluster responsible for tight attachment to surfaces (**Figure 7** and Supplementary Table ST9). The WCI includes flp-1, a gene encoding for the major structural component of the type IV Flp (fimbrial lowmolecular-weight protein) pili, as well as the tad (tight adherence) genes, and rcp pilus assembly genes (Planet et al., 2003). First characterized in Aggregatibacter actinomycetemcomitans, but studied in numerous other organisms including Caulobacter and Pseudomonas (Skerker and Shapiro, 2000; Bentzmann et al., 2006), the WCI genes assemble adhesive Flp pili that mediate tenacious surface adherence and biofilm formation (Kachlany et al., 2000, 2001; Planet et al., 2003). The CP strain genomes each contain flp-1 gene sequences that were confirmed to contain the conserved processing site motif GXXXXEY (Inoue et al., 1998; Kachlany et al., 2001), as well as tadABCDEGZ and rcpAC (Supplementary Table ST9). Both CP strain genomes have two predicted copies of tadE, one of which is likely tadF given the high sequence similarity of these two subunits (Tomich et al., 2006). Two WCI genes, tadV and rcpB, are not present in the CP genomes, which may be due to the general variability of WCI organization across bacteria or the potential for individual species to possess novel genes in place of individual WCI components (Tomich et al., 2007). For example, P. aeruginosa was demonstrated to encode a novel prepilin peptidase, FppA, instead of TadV (Bentzmann et al., 2006), suggesting that the CP strains could possess different versions of TadV and RcpB that would not be recognized by genomic analysis alone. The set of WCI genes in the CP strain genomes is also found in EKF-M39, but entirely absent from all other Zetaproteobacteria isolates, suggesting that only a small subset of Zetaproteobacteria can produce Flp pili. The mechanisms thought to regulate WCI Flp pilus production vary across species (Tomich et al., 2007) and could be controlled by c-di-GMP signaling in the CP strains. In each CP strain genome, predicted WCI components are adjacent to pilZ domain-containing genes (Supplementary Table ST9), which have been linked to c-di-GMP-regulated fimbriae production (Johnson et al., 2011; Wilksch et al., 2011). The CP strain pilZ sequences contain the c-di-GMP binding motifs RxxxR-(D/ N)x(S/A)xxG (Ryjenkov et al., 2006; Christen et al., 2007), and thus may connect Wsp system c-di-GMP synthesis to WCI Flp pilus production to promote a surface-attached biofilm lifestyle.

#### Adaptations of Pelagic Zetaproteobacteria in Estuarine Water Columns

In many ways, the Chesapeake strains are like other Zetaproteobacteria isolates: they are autotrophic, obligate Fe(II) oxidizers, with similar electron transport and carbon fixation mechanisms. However, the CP strains differ in that they produce distinctive dreadlock-shaped oxides that are much smaller than the twisted stalks common to other Zetaproteobacteria. This is likely a key difference between pelagic and benthic FeOB. Our previous work has shown that Fe biomineral stalks are the building blocks of Fe microbial mats (Chan et al., 2016). Millimeters-long stalks create a highly porous framework unlike any other biofilm or mat, in that the bulk is made of mineral, without much interstitial EPS. This architecture allows Fe(II)-bearing fluids to flow through mats, enabling FeOB to biomineralize and position themselves at a benthic Fe(II)/O<sup>2</sup> interface (e.g., hydrothermal vent on the seafloor). In contrast, for FeOB that need to maintain position in a water column, large stalk structures would be undesirable because of their weight. Instead, the CP strains make smaller dreads that can be constantly shed, thereby eliminating the heavy oxide by-products to avoid sinking out of the oxic–anoxic transition zone.

Instead of making Fe mats, it appears that the CP strains have the genes to form standard, EPS-bound biofilms, as is typical of organisms possessing the WCI and Wsp system. EPS production enables marine bacteria to colonize suspended particle surfaces (Decho, 1990), and is likely key for CP strains to attach to and access nutrients from Fe(II)-bearing minerals such as FeS. In seasonally stratified Chesapeake Bay waters, particulate FeS is formed by an O2-Fe-H2S catalytic cycle where sulfidic (H2S) bottom waters reduce solid Fe(III) oxyhydroxides to dissolved Fe(II) and react further to precipitate solid FeS particles (**Figure 8A**; MacDonald et al., 2014; Hansel et al., 2015; Field et al., 2016). Indeed, a majority of the Fe(II) in the Chesapeake Bay oxic–anoxic transition zone was found to be particulate (Field et al., 2016), likely as FeS. Still, FeS particles are likely to be sparse given the low overall Fe(II) concentration, necessitating strategies for the CP strains to recognize and firmly attach to these Fe(II)-bearing particles. The c-di-GMP produced by the Wsp system could stimulate WCI Flp pilus production and other pathways involved in biofilm production to facilitate tight cell attachment to suspended FeS particles (**Figure 8B**).

These small cell-mineral aggregates could further assemble and grow into larger flocs, incorporating other suspended particles and cells (**Figure 8C**). These flocs could trap more FeS for consumption (**Figure 8D**), and any trapped dreads could be recycled back to FeS if flocs settle into sulfidic bottom waters (**Figure 8E**). Floc formation can further benefit the CP strains by effectively increasing their O<sup>2</sup> tolerance, as previously described for freshwater floc-dwelling FeOB (Elliott et al., 2014). Diffusion of O<sup>2</sup> into flocs is slowed by the EPS matrix (Brezonik, 1993). If floc-dwelling bacteria consume O<sup>2</sup> faster than it can diffuse in, low oxic or anoxic microenvironments develop within the floc structure (Flemming and Wingender, 2001; Han et al., 2012). Should a CP strain floc be mixed into oxic layers of the water column, low [O2] microenvironments would provide a niche where the CP strains could still compete with abiotic oxidation for matrix-bound Fe(II). In all, biofilm-related genes would give the CP Zetaproteobacteria multiple advantages for persisting in the water column despite low Fe(II) and fluctuating O<sup>2</sup> conditions.

Mariprofundus aestuarium CP-5 and Mariprofundus ferrinatatus CP-8 add to the growing number of Zetaproteobacteria isolates that form a closely related phylogenetic cluster despite differing environmental origins and lifestyles. Our analyses suggest that key genes can confer specialized strategies for these organisms to live in diverse environmental niches. If broadly true, the Fe-oxidizing Zetaproteobacteria would be expected to live anywhere that Fe(II) and O<sup>2</sup> are available, and thereby be a widespread driver of marine Fe cycling.

#### Description of Mariprofundus aestuarium sp. nov.

Mariprofundus aestuarium [aes.tu.a'ri.um. L. n. aestuarium an estuary].

Cells are slightly curved, short rods (0.43 ± 0.05 µm × 1.01 ± 0.18 µm). Does not form spores. Mesophilic and neutrophilic. Microaerobic, growing with opposing gradients of Fe(II) and O2. Autotrophic. Grows at 10–30◦C (optimally at 20–25◦C), pH 5.5–8.3 (optimally at pH 6.9–7.2), and 7–31.5<sup>h</sup> salinity (optimally at 14–17.5h salinity). Utilizes ferrous iron as an energy source for lithotrophic growth. Does not utilize thiosulfate, sulfide, pyruvate, glucose, or acetate as an energy source. Produces extracellular dreadlock-like iron oxides around

the cell. The doubling time under optimal conditions is 19.5 h. The type strain is CP-5<sup>T</sup> , isolated from redox-stratified waters in the Chesapeake Bay, United States. The total DNA G+C content of the type strain is 51.5 mol%.

## Description of Mariprofundus ferrinatatus sp. nov.

Mariprofundus ferrinatatus [fer.ri.na'ta.tus. L. neut. n. ferrum iron; L. masc. n. natatus floating; N.L. masc. n. ferrinatatus floating iron].

Cells are slightly curved, short rods (0.45 ± 0.04 µm × 0.91 ± 0.08 µm). Does not form spores. Mesophilic and neutrophilic. Microaerobic, growing with opposing gradients of Fe(II) and O2. Autotrophic. Grows at 15–35◦C (optimally at 25–30◦C), pH 5.5–8.3 (optimally at pH 6.9–7.2), and 7–31.5<sup>h</sup> salinity (optimally at 14–17.5h salinity). Utilizes ferrous iron as an energy source for lithotrophic growth. Does not utilize thiosulfate, sulfide, pyruvate, glucose, or acetate as an energy source. Produces extracellular dreadlock-like iron oxides around the cell. The doubling time under optimal conditions is 27 h. The type strain is CP-8<sup>T</sup> , isolated from redox-stratified waters in the Chesapeake Bay, United States. The total DNA G+C content of the type strain is 53.7 mol%.

#### AUTHOR CONTRIBUTIONS

CC conceived of and directed the study, contributed to genome analysis, and wrote the paper. BC performed Fe(II) oxidation growth curve experiments, analyzed genomes, and

#### REFERENCES


wrote this paper. SK isolated strains CP-5 and CP-8, performed physiological growth tests and microscopy, prepared the strains for sequencing, and wrote portions of the paper. SM helped plan and prepare DNA for sequencing, performed genome quality control, and performed phylogenetic analyses. EF initiated and advised on genome analysis. All coauthors edited the manuscript.

#### FUNDING

This research was supported by NASA Exobiology Program Grant NNX12AG20G, NSF CAREER Geobiology grant EAR-1151682, and NSF Biological Oceanography grant OCE-1155290 (to CC), and a Grant-in-Aid for JSPS Overseas Research Fellow and Young Scientists (A)-16H06180 (to SK).

## ACKNOWLEDGMENTS

We would like to thank Deborah Powell and Shannon Modla for their assistance with SEM and TEM sample preparation and analysis. We also thank Brewster Kingham and Olga Shevchenko for advice and assistance in genome sequencing and analysis.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.01280/full#supplementary-material



zetaproteobacterium isolated from a deep-sea hydrothermal field at the Bayonnaise knoll of the Izu-Ogasawara arc, and a description of Mariprofundales ord. nov. and Zetaproteobacteria classis nov. Arch. Microbiol 199, 335–346. doi: 10.1007/s00203-016-1307-4


**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 Chiu, Kato, McAllister, Field and Chan. 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.

# Analysis of Bacterial Community Composition of Corroded Steel Immersed in Sanya and Xiamen Seawaters in China via Method of Illumina MiSeq Sequencing

Xiaohong Li1,2, Jizhou Duan<sup>1</sup> \*, Hui Xiao<sup>2</sup> \*, Yongqian Li1,2, Haixia Liu<sup>1</sup> , Fang Guan<sup>1</sup> and Xiaofan Zhai<sup>1</sup>

<sup>1</sup> Key Laboratory of Marine Environmental Corrosion and Biofouling, Institute of Oceanology, Chinese Academy of Sciences, Qingdao, China, <sup>2</sup> College of Marine Life Sciences, Ocean University of China, Qingdao, China

Edited by: Hongyue Dang, Xiamen University, China

#### Reviewed by:

Filomena De Leo, Università degli Studi di Messina, Italy Malin Bomberg, VTT Technical Research Centre of Finland, Finland

#### \*Correspondence:

Jizhou Duan jizhouduan@163.com Hui Xiao xiaoh28@163.com

#### Specialty section:

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

Received: 22 May 2017 Accepted: 25 August 2017 Published: 12 September 2017

#### Citation:

Li X, Duan J, Xiao H, Li Y, Liu H, Guan F and Zhai X (2017) Analysis of Bacterial Community Composition of Corroded Steel Immersed in Sanya and Xiamen Seawaters in China via Method of Illumina MiSeq Sequencing. Front. Microbiol. 8:1737. doi: 10.3389/fmicb.2017.01737 Metal corrosion is of worldwide concern because it is the cause of major economic losses, and because it creates significant safety issues. The mechanism of the corrosion process, as influenced by bacteria, has been studied extensively. However, the bacterial communities that create the biofilms that form on metals are complicated, and have not been well studied. This is why we sought to analyze the composition of bacterial communities living on steel structures, together with the influence of ecological factors on these communities. The corrosion samples were collected from rust layers on steel plates that were immersed in seawater for two different periods at Sanya and Xiamen, China. We analyzed the bacterial communities on the samples by targeted 16S rRNA gene (V3–V4 region) sequencing using the Illumina MiSeq. Phylogenetic analysis revealed that the bacteria fell into 13 phylotypes (similarity level = 97%). Proteobacteria, Firmicutes and Bacteroidetes were the dominant phyla, accounting for 88.84% of the total. Deltaproteobacteria, Clostridia and Gammaproteobacteria were the dominant classes, and accounted for 70.90% of the total. Desulfovibrio spp., Desulfobacter spp. and Desulfotomaculum spp. were the dominant genera and accounted for 45.87% of the total. These genera are sulfate-reducing bacteria that are known to corrode steel. Bacterial diversity on the 6 months immersion samples was much higher than that of the samples that had been immersed for 8 years (P < 0.001, Student's t-test). The average complexity of the biofilms from the 8-years immersion samples from Sanya was greater than those from Xiamen, but not significantly so (P > 0.05, Student's t-test). Overall, the data showed that the rust layers on the steel plates carried many bacterial species. The bacterial community composition was influenced by the immersion time. The results of our study will be of benefit to the further studies of bacterial corrosion mechanisms and corrosion resistance.

Keywords: bacterial community, MIC, carbon steel, Illumina MiSeq sequencing, 16S rRNA gene

## INTRODUCTION

fmicb-08-01737 September 12, 2017 Time: 16:30 # 2

Structural steel is widely used in marine environments because it is strong, readily available, easy to fabricate, and cost-effective, overall. However, steel is subject to corrosion. This is a serious worldwide problem and has a great social and economic impact (Hou et al., 2017). Corrosion is caused by complex chemical, physical and biological processes (Kip and Veen, 2015). Biological (in fact, microbiological influenced corrosion MIC) plays a critical role (Baboian, 2005). MIC is caused by electrochemical reactions created by those microorganisms that form 'biofilms' on immersed steel structures (Hamilton, 1991). Fungi are closely associated to this process (e.g., Arthrinium phaeospermum, Aspergillus niger, Chrysosporium merdarium and acidotolerant black yeast) (Lugauskas et al., 2009; Leo et al., 2013). Lugauskas et al. (2009) found that various strains of the same fungal species have different influences on submerged metal surfaces. However, bacteria are the main component of the biofilms, and contribute most to MIC (Bermont-Bouis et al., 2007) and the formation and transformation of corrosion products (Sun H. et al., 2014). The metabolic activities of bacterial communities within the biofilms interact with environmental factors, such as dissolved oxygen, pH, organic, and inorganic compounds, etc., to influence the electrochemical state of the metal and influence the rate of corrosion (Beech, 2004; Beech and Sunner, 2004; Coetser and Cloete, 2005; Videla and Herrera, 2005). It is also known that the bacterial surface associations within biofilms influence the electrochemical reaction rate (Dang and Lovell, 2016). Diverse bacterial populations can coexist in biofilms and often form synergistic communities (consortia) which contribute to the electrochemical processes via cooperative metabolic processes (Gonzalez-Rodriguez et al., 2008; Korenblum et al., 2008).

Some of the bacteria species that are associated with steel corrosion have been identified. They includes sulphate-reducing bacteria (SRB), sulphur-oxidizing bacteria (SOB), iron-reducing bacteria (IRB), and iron-oxidizing bacteria (IOB) (Sun J. et al., 2014), etc. SRB are regarded as the most influential (Duan et al., 2008), and are regarded as the main corrosion-accelerating factor in the context of the MIC of metals in marine environments (Angell and Urbanic, 2000). Other types of bacteria may also play an important role, e.g., methanogens and metal reducing-bacteria (Zhu et al., 2003; Gonzalez-Rodriguez et al., 2008). Moreover, what is interesting is that bacteria not only cause corrosion but can also inhibit or protect against corrosion, which is termed as MIC inhibition (MICI) (Zuo, 2007). There is currently a focus on exploiting bacteria and their metabolic by-products, including biofilm and extracellular polymeric substances (EPSs), to reduce MIC. The aim is to replace the biocides and toxic evaporative, organic compounds that are currently employed as rust retardants (Grooters et al., 2007). The mechanisms of MIC and MICI are not completely understood. They cannot be connected with a single biochemical reaction or a single bacterial species or cluster (Kip and Veen, 2015). It is therefore necessary, in this context, to learn more about the nature of the species complexes that form on corroding steel and rust that is immersed in seawater, so as to learn how to protect steel structures in marine environments.

Analyses of the bacterial communities of early developing biofilms in the rust layers of steel originally relied upon plate culturing techniques (Bermont-Bouis et al., 2007), which is laborious, imprecise, and time-consuming. Significantly, nearly all of the bacterial species from this environment do not reproduce on culture plates (Dunbar et al., 1999). Advances in molecular biology now permit us to analyze bacterial communities with considerable more precision. The techniques we adopted to investigate the composition of the bacterial communities were terminal restriction fragment length polymorphism (T-RFLP), denaturant gradient gel electrophoresis (DGGE), fluorescence in situ hybridization (FISH), and 16S rRNA gene libraries. Proteobacteria was recognized as the dominant bacterial group during the first 36 h of biofilm formation by using 16S rRNA gene libraries and T-RFLP (Lee et al., 2008). Citrobacter spp., Enterobacter spp. and Halanaerobium spp. were identified as the dominant bacteria of biofilms after 40-days immersion by ribosomal library and DGGE. FISH analysis was also used in the study of bacterial community composition, and the results showed that Alphaproteobacteria was the dominant community during the first few weeks of biofilm growth. In addition, it became apparent that the combination of FISH and confocal microscopy was of critical importance. It allowed us to define the relative importance of different bacteria in causing corrosion, and provided information both about the spatial structure of the corrosion biofilms, and quantitative information about the bacteria (Dang and Lovell, 2002a,b). Recently, high-throughput Illumina sequencing has been frequently used to investigate the bacterial community composition of various environments (Moreau et al., 2014; Sun J. et al., 2014; Chao et al., 2015). and allowed us to gain deeper insight into the bacterial community composition of the samples (Bokulich and Mills, 2012; Mayo et al., 2014). Our research was greatly enhanced by access to MiSeq sequencing which allowed us to obtain comprehensive information covering the composition of the bacterial communities we targeted. This follows Vigneron et al. (2016) who adopted this technique to reveal that Desulfovibrio species was the dominant bacteria on an offshore oil production facility. We consider that the application of this technology in the current area of research is in its infancy.

In this study we characterized the composition of the bacterial communities in corrosion samples that had been collected from rust layers on steel plates that had been immersed in seawater, by means of high-throughput Illumina MiSeq sequencing. In addition, we analyzed the influence of ecological factors on the bacterial communities. The results of our study have important implications for further study of bacterial corrosion mechanisms and anti-corrosion.

#### MATERIALS AND METHODS

#### Sample Sites and Collection

The plates of steel had the following composition (wt. %): C 0.16, Si 0.12, Mn 0.45, S 0.029, and P 0.019. Nine samples were collected in December 2014 for this study. Among them, six

samples (SE1, SE2, SE3, SE4, SE5, and SOH) were collected from the coastal zone of the Hongtang Bay which is located in Sanya City, Hainan Province. The sample identified as SOH provided us with rust layers from steel plates that had been immersed in seawater for 6 months. Samples identified as XE4, XE5 and XE6 were collected from the rust layers of steel plates that had been immersed in seawater for 8 years in a coastal zone of the island of Gulang, which is situated in Xiamen City, Fujian Province.

Large fouling organisms were removed with sterile forceps in sterile conditions from the steel plates as soon as they were removed from the sea. The surface of the test material was gently rinsed in sterilized seawater to remove unattached bacteria. The deposits were sampled with metallic spatulas, taking care not to crush the samples or expose them to air for too long. They were immediately placed in 10 ml sterile plastic centrifuge tubes, transported to the laboratory on dry ice, and were stored at −80◦C pending analysis (Païssé et al., 2013). Meanwhile, the salinity, temperature and pH of the seawater were measured by multiparameter water quality detector (CTD90M, Germany).

#### DNA Extraction

The total community genomic DNA of each sample was extracted according to the method of Zhou et al. (1996). Five microliter of each genomic DNA were subjected to 1% agarose gel electrophoresis to examine its integrity. The concentration of the DNA was measured with a UV-vis spectrophotometer (NanoDrop 2000c, United States) to identify that adequate amounts of high-quality total genomic DNA were extracted.

#### 16S rRNA Gene Amplification by PCR

V3–V4 region of the bacterial 16S rRNA gene was amplified by PCR (95◦C for 3 min followed by 27 cycles of 95◦C for 30 s, 55◦C for 30 s and 72◦C for 45 s and a final extension at 72◦C for 10 min using the primers 338F 5 0 -barcode-ACTCCTACGGGAGGCAGCAG-3<sup>0</sup> and 806R 5<sup>0</sup> -G GACTACHVGGGTWTCTAAT-3<sup>0</sup> (Dennis et al., 2013), where the barcode was an eight-base sequence that was unique to each sample. The PCR reactions were performed in triplicate in 20 µl reactions, containing 2 µl of 10× Ex Taq buffer, 2 µl of 2.5 mM dNTPs, 0.8 µl of each primer (5 µM), 0.2 µl Ex Taq polymerase, 0.2 µl of BSA, 14 µl of ddH2O and 10 ng of template DNA.

#### Illumina MiSeq Sequencing

The amplicons were extracted from 2% agarose gels and purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, United States) according to the manufacturer's instructions. The purified amplicons were quantified using QuantiFluorTM -ST (Promega, United States), pooled in equimolar ratios and subjected to paired-end sequencing (2 × 250) on an Illumina Miseq platform according to standard protocols. The raw reads were deposited into the NCBI Sequence Read Archive (SRA) database.

#### Processing Sequencing Data

The raw fastq files were demultiplexed and quality-filtered using QIIME (version 1.9.1) (Caporaso et al., 2010) with the following criteria: (i) The output data (reads) were truncated at any site receiving an average quality score < 20 over a 50 base pair (bp) sliding window. (ii) Primers were matched exactly allowing a two nucleotide mismatching, and reads containing ambiguous bases were removed. (iii) Sequences whose overlap was longer than 10 bp were merged according to their overlap sequence. Operational taxonomic units (OTUs) were clustered with a 97% similarity cut-off using UPARSE version 7.1<sup>1</sup> (Edgar, 2013). The normalization process followed OTU clustering. Chimeric sequences were identified and removed using UCHIME (Edgar, 2010; Edgar et al., 2011). The taxonomy of each 16S rRNA gene sequence was analyzed with RDP Classifier<sup>2</sup> (Wang et al., 2007) against the Silva (SSU128) 16S rRNA database using a confidence threshold of 70% (Quast et al., 2013).

The relative abundances of the phylum, class and genus levels were plotted as a bar graph. Heatmaps based on the relative abundance of OTUs at the phylum and genus levels were also generated with R program (R Development Core Team, 2013). A venn diagram was created using Mothur v.1.30.1 (Schloss et al., 2009) to identify the similarities and differences of the communities in the three kinds of samples (sample SOH, samples from Sanya, and samples from Xiamen). In alpha diversity analysis, alpha diversity parameters such as Chao, Ace, Simpson, and Shannon were estimated using mothur (version v.1.30.1<sup>3</sup> ) with a 97% similarity cut-off (Schloss et al., 2009). They provided a means of evaluating the potential total number of OTU and an estimate of the level of diversity in each sample. Rarefaction curves based on these metrics were generated. In beta diversity analysis, differences in the bacterial communities among the nine samples were preformed by a hierarchical cluster tree created using the unweighted pair-group method with arithmetic mean (UPGMA). A principal co-ordinates analysis (PCoA) plot was also obtained using Mothur with the calculation of Bray–Curtis (Schloss et al., 2009).

#### Data Accession Number

The obtained raw sequences were deposited in the NCBI database (Accession Number: PRJNA396473).

## RESULTS

#### Diversity Analysis and Richness of OTUs

A total 558,632 high-quality bacterial V3–V4 Illumina sequences, ranging from 47,920 to 77,230, were obtained for further analysis (**Table 1**). Data were normalized by subsampling the 16S rDNA data at 45,530 reads per sample to correct for unequal sequencing depth. The average length of the high-quality sequences from the nine corrosion samples was 441 bp. After random re-sampling at the 0.03 distance level, the average number of OTUs in the 8 years samples was 1,695. However, there were 6,020 OTUs in the sample immersed for 6 months. For the 8 years samples the average numbers were: OTUs 1,695, ACE 1,955 Chao1 index

<sup>1</sup>http://drive5.com/uparse/

<sup>2</sup>http://rdp.cme.msu.edu/

<sup>3</sup>http://www.mothur.org/wiki/Schloss\_SOP#Alpha\_diversity


TABLE 1 | Number of sequences analyzed, OTUs, estimated community richness indices (Chao and Ace), coverage, and community diversity indices (Shannon and Simpson) of the 16S rRNA libraries of the corrosion samples.

1826, Shannon and Simpson index 4.70 and 0.0378. For the 6 months sample the numbers were of OTUs 6,020, ACE 7,026 and Chao1 index 6,422, Shannon and Simpson index 7.11 and 0.0050 (**Table 1**). The rarified Chao1 and Shannon diversity indexes showed remarkable differences across the 8 years samples and the 6 months sample (P < 0.001, Student's t-test). Bacterial diversity and richness were higher in the 6 months immersion sample compared to the 8 years immersion samples, as described by the Shannon and Simpson diversity indices, ACE and Chao1 index. This was also confirmed by rarefaction curve analyses of the OTUs (**Figure 1**). Meanwhile, the average richness of 8 years immersion samples from Sanya (ACE index 2,249 and Chao1 index 2,092) was higher than that from Xiamen (ACE index 1,463 and Chao1 index 1,383). However, both of the rarified Chao1 and Shannon diversity indexes were not significantly different (P > 0.05, Student's t-test) across the 8 years samples from Sanya and Xiamen.

The Venn diagram showed that SE (8 years immersion samples from Sanya) and XE (8 years immersion samples from Xiamen) shared 1,180 OTUs, SE and SOH shared 1,887 OTUs, XE and SOH shared 622 OTUs and 480 OTUs were shared by all nine samples (**Figure 2**). The average number of OTUs in the 8 years samples from Sanya and Xiamen were 1,387 and 1,013, respectively. In addition, the Good's coverage values (**Table 1**) and the rarefaction curves of all corrosion samples (**Figure 1**) indicated that the 16S rRNA gene sequences derived from these corrosion samples could represent the total bacterial community in this study.

#### Analysis of Bacterial Communities

At the phylum level, more than 13 prokaryotic phyla were found in the nine samples accounting for 95.35% of the total community, namely Proteobacteria (63.44%), Firmicutes (19.12%), Bacteroidetes (6.28%), Tenericutes (1.57%), Actinobacteria (0.99%), Chloroflexi (0.86%), Thermotogae (0.81%), Cyanobacteria (0.54%), Acidobacteria (0.49%), Planctomycetes (0.36%), Spirochaetae (0.32%), Nitrospirae

in sample SE (five corrosion samples that were collected from Sanya and had been immersed in seawater for 8 years). There were 3,040 OTUs in sample XE (three corrosion samples that were collected from Xiamen and had been immersed in seawater for 8 years). There were 6,020 OTUs in sample SOH.

(0.34%) and lgnavibacteriae (0.23%) (**Figure 3** and **Table 2**). Proteobacteria, Firmicutes and Bacteroidetes were the core phyla, accounting for nearly 88.84% of the total. For the majority of corrosion samples, Proteobacteria was the dominant phylum, ranging from 26.82 to 82.93% of the total number of phyla. Firmicutes was the second most represented phylum, ranging from 0 to 62.14% of the total number of phyla. Bacteroidetes was the third most dominant phylum, ranging from 2.39 to 12.00% of the total number of phyla. However, in SE5, Firmicutes (62.14%) and Proteobacteria (26.82%) were the first and second most abundant phyla, which was markedly different to the distribution in the other samples. The remaining 10 phyla were represented at a low level on individual samples. Furthermore, the hierarchical clustering heat map of the in-depth taxonomic analysis was plotted to compare the membership and structure of each sample at the phylum level. It also indicated that Proteobacteria, Bacteroidetes and Firmicutes were the three dominant bacterial communities (**Figure 4**).

At the class level, more than 21 classes of prokaryote were found overall and accounted for 86.80% of the total community (**Figure 5** and **Table 3**). For the majority of corrosion samples, Deltaproteobacteria was the most abundant class, ranging from 16.50 to 71.56% according to the samples. Clostridia came second and ranged from 0 to 61.85% of the whole community. Gammaproteobacteria was the third most dominant class, ranging from 1.78 to 22.0% of the whole bacterial community. Some other classes (e.g., Alphaproteobacteria and Bacteroidia) also occupied a relatively large proportion of


the bacterial community composition, based on the average abundance analysis (**Figure 5** and **Table 3**). In addition, some classes, which occupied a relatively small proportion of the community composition, but which have been associated with corrosion (such as Zetaproteobacteria), were also found in this study (**Figure 5**). However, the community composition of some samples was unique at the class level. For example, Clostridia (61.85%) and Deltaproteobacteria (16.50%) were the first and second dominant classes in SE5. Bacteroidia (9.06%) and Gammaproteobacteria (12.59%) were the second dominant bacterial class in XE4 and XE6, respectively. Alphaproteobacteria was the third dominant bacterial classes in SE4 (15.48%) and SOH (19.91%). Furthermore, the hierarchical clustering heat map was also plotted to compare the membership and structure of each

sample at the class level. It also indicated that Deltaproteobacteria, Clostridia and Gammaproteobacteria were the dominant three bacterial communities among the top 50 classes across all the samples (**Figure 6**).

More than 56 genera were identified (**Figure 7** and **Table 4**). For the majority of corrosion samples, Desulfovibrio was the most abundant, ranging from 3.59 to 42.04% of the total number of genera. Desulfobacter came second (2.70–18.75%), and Desulfotomaculum was the third (0–56.04%). Other genera were well represented, based on the average abundance analysis (e.g., Sulfurimonas and Desulfonatronum) (**Figure 7** and **Table 4**). The generic profile of some samples was unique. For example, Desulfotomaculum (56.04%) and Desulfobacter (18.01%) were the dominant genera in SE5 and XE5, respectively. Sulfurimonas (10.31%) was the second most dominant genus in samples XE6. The hierarchical clustering heat map indicated that Desulfovibrio, Desulfobacter and Firmicutes were the dominant three bacterial genera among the top 100 genera (**Figure 8**).

#### Beta Diversity Analysis of the Nine Corrosion Samples

Two methods were adopted to analyze the beta diversity of the nine samples (**Figures 9**, **10**). Firstly, a hierarchical cluster tree of the bacterial communities was constructed by means of the UPGMA at a 97%-similarity OTU level. This showed that the data were clustered in two distinct groups (**Figure 9**). Group 1 contained the 6 months immersion sample (SOH) and one 8 years immersion sample (SE4). Group 2 included the other 8 years immersion samples. Afterwards, a principal coordinates analysis (PCoA) then targeted major bacterial clades, and confirmed the output of the first method, and explained 51.06% of the observed variation (**Figure 10**). Eight years


TABLE 3 | Microbial community compositions at class level.

immersion samples (except SE4) were grouped to the right of the graph along PC1. SOH was separated from the 8-years immersion samples and grouped to the left of the graph along PC1. Whereas SE4 was grouped in the middle of the graph between SOH and the other 8-year immersion samples. There was a clear distinction between SOH and the other corrosion samples along the first axis. Furthermore, bacterial communities were separated by the second axis. The results of the two methods indicated that the bacterial diversity (bacterial community composition) was clearly correlated to the immersion period. The sea area had no influence on the composition of the bacterial community.

## DISCUSSION

## Analysis of Composition of the Bacterial Community of the Corrosion Samples

Compared with samples from other environments, such as marine sediments (Liu et al., 2015), and seawater samples (Suh et al., 2015; Yang et al., 2015), the composition of the bacterial communities on the corrosion samples was similar at the phylum level, but significantly different at the genus level. In this study, Proteobacteria, Bacteroidetes and Firmicutes were the three core phyla in all samples. Proteobacteria was dominant in the majority of samples. This was also observed by Vigneron et al. (2016). They analyzed the bacterial community composition of corrosion samples taken from an offshore oil production facility. Proteobacteria also emerged as the dominant bacterial phylum in the initial stage of biofilm formation on carbon steel (Bermont-Bouis et al., 2007; Jones et al., 2007; Lee et al., 2008; Dang et al., 2011; McBeth and Emerson, 2016). We can point to a number of reasons for the dominance of Proteobacteria in rust samples. Members of this phylum are pioneer surface colonizers and important biofilm 'builders.' The 'facilitation' of biofilm formation is an important step in the further development of diverse populations, and their on-going stability (Slightom and Buchan, 2009; Dang et al., 2008, 2011). It is noteworthy that Proteobacteria is also the largest bacterial phylum and the most abundant across a range of environmental conditions (Liu et al., 2015; Qi et al., 2016) and in seawater (Suh et al., 2015; Yang et al., 2015; Mancuso et al., 2016).

Bacteroidetes was the second most abundant phylum. It was found to be dominant in biofilms formed on steel plates immersed in the sea for 40 days (Dang et al., 2011; McBeth


TABLE 4 | Microbial community compositions at genus level.

fmicb-08-01737 September 12, 2017 Time: 16:30 # 10

and Emerson, 2016). Bacteroidetes is a dominant phylum in marine environments (Kirchman, 2002), and has also been linked to biological corrosion. Bacteroidetes can also contribute to the survival of other of surface colonizers and the formation and development of biofilms (Dang et al., 2011). The composition and abundance of surface-associated bacterial colonies may be influenced by 'predatory' members of the phylum Bacteroidetes (Dang and Lovell, 2016). Bacteroidetes members are known to degrade complex biopolymers (Kirchman, 2002), which may assist in the creation of an aerobic environment with a biofilm, that is conducive to the growth of colonizing bacteria. Firmicutes was the third most dominant phylum in the majority of corrosion samples. It was found to be abundant in biofilms in the rust layer, based on 16S rRNA gene (Zhang and Fang, 2001; Luan et al., 2012). The presence of a Firmicutes member, Tindallia texcoconensis, isolated from lake Texcoco, Mexico by Alazard et al. (2007), was associated with hydrogen production, that provided for SRB. Some members of this phylum generate H2S and organic acids that can cause corrosion. For example, Acetobacterium carbinolicum produces acetic acid which can corrode steel (Paarup et al., 2006).

At the class level, Deltaproteobacteria was the dominant class in the majority of corrosion samples. This observation parallels information from studies of the bacterial communities of samples collected from water-flooded petroleum reservoirs, water injection systems of Brazilian offshore oil platforms, and corrosive petroleum reservoirs in Yangzhou (Korenblum et al., 2010; Li et al., 2016; Tian et al., 2017). There are many important SRB groups belonging to this class, e.g., Desulfovibrio, Desulfobacter and Desulfonatronum. Some species of SRB in the Deltaproteobacteria can promote the production of corrosive hydrogen sulfide from metallic sulfates (Kan et al., 2011). Clostridia was the second most abundant class in the majority of corrosion samples. This was similar observation to the results of a study of the composition of the bacterial community composition of biofilms from metal surfaces of an alkaline district heating system, and samples collected from water-flooded petroleum reservoirs (Kjeldsen et al., 2007; Tian et al., 2017). Some important SRB groups also belong to this class, for instance Desulfotomaculum. Some Clostridia produce acetic, butyric, or formic acids so that their presence may also lead to corrosion (Broda et al., 2000). Some are homoacetogens meaning that they convert carbon dioxide and hydrogen into acetate and propionate (Boga and Brune, 2003). The third most abundant class was the Gammaproteobacteria. Dang et al. (2011) reported that Gammaproteobacteria (mainly Alteromonadales and Oceanospirillales) are pioneer and long term surface colonizers, and can also contribute to the initiation and on-going development of biofilms. There are some other classes that were identified by this study that are known to contribute to steel corrosion, for instance Epsilonproteobacteria and Zetaproteobacteria (Dang et al., 2011; McBeth et al., 2011). Related research showed that Epsilonproteobacteria were the possible cause of microbial corrosion in pipelines injected with bisulfite (An et al., 2015).

At the genus level, three SRBs genera, Desulfovibrio spp., Desulfotomaculum spp. and Desulfobacter spp. formed a large proportion of the bacterial communities that were analyzed in this study. Desulfovibrio spp. were the most abundant. This complies with the data from a study of the corrosive marine biofilms of carbon steels immersed in seawater for 8 months (Bermont-Bouis et al., 2007). Desulfovibrio was also the most abundant bacterial genus in corrosion samples from oil pipelines in the Southeast of Mexico (Zhang and Fang, 2001; Neria-González et al., 2006; Vigneron et al., 2016). This agrees with previous studies that show that this genus is often the main cause of bacteria related corrosion (Miranda et al., 2006; Ilhan-Sungur et al., 2007). Members of genus Desulfovibrio are metabolically diverse and can reduce iron sulfate and, with hydrogen, produce H2S (Dinh et al., 2004). Significantly, the pH of an aquatic environment is modified by the presence of H2S, leading to the generation of a corrosion product, FeS in the presence of iron. The steel corrosion capacity of Desulfovibrio spp., such as D. vulgaris (Zhang et al., 2015, 2016), D. alaskensis (Wikieł et al., 2014) and D. desulfuricans (Lopes et al., 2006) has been extensively studied in laboratory experiments. Different mechanisms of corrosion development caused by Desulfovibrio spp. have been described and show that members of this genus have a interact with Clostridium species (Zhang and Fang, 2001), which was the second most abundant bacterial genus in sample SE4.

Desulfotomaculum spp. (the second most abundant genus) is a gram-positive SRB and is thermophilic. It plays an important role in MIC (Cetin et al., 2007) by accelerating cathodic depolarization

FIGURE 9 | A hierarchical cluster tree created using UPGMA with Bray–Curtis at the level of OTU. Microbial community distribution patterns at a 97%-similarity OTU level.

and decelerating anodic depolarization (Cetin and Aksu, 2009). The ability of members of this genus to corrode steel has been studied extensively in laboratory experiments: D. nigrificans (Mystkowska et al., 2015), D. orientis (Ren and Wood, 2004), and D. kuznetsovii (Anandkumar et al., 2015). Members of this genus are usually associated with oil, and have been isolated from the crude oil field, oil production wells, or even the cooling towers of a petroleum refinery (Cetin et al., 2007; Cetin and Aksu, 2009; Anandkumar et al., 2015). Desulfobacter spp., is a mesophilic, gram-negative genus with an oval morphology in the marine environment. Its ability to oxidize acetic acid is a characteristic (Widdel, 1988). It can also reduce organic substrates to CO<sup>2</sup> in a strictly anaerobic environment. However, the roles that Desulfobacter spp. play in steel corrosion are still unknown. Further study of its corrosive properties are needed. Some bacteria were found to inhibit MIC by the formation of a biofilm on the surface of steel. They included gramicidin-producing Bacillus brevis (Nikolaev and Plakunov, 2007), although Vibrio neocaledonicus may have the highest known level of corrosion inhibition (Moradi et al., 2015a,b). They did not appear in our results, but this might mean that they were present but at levels that were too low for detection, or that they were present at higher levels but were not detectable by techniques we adopted.

#### Comparative Analysis of Bacterial Community Composition

It is well-known that biofilm maturity significantly affects the bacterial communities of biofilms (Neria-González et al., 2006). The succession pattern of these bacterial communities is tied to the immersion time of the steel. The steel could be exposed to local acidification with a decrease in the redox potential over time. This might stabilize the conditions so that the anaerobes are better accommodated. Also, the increase in the local concentration of dissolved iron salts may affect the biofilm community. Although SRB were dominant in all of our samples, the bacterial community composition of samples immersed for 8 years was significantly different to that of the sample immersed for 6 months. The bacterial diversity of the 6-months sample was higher than that the others. This result was consistent with previous studies. Bermont-Bouis et al. (2007) reported that there was a big difference between the bacterial community composition of 8 months immersion samples and 1 month immersion samples. They found that SRB were also the dominant population in mature biofilms after an 8 months immersion, but Vibrio spp. (Gammaproteobacteria) was the main component in samples that had been immersed for 1 month (Bermont-Bouis et al., 2007). This may be because biofilms form in a consistent series of discrete steps, or as a time series, each being associated with a different bacterial community (Stoodley et al., 2002).

Oxygen is consumed throughout the formation of biofilms. For instance, members of the Bacteroidetes may contribute to a decrease the quantity of oxygen emitted: these bacteria degrade high-molecular weight organic matter (Kirchman, 2002). The reduction of oxygen in the biofilms generates the anaerobic environment, which is needed to induce SRB growth. Over time, an increasingly acidic and anaerobic environment develops, and this is believed to result in the succession of membership of the biofilm community. In this study, the bacterial diversity of the samples immersed for 6 months was higher than that of the other samples. We believe that the anaerobic environment formed after 8 years was more suitable for the growth of SRB than that of the samples that had been immersed for 6 months. This implies that the anaerobic environment formed after 8 years was clearly unsuitable for the aerobic bacteria (the early colonizers), which is the reason for there being a reduction in bacterial diversity over time. That SRB may be only a minor component at the initial stage of biofilms is supported by Dang et al. (2011). Earlier studies

have shown that the anaerobic zone will form underneath the upper aerobic layer when it is 10–25 mm thick (Coulter and Russell, 1976). At that point, the biofilm is clearly composed of a complex consortium of aerobic and anaerobic bacteria (Baker et al., 2003; Zhang et al., 2003). In addition, the bacterial community composition of biofilms is also changed most at the beginning of immersion (Dang and Lovell, 2000; Lee et al., 2008).

The methods of analysis can also have an impact on our understanding of the structure of bacterial communities. Proteobacteria was the dominant group for all corrosion samples no matter what methods were used. But the numbers of phyla and genera obtained from the corrosion samples were influenced by the methodology. We obtained more than 50 phyla and 100 genera by high-throughput Illumina sequencing (**Figures 4**, **8**): that is many more than have previously been detected (Luan et al., 2012; Chen et al., 2014). Equally important, the number of OTUs obtained by high-throughput sequencing was greater than had been revealed by the traditional methods (plating) and T-RFLP technique. 19,581 OTUs were found in the nine corrosion samples by high-throughput Illumina sequencing in this study, whereas only 64 OTUs and 24 OTUs were previously identified in corrosion samples by PCR-RFLP (Luan et al., 2012; Chen et al., 2014). The high-throughput Illumina sequencing method is clearly ideal because we were able to achieve in depth quantitative analyses of microbial communities (Bokulich and Mills, 2012; Mayo et al., 2014).

Many environmental factors can affect the composition of bacterial communities. In this study, the richness of immersed steel was related to the sea location. The average richness of 8-years immersion samples from Xiamen was numerically lower than that of Sanya, although the difference was not significant. Among the environmental factors, salinity, pH and temperature generally have a significant effect on the bacterial community composition. Parallel research has shown that saline water irrigation can change bacterial metabolic activities and community structures (Chen et al., 2017). Cell growth rate was inhibited by high salinity, but the viability and integrity of the bacterial membrane were increased (Kim and Chong, 2017). The functional structure of a bacterial community was significantly affected by pH (Joshi et al., 2017). However, in our study, there was little difference in salinity at the two locations (Sanya 33.97h and Xiamen 31.96h) and pH (Sanya 8.48 and 8.56). As the seawater temperature at Sanya (25.14◦C) was higher than that of Xiamen (19.27◦C), we speculate that temperature caused the small difference. Many studies have shown that temperature is a major influence on the composition of bacterial communities in the marine environment. The composition of the bacterial community of crude oil-contaminated marine sediments or seawaters were shown by Bargiela et al. (2015) and Meng et al. (2016) to be strongly linked to temperature. Even more important, studies have shown that temperature is also related to corrosion levels. In the aquatic system, temperature plays an important role in the changes of most biofilm parameters, and in their propagation and metabolism (Bott, 1996; Rao, 2010). In addition, the amount of bacteria (whether aerobic or anaerobic bacteria) in biofilms was also temperature dependant (Bott, 1996; Guo et al., 2006). It has been reported that the amount of bacteria in the rust layer of immersed carbon steel in Yulin station was more abundant than that of Qingdao station because of the different temperature (Guo et al., 2006). In this study, the seawater temperature at Sanya was nearly 5◦C higher than at Xiamen. The temperature of the Sanya coast was much more appropriate for the growth of bacteria. The degree of corrosion was enhanced by the presence of many more large fouling organisms in the warmer water. This resulted in the provision of higher levels of soluble nutrients provided by the decomposition of the other organisms. However, except for them, many other factors (like nutrients, dissolved oxygen and so on) could also influence the bacterial communities. The difference of the diversity was probably the result of integrated effects of the multiple factors. So far, it is hard to explain how the 6 months immersion sample (SOH) and 8 years immersion sample (SE4) were clustered in the same group in the multisample dendrogram. Our next step is to study the bacterial communities from samples from different substrates with a wider range of immersion times in a wider range of seawaters. This will greatly improve our knowledge of the relationships between environmental factors and bacterial community structure.

## CONCLUSION

The bacterial community composition of corrosion samples collected from rust layers of steel plates immersed in seawater for 6 months and 8 years at Sanya and Xiamen was revealed by means of Illumina MiSeq sequencing. We identified members of 13 phyla. Proteobacteria, Firmicutes and Bacteroidetes three dominated and accounted for nearly 89.03% of the total. Desulfovibrio spp., Desulfotomaculum spp. and Desulfobacter spp. were the core genera. The bacterial diversity from steel plate that has been immersed for 6 months was significantly higher than that taken from plates that had been immersed for 8 years. The average richness of biofilms removed from steel plates immersed for 8 years from Sanya was numerically but not significantly higher in similar samples taken from Xiamen at the same time. We identified bacteria that had not previously been found in this niche, although we do not know if they are involved in the corrosion of steel.

## AUTHOR CONTRIBUTIONS

The article and experiment done by XL, YL took part in the experiment, experimental design done by JD and HX. The other authors took part in the sample collection.

## ACKNOWLEDGMENTS

This work was supported by the National Basic Research Program (No. 2014CB643304) and National Natural Science Foundation of China (No. 41576080). We thank the Qingdao Research Institute for Marine Corrosion for providing 8-year immersion samples.

#### 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 © 2017 Li, Duan, Xiao, Li, Liu, Guan and Zhai. 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.

# Genome-Wide Detection of Small Regulatory RNAs in Deep-Sea Bacterium Shewanella piezotolerans WP3

Muhammad Z. Nawaz1,2, Huahua Jian<sup>1</sup> , Ying He1,2, Lei Xiong<sup>1</sup> , Xiang Xiao1,2 and Fengping Wang1,2 \*

<sup>1</sup> State Key Laboratory of Microbial Metabolism, School of Life Sciences and Biotechnology, Shanghai Jiao Tong University, Shanghai, China, <sup>2</sup> State Key Laboratory of Ocean Engineering, Shanghai Jiao Tong University, Shanghai, China

Shewanella are one of the most abundant Proteobacteria in the deep-sea and are renowned for their versatile electron accepting capacities. The molecular mechanisms involved in their adaptation to diverse and extreme environments are not well understood. Small non-coding RNAs (sRNAs) are known for modulating the gene expression at transcriptional and posttranscriptional levels, subsequently playing a key role in microbial adaptation. To understand the potential roles of sRNAs in the adaptation of Shewanella toward deep-sea environments, here an in silico approach was utilized to detect the sRNAs in the genome of Shewanella piezotolerans WP3, a piezotolerant and psychrotolerant deep-sea iron reducing bacterium. After scanning 3673 sets of 5<sup>0</sup> and 3<sup>0</sup> UTRs of orthologous genes, 209 sRNA candidates were identified with high confidence in S. piezotolerans WP3. About 92% (193 out of 209) of these putative sRNAs belong to the class trans-encoded RNAs, suggesting that trans-regulatory RNAs are the dominant class of sRNAs in S. piezotolerans WP3. The remaining 16 cis-regulatory RNAs were validated through quantitative polymerase chain reaction. Five cis-sRNAs were further shown to act as cold regulated sRNAs. Our study provided additional evidence at the transcriptional level to decipher the microbial adaptation mechanisms to extreme environmental conditions.

#### Edited by:

Martin G. Klotz, Queens College (CUNY), United States

#### Reviewed by:

Chiaki Kato, Japan Agency for Marine-Earth Science and Technology, Japan Brett Pellock, Providence College, United States

> \*Correspondence: Fengping Wang fengpingw@sjtu.edu.cn

#### Specialty section:

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

Received: 06 March 2017 Accepted: 30 May 2017 Published: 15 June 2017

#### Citation:

Nawaz MZ, Jian H, He Y, Xiong L, Xiao X and Wang F (2017) Genome-Wide Detection of Small Regulatory RNAs in Deep-Sea Bacterium Shewanella piezotolerans WP3. Front. Microbiol. 8:1093. doi: 10.3389/fmicb.2017.01093 Keywords: sRNA, Shewanella piezoloterans WP3, deep-sea, adaptation, gene regulation

## TOOLS DESCRIPTION:

QRNA, A tool used for the detection of the conserved RNA secondary structures, including both ncRNA genes and cis-regulatory RNA structures.

RNAz, A program for predicting structurally conserved and thermodynamically stable RNA secondary structures in multiple sequence alignments.

RNAalifold, A software used for the prediction of a consensus structure for a set of related RNAs. sRNAPredict3/SIPHT, sRNApredict3/SIPHT are recent versions of the sRNApredict suite that are used in the efficient prediction of sRNAs, with a high level of specificity. SIPHT is a compatible web version of sRNAPredict3 that searches approximately 1900 bacterial replicons from the NCBI database and predicts putative sRNA locations. sRNApredict3 is inclusive of sequence comparable options to look for conserved sRNAs.

sRNAscanner, The sRNAscanner is a computational tool used to detect the intergenic small RNA specific transcriptional units (TU) in the completely sequenced bacterial genome.

NAPP (Nucleic acid phylogenetic profiling), is a clustering method that efficiently identifies non-coding RNA (ncRNA) elements in a bacterial genome.

Rfam, The Rfam database is a collection of RNA families, each represented by multiple sequence alignments, consensus secondary structures and covariance models (CMs).

RNA Infernal, Infernal ("INFERence of RNA ALignment") is a tool for searching DNA sequence databases for RNA structure and sequence similarities.

CLUSTALW, ClustalW is a general purpose DNA or protein multiple sequence alignment program for three or more sequences.

PAM, PAM (point accepted mutation) matrices are used as substitution matrices to score sequence alignments in CLUSTALW.

FastTree, is open-source software to infer approximately maximum-likelihood phylogenetic trees from alignments of nucleotide or protein sequences.

## INTRODUCTION

Until the early 1990s, non-coding DNA was assumed to be non-functional and referred to as Junk DNA (Pagel and Johnstone, 1992). From the past two decades, these non-coding sequences were shown to act as modulators and regulators of gene expression in response to environmental signals (Gottesman, 2005; Vogel and Wagner, 2007; Waters and Storz, 2009), and were usually small in size (50–500 nucleotide [nt]) (Vogel and Papenfort, 2006; Gottesman and Storz, 2011) called small noncoding RNA (sRNA) and were shown to involve in a variety of biological processes, including quorum sensing, bacterial virulence, iron homeostasis, stress responses and so on (Masse and Gottesman, 2002; Lenz et al., 2004; Thomason et al., 2012). The sRNA-encoding genes are widespread in bacterial genomes (Luban and Kihara, 2007) and sRNAs can regulate the gene expression both at transcriptional and post-transcriptional level (Gottesman and Storz, 2011). The sRNA that binds to target mRNA can act either as cis or trans. Cis-encoded sRNA is typically encoded adjacent to its regulatory target on the same strand as a riboswitch or on the opposite strand to an antisense sRNA, with a perfect base-pairing region between their transcripts (Wagner et al., 2002; Brantl, 2007). On the other hand, trans-encoded sRNA is separated from the target gene, where an imperfect base-pairing often occurs between their transcripts (Gottesman, 2002, 2005; He and Hannon, 2004; Aiba, 2007). Trans-regulatory sRNAs are believed to be involved in the regulation of several biological processes including iron homeostasis and quorum sensing (Gottesman and Storz, 2011) while most of the cis-regulatory RNAs are known to maintain the appropriate copy number of the mobile elements (Wagner et al., 2002; Brantl, 2007). To date, almost all the sRNA species identified are encoded in the intergenic regions (IGRs) (Koo et al., 2011). IGRs are under lower selection pressure when compared to the rest of genomic regions, allowing more room to mutate and evolve in response to various environmental stimuli.

Since the first discovery of sRNA using electrophoresis in Escherichia coli in 1967 (Hindley, 1967), recent advances in 2D-gel electrophoresis, Northern blotting, direct labeling and RNA sequencing, DNA microarray, shotgun sequencing, co-purification, and genomic SELEX (Systematic Evolution of Ligands by Exponential enrichment) have led us to enhanced identification of sRNAs (Vogel and Sharma, 2005; Vogel et al., 2005; Huttenhofer and Vogel, 2006; Altuvia, 2007; Livny and Waldor, 2007; Liu et al., 2009). Previous sRNA identification studies have mainly been carried out on extensively studied bacteria including E. coli (Hershberg et al., 2003), Salmonella (Hebrard et al., 2012), Pseudomonas aeruginosa (Livny et al., 2006), Staphylococcus aureus (Marchais et al., 2009), Synechocystis PCC 6803 (Voss et al., 2009), Burkholderia pseudomallei (Khoo et al., 2012), B. cenocepacia J2315 (Ramos et al., 2012), Clostridium difficile (Soutourina et al., 2013), Brucella abortus 2308 (Dong et al., 2014), and Listeria monocytogenes (Toledo-Arana et al., 2009). E. coli has been the most studied microbe in this context, almost 80 sRNAs, including 30 Hfq (host factor for Q2)-dependent ones, have been validated using various experimental approaches such as Northern blot and microarray (Altuvia, 2007; Waters and Storz, 2009). Hfq belongs to the large family of Sm and Sm-like proteins (a family of RNA binding proteins), that promotes the binding between sRNA and its target mRNA through conserved sequence motif (Møller et al., 2002). sRNAs such as RyhB (regulate iron homeostasis) have been characterized using microarrays. OxyR (oxidative stress induced RNA), and CsrB (carbon storage regulator) were discovered by co-purification with overproduced CsrA protein (Altuvia et al., 1997; Romeo, 1998). Growth phase dependent sRNA genes in E. coli and S. aureus were identified using DNA microarray along with comparative genome analysis (Wassarman et al., 2001; Pichon and Felden, 2005; Silvaggi et al., 2006). The sRNAs have also shown to play regulatory roles in response to fluctuating conditions, for instance, RyhB has shown to involve in regulation of tricarboxylic acid cycle under iron-limiting conditions, through modulating the fur (ferric uptake regulator) gene expression in E. coli (Modi et al., 2011; Salvail and Masse, 2012; Michaux et al., 2014).

However, experimental methods are tedious and time-consuming. Moreover, expressions of sRNAs are condition-dependent (Stubben et al., 2014), therefore, experimental verification of sRNAs are less effective and inconclusive (McHugh et al., 2014). As a result, many of the predicted sRNAs could not be verified using experimental methods (Gottesman and Storz, 2011). Alternatively, with the availability of in silico sRNA prediction algorithms, computational screening of sRNAs in a large/genomic scale becomes efficient and complementary to experimental approaches (Livny and Waldor, 2007; Khoo et al., 2012). Biocomputationally predicted sRNAs are subsequently validated through experiments (Argaman et al., 2001; Rivas and Eddy, 2001; Wassarman et al., 2001). Recently, computational tools based on different features, such as RNA secondary structures,

thermodynamic stability, conservation of sequence and structure, transcriptional termination signals, and non-coding sequence clusters based on cross-genome conservation profiles (Vogel and Sharma, 2005; Lu et al., 2011), have greatly facilitated the efficient prediction of sRNAs in diverse bacterial species (Lu et al., 2011). Some of the widely used de novo search tools (**Table 1**) include QRNA (Rivas and Eddy, 2001), RNAz (Washietl and Hofacker, 2004), sRNAPredict3/SIPHT (Livny and Waldor, 2007), sRNAscanner (Aziz et al., 2010), RNAalifold (Bernhart et al., 2008), and NAPP (Nucleic acid phylogenetic profiling) (Ott et al., 2012). RNAz predicts evolutionarily conserved and thermodynamically stable RNA secondary structures in multiple sequence alignments, which is not only accurate as compared to other available tools (Xu et al., 2009) but also efficient as well (Washietl et al., 2005). Knowledge-based approaches, with homologs of identified sRNAs for profiling, can be used as complementary to de novo ones. RNA Infernal (Nawrocki and Eddy, 2007), one of the knowledge-based sRNA identification tool, together with a de novo tool RNAz were used in this study.

Not all predicted sRNAs can be verified by experimental techniques due to the fact that the functioning of sRNAs is condition-dependent. Due to the difficulty in experimental verification and characterization, only a small portion of computationally identified sRNA is subjected to testing. Up to date, a few studies on sRNA prediction and characterization have been conducted at a large scale, where at least over 100 sRNAs have been identified in each study (**Table 1**). In a study by Livny et al. (2006), in total 2759 sRNAs were predicted in Pseudomonas aeruginosa, but only 31 were tested and 17 were validated. According to Khoo et al. (2012), over 1300 sRNAs were identified in Burkholderia pseudomallei, 15 of which were tested and 8 were validated. Different computation tools have different sensitivities in generating sRNA candidates, varying from hundreds to thousands (**Table 1**), integrated use of tools provide a moderate and more accurate dataset for validation. The percentage of validated sRNAs relative to the total number of tested candidates can be as high as 100% in the case for B. cenocepacia J2315 (Ramos et al., 2012) and Synechocystis PCC6803 (Voss et al., 2009), and as low as 29% in Staphylococcus aureus (Marchais et al., 2009) (**Table 1**).

Shewanella are widely distributed aquatic organisms and one of the most abundant Proteobacteria in the deep sea, with the ability to grow on minimal medium and utilize a variety of compounds, such as iron, manganese, sulfite, oxygen, chromium, uranium, nitrate, fumarate, trimethylamine N-oxide (TMAO), and DMSO (Dimethyl sulfoxide), as terminal electron acceptors (Kato, 1999; Kato and Nogi, 2001). In addition, Shewanella has been shown to adapt to harsh and diverse environmental conditions with extreme temperature, pH, salinity, and pressure (Hau and Gralnick, 2007). Shewanella piezotolerans WP3 (hereafter referred as WP3), a piezotolerant and psychrotolerant Gram-negative gammaproteobacteria, was isolated from west Pacific deep-sea sediment at a depth of 1914 m (Wang et al., 2004). WP3 is considered as a good candidate for studying adaptation of Shewanella to deep-ocean (Xiao et al., 2007), as it grows in the pressure range 0.1–50 MPa with optimal growth at 20 MPa, and in the temperature range of 0–28◦C with optimum growth temperature of 20◦C (Xiao et al., 2007). WP3 has shown capable of adapting to a broad range of physical environmental conditions (Wang et al., 2008; Yang et al., 2013). Among the closely related species of WP3 from Shewanella genus, S. oneidensis MR-1 (referred as MR-1) is extensively studied and environmentally important species of the Shewanella genus because of its ability to use more than ten respiratory electron acceptors including nitrate, chromium, and uranium (Tiedje, 2002). S. psychrophila WP2 (referred as WP2) was isolated along with WP3 (Xiao et al., 2007), and shown to be a facultative anaerobic and psychrophilic, growing optimally at about 10–15◦C. S. violacea DSS12 (referred as DSS12) is a piezo- and psychrophilic deepsea bacterium which grows optimally at 8◦C and pressure of 30 MPa (Kato et al., 1995). DSS12 and S. benthica KT99 (referred as KT99), two piezophiles (Lauro et al., 2013), are the only species belonging to Shewanella genus which is found to be present at the depth of more than 2,000 m in the ocean (Wang et al., 2004). Details of these 26 sequenced genomes of Shewanella genus used in this study are available in Supplementary Table S1. In the present study, we used an in silico approach to detect the sRNAs in the genome of Shewanella piezotolerans WP3 and predicted sRNAs were further characterized, validated through transcriptions and quantitative polymerase chain reaction (qPCR). Discovery of novel sRNAs in WP3 and studying their conservation patterns across deep-sea bacterial lineages will lead us to elucidate gene regulation



and molecular mechanisms of bacterial adaptation to extreme deep-sea environmental conditions.

#### MATERIALS AND METHODS

#### Intergenic Region Extraction and Orthologs Prediction

As bacterial regulatory motifs were found to be encoded in IGRs of the genome, each IGR dataset was generated by extracting sequences from both 5<sup>0</sup> and 3<sup>0</sup> -untranslated regions (UTRs) of each gene in WP3 (described in details below), together with its orthologous IGRs in other reference genomes (Thomason and Storz, 2010). IGRs from 5<sup>0</sup> and 3<sup>0</sup> -UTRs of the orthologous sets of genes were extracted and considered as orthologous sets of IGRs. Orthologous relationship of genes across all the reference genomes was predicted by using OrthoMCL (Li et al., 2003). Afterward, IGRs of each gene belonging to same orthologous groups (OG) in WP3 and reference genomes were retrieved as a set of orthologous IGR.

For each IGR dataset from 5<sup>0</sup> -end, up to 250 nt upstream (depending upon the length of IGR) and 20 nt downstream of the translation initiation sites of the corresponding genes in the positive direction were extracted and considered as one orthologous set of 5<sup>0</sup> -UTRs and subject to RNA secondary structure prediction. For the purpose of increased specificity and sensitivity, three overlapping windows, that is −250∼−100, −200∼−50 and −150∼+20 (1 corresponds to the translation start site, "−" referred to upstream, and "+" to downstream) were generated from each original IGR dataset and analyzed separately. For genes transcribing in the negative direction, 250 nt downstream and 20 nt upstream of translation initiation sites were extracted and divided into three overlapping windows in the same manner. Similarly, for the extraction of orthologous IGR datasets from 3<sup>0</sup> -end, up to 250 nt (depending upon the length of IGR) sequences downstream and 20 nt upstream of the translation initiation sites of genes in the positive direction were retrieved as one orthologous set and subject to RNA secondary structure prediction. This IGR region was divided into three overlapping windows that are −20∼150, 50–200, 100–250. For genes transcribing in the negative direction, 250 nt upstream and 20 nt downstream of translation initiation site were extracted and divided into three overlapping windows.

#### sRNA Prediction by RNAz and RNA Infernal

We applied a comparative genomics-based approach for prediction of bacterial sRNA by integrating the RNA secondary structure prediction tool RNAz (Washietl et al., 2005) and RNA motif searching tool RNA-Infernal<sup>1</sup> . RNAz is widely used and comparatively reliable for de novo detection of structured non-coding RNAs from comparative genomics analysis (Washietl et al., 2005; Gruber et al., 2010). This program predicts structurally conserved and thermodynamically stable RNA secondary structures from multiple sequence alignments, with training data from multiple Rfam families and RNAalifold (Bernhart et al., 2008) for common RNA structure prediction. RNA Infernal is a homology-based tool that searches sequence databases for homologs of structural RNA sequences and then generates structure-based RNA sequence alignments. As stated above, IGRs were extracted from 5<sup>0</sup> and 3<sup>0</sup> -UTRs of genes in target genome and its orthologous regions in reference genomes. Afterward, multiple sequence alignment for each IGR orthologous dataset was performed with CLUSTALW (Thompson et al., 1994) (using PAM substitution matrix). RNAz was then applied on multiple aligned sequences, with default parameters. Orthologous UTR sets that were predicted as RNA by RNAz were searched in Rfam database using the RNA motif searching software Infernal (Nawrocki and Eddy, 2007), where highly confident candidates were those, which predicted with an infernal score higher than 10 bit (CM score > 10).

#### Conservation of Predicted sRNAs

We used the same approach to find the conservation or genome specificity of predicted cis-sRNAs in few closely related species of Shewanella genus and well-studied piezophilic and psychrophilic bacteria. We retrieved the 5<sup>0</sup> -UTR of a gene in WP3 and 5<sup>0</sup> -UTRs of orthologs of this gene in other species of Shewanella genus or other genera as a set of orthologous IGR. We applied RNAz on these orthologous sets of IGRs, positively predicted sets by RNAz were further subject to RNA Infernal. If the sequence in the set of orthologous IGR that belongs to 5<sup>0</sup> -UTR in WP3 and have infernal score higher than 10, it was declared as sRNA in WP3. Other sequences belonging to other species in the same set of orthologous IGR were considered as conserved if they also gave an infernal score higher than 10. These predicted sRNAs are conserved sRNAs because they are orthologous to each other as they were found at the 5<sup>0</sup> -UTR of the orthologous genes.

#### Phylogenetic Tree Construction

Maximum-likelihood tree based on 16S rRNA gene sequences was constructed to represent the taxonomic relationships of the species used in this study and tree was used to demonstrate the conservation pattern of 16 identified cissRNAs across species used in present study. 16S rRNA gene sequences of all the representative species were aligned using CLUSTALW and tree was constructed with FastTree using default parameters. Bootstrap values are based on 1000 replicates and are shown with white (≥80%) and black (≥90%) circles (**Figure 1**).

#### Culture Conditions and RNA Isolation

Shewanella piezotolerans WP3 was cultured in a modified marine medium 2216E (5 g/liter tryptone, 1 g/liter yeast extract, 0.1 g/liter FePO4, 34 g/liter NaCl) (Xiao et al., 2007). For aerobic cultivation, the single clone of WP3 strains were firstly inoculated into a 5 ml test tube, then the culture was diluted 1000 fold in the same medium with shaking (220 rpm) at 0.1 MPa (1 atm) and 20 and 4◦C, respectively. For anaerobic cultivation,

<sup>1</sup>http://eddylab.org/infernal/

media was prepared without any electron acceptor under nonsterile condition. Prepared media was dispensed into serum bottles filled with O2-free nitrogen, and bottles were covered with stoppers. Metal seals were used to seal the caps and media was autoclaved by inserting a needle into the stopper. Needles were plucked off immediately after the autoclave was done. Then serum bottles were filled by a gassing manifold system after the media was cooled down (Balch and Wolfe, 1976). Nitrate solution (0.4 M, sodium nitrate) was filter sterilized in a vinyl anaerobic airlock chamber (Coy Laboratory Products Inc., Grass Lake, MI, United States) and added to the concentration needed. The serum bottles, stoppers, and metal seals were bought from Wheaton Science Products (Millville, NJ, United States). In order to incubate WP3 at high pressure, log phase cultures of WP3 cells at atmospheric pressure were grown and then diluted 1000-fold with the same medium for anaerobic cultivation. After that cells were transferred into sterile injection syringes and were placed inside the pressure vessels. The syringes were then incubated at a hydrostatic pressure of 20 MPa (200 atm) at 4◦C. Pin closure pressure vessels were used in this study (Feiyu Petrochemical Instrument Equipment Inc., Nantong, China). Pressure was applied using a hand-operated pump with a quick-fit connector to the pressure vessel.

The growth of the WP3 strains was determined using turbidity measurements at 600 nm with a spectrophotometer (UV-2550, Shimadzu, Kyoto, Japan). The culture of WP3 was collected immediately when the cells reached mid-exponential phase (OD600≈0.8 and 0.2 for aerobic and anaerobic cultivation, respectively). The samples were centrifuged for 30 s at the maxim speed (16000 × g). The cells were immediately frozen in liquid nitrogen for subsequent RNA extraction.

TRI reagent-RNA/DNA/protein isolation kit (MRC, Cincinnati, OH, United States) was used to isolate the total RNA. After treating the RNA samples with DNase I at 37◦C for 1 h, RNA was purified by using RNeasy Mini Kit (Qiagen, Hilden, Germany). The quality of RNA samples were determined by visualizing the nearly 2:1 ratio of 23S:16S ribosomal RNA by running a 1.0% TAE (Tris-Acitate-EDTA) agarose gel. The total RNA was treated with DNase I at 37◦C for 1 h to remove DNA contamination and the purity was checked by PCR amplification with RNA as template. The quantity and integrity of RNA was evaluated with a UV spectrophotometer (Thermo Fisher, Waltham, MA, United States). In general, the ratio of 260 nm/280 nm > 2 and 260 nm/280 nm≈2.2 indicate the RNA is pure and could be used for the follow-up analysis. cDNA was synthesized from the purified RNA samples with the RevertAid First Strand cDNA Synthesis Kit (Fermentas, Glen Burnie, MD, United States).

#### Microarray Analysis of Differentially Expressed Genes at Low Temperature

The expression profiles of genes in WP3 genome at low temperature (4◦C compared to 25◦C) were studied using microarray data (NCBI GEO dataset accession: GSE80668). A microarray that contained 95% of the total predicted genes of WP3 was designed and manufactured (CapitalBio, Beijing, China). Differentially expressed genes (DEGs) were the ones with a significant change (fold ≥ 2) in expression pattern. Microarray signals with P-values < 0.001 in the F-test were regarded as DEGs. All of the DEGs were confirmed with the Significance Analysis of Microarrays (SAM) software. Detailed descriptions of the microarray procedure for WP3 under cold temperature was described somewhere else (Jian et al., 2016).

#### Real-Time qPCR

Primer Express software (ABI) was used to design the primer pairs for the selected genes for real-time qPCR (qPCR). PCR cycling was conducted using 7500 System SDS software (ABI, Foster City, CA, United States) in reaction mixtures with total volumes of 20 µl containing 1× SYBR Green I Universal PCR Master Mix (ABI, Foster City, CA, United States), 0.5 µM each primer, 1 µl cDNA template. The amount of target was normalized to that of the reference gene swp2079, whose expression remains constant under various conditions relative to the calibrator (The transcription levels of the genes at 20◦C and 0.1 MPa were set as 1) (Li et al., 2006). qPCR assays were performed in triplicate for each sample, and a mean value and standard deviation were calculated for the relative RNA expression

levels. Primers designed for qPCR are shown in Supplementary Table S2.

#### RESULTS

sRNAs are considered to be evolutionarily conserved in their secondary structures among the closely related species (Livny and Waldor, 2007). Here, we used a comparative genomicsbased approach for genome-wide screening of sRNAs. At present, various species of Shewanella genus from various environments have been sequenced. In this comparative genomics-based approach, 26 representatives sequenced genomes of Shewanella genus were selected based on niche and phylogenetic relatedness and being well-studied and were used as reference genomes, including one partial sequenced genome (Shewanella benthica KT99), being closely related to WP3. We have also included a genome sequence of Shewanella psychrophila WP2 isolated previously along with WP3 (Wang et al., 2004). This unpublished completely sequenced genome was also studied and discussed here in this study. In addition, we also investigated the distribution pattern of the identified cis-regulatory RNAs in WP3, across 10 bacterial species for piezophilic and psychrophilic adaptation features, including four selected Shewanella species (MR-1, WP2, DSS12, and KT99), three piezophilic model bacteria [Photobacterium profundum SS9 (Vezzi et al., 2005), Marinitoga piezophila KA3 (Lucas et al., 2012), Desulfovibrio piezophilus C1TLV30 (Khelaifia et al., 2011), referred as SS9, KA3, and C1TLV30], and three well studied psychrophilic bacteria (Colwellia psychrerythraea 34H, Psychromonas ingrahamii 37, Psychrobacter arcticus 273-4, referred as Cp34H, Pi37, and Pa273-4).

#### sRNA Prediction in WP3 and Other Shewanella Species

Using an integrated approach 209 RNA motifs in WP3 were predicted as reliable sRNA candidates. For comparative analysis of occurrence of sRNAs and determining the conservation of sRNAs, genomes of closely related species including MR-1, WP2, DSS12, and KT99 were scanned for sRNAs identification, using the same approach (**Table 2**). From total sets of 5<sup>0</sup> and 3<sup>0</sup> - UTRs in WP3 and other four Shewanella species, about 3 to 8% were predicted as sRNA candidates. As cis-regulatory RNAs are often located in the 5<sup>0</sup> -UTR of the mRNA (Xu et al., 2009), predicted sRNAs from 5<sup>0</sup> -UTRs are considered as cisregulatory and sRNAs from 3<sup>0</sup> -UTRs as trans-regulatory RNAs (Papenfort et al., 2015). Cis-regulatory RNAs are mainly known to maintain the appropriate copy number of the mobile elements (Wagner et al., 2002; Brantl, 2007), while trans-regulatory RNAs are associated with almost every global response in bacteria (Gottesman and Storz, 2011). Because of their ubiquitous roles, trans-acting sRNAs are extensively studied and well characterized in Gram-negative bacteria (Man et al., 2011). However, compared to trans-sRNAs, cis-sRNAs are much less understood. In the present study, the proportion of predicted cis-encoded sRNA was lower than that of trans-encoded sRNAs in WP3, with a ratio for


cis- and trans-sRNA 1:12, and the roles of all the 16 predicted cissRNA (referred as cis1-16 herein) in WP3 will be discussed here in this paper.

## Conservation of Identified Cis-sRNA

The distribution pattern of the 16 identified cis-regulatory RNAs in WP3 across selected piezophilic and psychrophilic bacterial species (details in Materials and Methods) was displayed in **Figure 1**. Seven sRNAs (cis2, cis3, cis7, cis12, cis13, cis14, and cis16) were found limited to Shewanella genus, so they are more likely to be Shewanella specific sRNAs. Nine out of the 16 predicted cis-regulatory RNAs have been found universally conserved in at least half of all the reference genomes presented in this study (**Figure 1**). Only one piezophilic species Photobacterium profundum SS9 shared four conserved cis-encoded sRNA with WP3, while in the rest of two piezophiles (Marinitoga piezophila KA3 and Desulfovibrio piezophilus C1TLV30) no common sRNAs was identified. One of these four conserved cis-sRNAs in Photobacterium profundum SS9 (conserve with cis15 in WP3) was already identified in a study by Campanaro et al. (2012) (locus tag = PBPRA0551b). All the four psychrophilic species used in the present study shared conserved cis-sRNAs with WP3, where five sRNA (cis4, cis5, cis6, cis10, and cis15) were conserved in at least half of the all psychrophilic species used in this study (**Figure 1**).

## Function Inference from Genome Annotation

All 16 cis-regulatory RNAs were annotated against the known regulatory motifs in Rfam database (**Table 3**). Among the five cis-sRNAs conserved in psychrophilic bacteria, cis4 was annotated as asX2, an sRNA which appears to function in virulence (Schmidtke et al., 2012). cis5 was annotated as rimP with a possible role in the regulation of NusA protein (a protein which functions in 30S ribosomal subunit maturation) (Nord et al., 2009; Naville and Gautheret, 2010), cis6 as nsiR with a function in cell differentiation (Muro-Pastor, 2014). cis10 was with the best hit as Atu\_C9, with reported roles in growth (Wilms et al., 2012), while cis15 was with the best hit to STnc180 with unknown function (Sittka et al., 2008). According to the genome annotation of WP3, the 5<sup>0</sup> genes of both cis4 and cis5 is anthranilate synthase component I (TrpE), an enzyme involved in the conversion of chorismate to anthranilate, while the 5<sup>0</sup> gene of cis6 is 2-isopropylmalate synthase, which plays an important role in the biosynthesis of l-leucine and pyruvate metabolism (Webster and Gross, 1965; Cole et al., 1973). The 5<sup>0</sup> gene for cis10 is ATP phosphoribosyltransferase (biosynthesis of histidine), and for cis15 is aspartate kinase (phosphorylation of the amino acid aspartate). Locations of identified cis-sRNAs and their neighborhood genes in the genome of WP3 are shown in **Supplementary Figure S1**.

## Validation Using Microarray Data and qPCR

As cis-regulatory RNAs are responsible for regulating the gene expression of their 5<sup>0</sup> -associated genes, gene expression of identified cis-sRNAs and their associated genes at optimum and cold temperature conditions were determined using RT-qPCR and microarray data, respectively. Five cis-sRNA (cis4, cis5, cis6, cis10, and cis15) were appeared conserve in psychrophilic species and were considered to have a role in cold adaptation and later qPCR analysis showed their increased expression under all cold conditions (**Figure 2**), therefore regarded as cold-regulated sRNAs. Interestingly, 5<sup>0</sup> -genes of all the five cold-regulated sRNAs were shown to be up-regulated under cold temperature conditions (**Table 3**). As cis-regulatory RNAs in bacteria are typically located at the 5<sup>0</sup> -end of their target genes and are involved in regulating the gene expression of their 5<sup>0</sup> -associated gene (Weinberg et al., 2010). Therefore, it is inferred that these predicted cis-sRNAs present at the 5<sup>0</sup> -end of the gene are contributing toward regulating the expression of their 5<sup>0</sup> -associated gene. All the 16 cis-sRNAs were also validated through RT-qPCR except cis5, a possible explanation of this exception can be that it might be expressed under some other specific conditions. qPCR results showed that 9 cis-sRNAs (cis4, cis6-11, cis14-15), including four cold regulators were present with an increase in transcription under all three low-temperature conditions (**Figure 2**). As these cold-regulated sRNAs were found with increased transcription under cold temperature conditions consistent with their target regulatory genes, it shows that they have positive regulatory effect over their 5<sup>0</sup> -associated genes, i.e., in response to cold temperature condition, increase in the sRNA transcription takes place which promotes the transcription (up-regulation) of their target regulatory genes.

## DISCUSSION

In present study, we extracted the IGR both from 5<sup>0</sup> and 3 0 -UTR of the orthologous genes in WP3 and reference genomes and applied combination of RNAz and RNA Infernal tool to predict sRNAs in our target bacteria WP3 and four related species of Shewanella genus, i.e., S. oneidensis MR-1, S. psychrophila WP2, S. violacea DSS12, and S. benthica KT99. Positively predicted sRNAs based on thermodynamic stability by RNAz were subjected to motif searching tool RNA infernal. sRNA candidates present at the 5<sup>0</sup> -UTRs which were positively predicted by RNAz and giving an infernal score higher than 10 bit were called as "confident candidates" (Xu et al., 2009), and were subsequently selected for verification through RT-qPCR. In addition to CM bit Score (Covariance model), trusted cutoffs (TC) bit scores thresholds were also used in the model to evaluate sRNAs candidates. As TC thresholds are generally considered to be the score of lowest-scoring known true positive, about half of putative cis-sRNAs validated through RT-qPCR were missed, when TC bit score thresholds (–cut\_tc parameter for using CM's trusted cutoff thresholds) was used. In this study, we mainly focused on the less understood type of sRNAs (cissRNA) and reported that cis-regulatory RNAs might have a role in adaptation to extreme conditions. We also explored the conservation of cis-sRNAs in other reference species and their conservation pattern demonstrates that most of the sRNAs


Frontiers in Microbiology | www.frontiersin.org

TABLE 3


cis-encoded

 sRNAs in WP3 and changes in expression

 profiles of 5

0-associated

 genes of identified

cis-sRNAs

 under cold temperature

 conditions.

June 2017 | Volume 8 | Article 1093

fmicb-08-01093 June 14, 2017 Time: 16:45 # 8

tend to remain to conserve within the genus. Conservation pattern of cis-sRNAs showed that five of them were conserved not only in psychrophilic species from Shewanella genus but also in species belonging to other genera. Furthermore, we compared the expression of their 5<sup>0</sup> -associated genes at an optimum and cold temperature and surprisingly, expression of all the 5<sup>0</sup> -associated genes were found with an increase in expression at low temperature. Regulated genes of all the five conserved sRNAs in psychrophiles are involved in the biosynthesis and metabolism of amino acids, secondary metabolites, and antibiotics, suggesting that bacterium tends to increase the metabolism of amino acids, secondary metabolites, and antibiotics at cold temperature.

As most studied class of sRNAs, trans-regulatory RNAs were found to be associated with almost every global response in bacteria and were reported as an abundantly existing class of regulatory RNAs in bacterial genomes (Gottesman and Storz, 2011). Likewise, about 92% (193 out of 209) of the sRNAs identified in this study, belonging to the class trans-encoded RNAs, which reflects that trans-regulatory RNAs are the dominant class of regulatory RNAs in S. piezotolerans WP3. We screened the orthologs of trans-encoded sRNAs in Rfam database where most of the trans-sRNAs shared homology with known trans-regulatory RNAs in E. coli, S. enterica, Staphylococcus aureus, Cyanobacterium, Xanthomonas campestris, Vibrio cholera, Listeria monocytogenes, and Rhizobiales species performing a variety of biological roles. Detailed information about the trans-sRNAs in the genome of S. piezotolerans WP3 and their orthologs in Rfam are available in Supplementary Table S3.

In short, we identified sRNAs in a piezo- and psychrotolerant, iron reducing, deep-sea bacterium, Shewanella piezotolerans WP3 and four closely related species of WP3. In total, we identified 209, 217, 181, 268, and 157 sRNAs in WP3, MR-1, WP2, DSS12, and KT99, respectively. Out of 209 predicted sRNAs in WP3, 16 are the cis-sRNAs (cis1-16) and were further characterized, validated through qPCR. Seven of this 16 cissRNAs (cis2, cis3, cis7, cis12, cis13, cis14, and cis16) were found as Shewanella-specific while rest of nine cis-regulatory RNAs were shown to be universal in at least half of all the genomes presented in this study. Five cis-sRNA (cis4, cis5, cis6, cis10, and cis15) were found as regulators commonly present in most of psychrophilic species used in this study. Expression analysis of 16 cis-sRNAs and their target genes demonstrated significant change under low-temperature conditions. Although the present study does not provide insights into the functioning of all the identified sRNAs (including cis and trans) in Shewanella piezotolerans WP3, further analysis is required to reveal the possible roles of these sRNAs in WP3. However, our study also provides evidence that not only trans- but cis-sRNAs could also play their roles in adaptation to extreme conditions. It should be very useful to explore the conserved sRNAs in extremophiles adapted to diverse environmental conditions.

#### AUTHOR CONTRIBUTIONS

FW, XX, YH, and MN designed the experiment and analysis. MN designed the pipeline and performed the computational analysis. LX conducted experiments. MN, HJ, YH and FW wrote the manuscript, in consultation with all other authors.

#### ACKNOWLEDGMENTS

fmicb-08-01093 June 14, 2017 Time: 16:45 # 10

We are thankful to National Natural Science Foundation of China (Grant No. 31290232), China Ocean Mineral Resources R&D Association (Grant No. DY125-22-04), and National Natural Science Foundation of China (Grant No. 41576129, 91228201, 91428308) for supporting our research. Moreover Higher Education Commission (HEC), Pakistan is gratefully acknowledged for providing financial support to MN.

#### REFERENCES


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.01093/full#supplementary-material

FIGURE S1 | Identified cis-sRNAs in WP3 genomes and their flanking genes. Neighborhood genes are shown in green wedges and sRNAs are shown in blue colored round corner rectangles. Genes without any labels are the hypothetical proteins.




**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 Nawaz, Jian, He, Xiong, Xiao and Wang. 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.

# Biofilm Formation and Heat Stress Induce Pyomelanin Production in Deep-Sea Pseudoalteromonas sp. SM9913

Zhenshun Zeng<sup>1</sup>† , Xingsheng Cai<sup>1</sup>† , Pengxia Wang<sup>1</sup> , Yunxue Guo<sup>1</sup> , Xiaoxiao Liu<sup>1</sup> , Baiyuan Li1,2 and Xiaoxue Wang<sup>1</sup> \*

<sup>1</sup> Key Laboratory of Tropical Marine Bio-resources and Ecology, The South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, China, <sup>2</sup> Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Institute of Microbiology, Guangzhou, China

#### Edited by:

Charles Lovell, University of South Carolina, United States

#### Reviewed by:

Bin-Bin Xie, Shandong University, China Marinella Silva Laport, Federal University of Rio de Janeiro, Brazil

\*Correspondence: Xiaoxue Wang xxwang@scsio.ac.cn †These authors have contributed

#### Specialty section:

equally to this work.

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

Received: 27 May 2017 Accepted: 06 September 2017 Published: 21 September 2017

#### Citation:

Zeng Z, Cai X, Wang P, Guo Y, Liu X, Li B and Wang X (2017) Biofilm Formation and Heat Stress Induce Pyomelanin Production in Deep-Sea Pseudoalteromonas sp. SM9913. Front. Microbiol. 8:1822. doi: 10.3389/fmicb.2017.01822 Pseudoalteromonas is an important bacterial genus present in various marine habitats. Many strains of this genus are found to be surface colonizers on marine eukaryotes and produce a wide range of pigments. However, the exact physiological role and mechanism of pigmentation were less studied. Pseudoalteromonas sp. SM9913 (SM9913), an non-pigmented strain isolated from the deep-sea sediment, formed attached biofilm at the solid–liquid interface and pellicles at the liquid–air interface at a wide range of temperatures. Lower temperatures and lower nutrient levels promoted the formation of attached biofilm, while higher nutrient levels promoted pellicle formation of SM9913. Notably, after prolonged incubation at higher temperatures growing planktonically or at the later stage of the biofilm formation, we found that SM9913 released a brownish pigment. By comparing the protein profile at different temperatures followed by qRT-PCR, we found that the production of pigment at higher temperatures was due to the induction of melA gene which is responsible for the synthesis of homogentisic acid (HGA). The auto-oxidation of HGA can lead to the formation of pyomelanin, which has been shown in other bacteria. Fourier Transform Infrared Spectrometer analysis confirmed that the pigment produced in SM9913 was pyomelanin-like compound. Furthermore, we demonstrated that, during heat stress and during biofilm formation, the induction level of melA gene was significantly higher than that of the hmgA gene which is responsible for the degradation of HGA in the L-tyrosine catabolism pathway. Collectively, our results suggest that the production of pyomelanin of SM9913 at elevated temperatures or during biofilm formation might be one of the adaptive responses of marine bacteria to environmental cues.

Keywords: pyomelanin, temperature, Pseudoalteromonas, biofilm, pellicle formation

## INTRODUCTION

Biotic and abiotic surfaces in various marine environments are rapidly colonized by microorganisms, and surface colonization by marine microbes often involves biofilm formation, which aids survival in extreme marine environments (Dang and Lovell, 2016). During biofilm formation, marine bacteria often produced biologically active compounds to adapt the

extreme environmental conditions, such as high pressure, hydrothermal vent, and depletion of micronutrients (de Carvalho and Fernandes, 2010). Pseudoalteromonas is a genus of Gammaproteobacteria that is widespread in the world's ocean and have been shown to produce bioactive compounds that inhibit settling of several fouling invertebrates and algae during biofilm formation (Holmstrom and Kjelleberg, 1999; Egan et al., 2002; Ballestriero et al., 2010). Our recent work showed that biofilms formed by genetic variants of Pseudoalteromonas lipolytica with different capsular polysaccharide production or cellulose production varied with their ability to induce the larval settlement and metamorphosis of the mussel Mytilus coruscus (Zeng et al., 2015).

Production of the bioactive pigment is prevalent in marine bacteria and is important to cellular physiology and survival. Bioactive pigments with broad-ranging pharmacological activities has been received extensively studied, and it has been suggested that pigmented Pseudoalteromonas spp. produce more bioactive molecules (Bowman, 2007). Pseudoalteromonas spp. can be divided into pigmented and non-pigmented clades based on their ability to produce pigments. Of the twenty-three pigmented Pseudoalteromonas strains identified to date, three of them have been shown to produce melaninlike pigments, including Pseudoalteromonas nigrifaciens (Baumann et al., 1984), Pseudoalteromonas aliena (Ivanova et al., 2004), and Pseudoalteromonas distinct (Ivanova et al., 2004). A number of marine bacteria, such as Vibrio cholerae (Coyne and Alharthi, 1992), Shewanella colwelliana (Fuqua and Weiner, 1993), Cellulophaga tyrosinoxydans (Kahng et al., 2009), and Marinomonas mediterranea (Solano et al., 1997) also have been shown to produce dark-brown melanins. Melanins constitute a general class of complex polyphenolic heteropolymers that include eumelanin, pheomelanin, pyomelanin, and a number of pathways are known to covert tyrosine into melanin (Ito and Wakamatsu, 2003; Rzepka et al., 2016). Pyomelanin was produced by the accumulation of homogentisic acid (HGA), which is synthesized via 4-hydroxyphenylpyruvate dioxygenase (4-HDDP) in the process of L-tyrosine degradation; HGA is excreted to extracellular and auto-oxidized, followed by polymerized to form pyomelanin (Schmaler-Ripcke et al., 2009). The dark brown pigment synthesized by certain strains of Vibrio cholerae and Shewanella colwelliana is a type of pyomelanin derived from HGA (Coyne and Alharthi, 1992; Coon et al., 1994). Although melanin-like pigments are produced in different marine bacteria, little is known about the melanin precursors or biosynthetic enzymes as well as the regulation of melanin production in Pseudoalteromonas spp.

In this study, we studied a deep-sea derived non-pigmented Pseudoalteromonas sp. SM9913. SM9913 is a psychrotolerant bacterium and can grow at a wide range of temperatures from 4 to 37◦C (Qin et al., 2007). Comparative genomic analysis of SM9913 with the surface sea-water psychrophilic bacterium of Pseudoalteromonas haloplanktis TAC125 has revealed some special genetic features that may have allowed it to adapt to the sediment-attached lifestyle (Qin et al., 2011). SM9913 has both the polar and lateral flagella systems, and produces a large quantity of exopolysaccharides and proteases, which favor degrading the sedimentary particulate organic nitrogen and adopting a particle-associated biofilm lifestyle in the sediment (Chen et al., 2003; Mi et al., 2015; Liu et al., 2016). In this study, we first assessed the biofilm formation of SM9913 under various conditions. SM9913 formed attached biofilm and pellicles at 4–37◦C. Interestingly, we found that SM9913, considered to be a non-pigmented Pseudoalteromonas strain, produces a dark brown pigment when cultured at higher temperatures or when forming biofilm. We demonstrated that higher temperatures and biofilm formation induced pyomelanin production in SM9913 via accumulation of the melA gene product, 4-HPPD. Transcription of melA was induced more than three other genes (hmgA, maiA, and fahA) during the catabolism of L-tyrosine at higher temperatures and during biofilm formation, but not upon the addition of L-tyrosine to the culture medium at 15◦C. Knockout melA gene in SM9913 completely abolished pyomelanin production when cultured at higher temperatures, while ectopic expression of melA in melA knockout strain restored pyomelanin production at higher temperatures. Finally, knockout of melA reduced resistance to heat stress in SM9913. These findings demonstrate that pyomelanin production may be one of the adaptive strategies of marine bacteria.

## MATERIALS AND METHODS

#### Bacterial Strains, Plasmids, and Growth Conditions

The bacterial strains and plasmids used in this study are listed in **Table 1**. Pseudoalteromonas sp. SM9913 was a generous gift from Professor Yuzhong Zhang at Shandong University, China, and was originally isolated from deep-sea sediment at a water depth of 1855 m near the Okinawa Trough (Chen et al., 2002). SM9913 was routinely cultured in a nutrient-enriched medium SWLB (seawater Luria-Bertani: 10 g peptone, 5 g yeast extract, 1 L artificial seawater) or a nutrient-less marine broth 2216E at 15◦C (Becton, Dickinson and Company, Franklin Lakes, NJ, United States) (Qin et al., 2011). For Escherichia coli, experiments were conducted in LB medium at 37◦C. When needed, antibiotics were added to the medium at the following concentrations: erythromycin at 25 µg/mL and chloramphenicol at 30 µg/mL. DAP (2, 6-diamino-pimelic acid) was added to the medium at a concentration of 0.3 mM to culture E. coli WM3064 strain.

## Attached Biofilm and Pellicle Formation Assay

Attached biofilm formation was assayed in 96-well polystyrene plates using crystal violet staining as described previously (Pratt and Kolter, 1998). Briefly, cells were grown in SWLB or 2216E medium without shaking at different temperatures. Attached biofilm was measured at the indicated time points. To remove the growth effects, biofilm formation was normalized by dividing the total biofilm by the maximal bacterial growth as measured by turbidity OD<sup>620</sup> for each strain. Three independent cultures were used for each strain. To form the pellicle (air–liquid biofilm), cells were grown in SWLB or 2216E medium in glass beakers


Chloramphenicol (30 µg/mL) was used to maintain pCA24N-based and pBBR1MCS-Cm-based plasmids, while erythromycin (25 µg/mL) was used to maintain pK18mobsacB-Ery-based plasmids.

without shaking for the specified number of days at different temperatures. Pellicles were assayed by visual inspection of the air–liquid interface of the standing culture. Morphology was observed and photographed at the indicated time points.

#### Pigment Production and Quantification

Wild-type SM9913 strain was grown in SWLB or 2216E medium with or without the L-tyrosine or D-tyrosine at a concentration of 1 mg/mL for different days. E. coli K12 carries the pCA24N-melA or empty vector were grown in LB medium supplemented with 30 µg/mL chloramphenicol and 0.5 mM IPTG (isopropyl-beta-D-thiogalactopyranoside) for 2 days. The production of pigment was observed and photographed at the indicated time points. To measure the pigment production, the cultures were centrifuged at the 17000 g for 15 min, and the supernatants were collected and quantified by measuring the absorbance at 400 nm (Turick et al., 2002).

#### SDS-PAGE and Mass Spectrometry

Wild-type SM9913 strain was grown in SWLB medium at 15 or 37◦C for 24 h, respectively. The cultures of 5 mL were centrifuged at 13,000 rpm for 2 min and the supernatants were removed. Cell pellets were re-suspended in 1 mL lysis buffer [50 mM Tris (pH 8.0), 100 mM NaCl, and protease inhibitor cocktail (Sigma– Aldrich, St. Louis, MO, United States)]. Then samples were sonicated twice at level 2 for 5 min using a Sonic Dismembrator (Ningbo Scientz Biotechnology, Co., Ltd., Ningbo, China). The supernatant was collected after centrifugation at 13,000 rpm for 5 min and the protein concentration was measured using a Bi Yuntian BCA assay kit (Beyotime Biotechnology, Co., Ltd., Haimen, China). Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) was performed by loading 25 µg of each sample. Gels were stained with Coomassie Brilliant Blue R-250 and 10 protein bands were cut and subjected to in-gel tryptic digestion. In-gel digestion and mass spectrometry (MS) analysis of the gel slices were performed at the Research Center for Life Sciences University of Science and Technology (Hefei, Anhui, China) as previously described (Shevchenko et al., 2006).

## Fourier Transform Infrared Spectrometer Analysis

The pigment of SM9913 induced at 37◦C was extracted according to a previously published method with minor modifications (Schmaler-Ripcke et al., 2009). Briefly, SM9913 wild-type cells were grown in SWLB medium at 37◦C for 48 h. A 200 mL culture was collected and centrifuged at 13,000 rpm for 5 min. The supernatants were then filtrated through 0.45 µM filter membrane. The filtrate was acidified with 6 M HCl to a pH of 2.0, and it was allowed to precipitate overnight at room temperature. After centrifugation (16, 500 × g, 30 min), the pellet was re-suspended in 2.5 mL deionized water at pH of 12 and dialyzed in 3.5-kDa dialysis tubing for 24 h against deionized water. The dialyzed pigment was then dried using the vacuum centrifugal concentrator (Tokyo Rikakikai, Co., Ltd., Tokyo, Japan). Synthetic pyomelanin (HGA-melanin) was used as a control compound in Fourier Transform Infrared Spectrometer (FTIR) analysis. Synthetic pyomelanin was produced by autooxidation of a 10 mM HGA (Tokyo Chemical Industry, Co., Ltd., Tokyo, Japan) at pH 10 with constant stirring on a magnetic stirrer for 3 days (Thermo Fisher Scientific, Waltham, MA, United States). A 200 mL sample was collected and precipitated by adjusting the pH to 2 with 6 M HCl. Further treated is performed as described above. The pigment prepared from in vitro cultures of SM9913 and synthetic pyomelanin were analyzed by using FTIR spectrophotometer (IR Affinity-1, Kyoto, Japan).

## Construction of Strains and Vectors

In-frame deletion of a single gene in SM9913 was performed using the fusion PCR method we developed recently (Wang et al., 2015). Briefly, two primer pairs (melA-up-S/melA-up-A and melA-down-S/melA-down-A) were used to amplify the upstream and downstream of the target region from wild-type SM9913 genomic DNA. The resulting 842 and 836 bp fragments were digested with SphI/EcoRI and EcoRI/SalI, respectively, and then cloned into the SphI/SalI sites of the suicide plasmid pK18mobsacB-Ery. The constructed vector was then transformed

into the E. coli WM3064 strain and verified by DNA sequencing using primer pair pK18-F/pK18-R. The recombinant suicide plasmid was mobilized from E. coli WM3064 into SM9913 by intergeneric conjugation at 15◦C. Cells integrated the recombinant plasmid via a single crossover event was selected for by erythromycin resistance and confirmed by PCR using primer pair pK18-F/pK18-R. The deletion mutant was screened by plating the single-crossover strain on SWLB medium containing 15% sucrose. Further confirmation of the deletion mutant was carried out by PCR followed by DNA sequencing using primer pair melA-F/melA-R.

The broad-host-range plasmid pBBR1MCS-Cm was used for gene complementation (Wang et al., 2015). Gene fragment of melA was PCR amplified using SM9913 genomic DNA as template, and then inserted into the EcoRI/XhoI sites of pBBR1MCS-Cm. After verification by DNA sequencing using primer pair pBBR1MCS-F/pBBR1MCS-R, the expression plasmid was introduced into the 1melA strain by conjugation. The plasmid pCA24N was used to express the melA product in E. coli K12. The melA gene was amplified from SM9913 genomic DNA using the primer pair pCA24N-melA-F/pCA24NmelA-R. The PCR product was phosphorylated, purified, and ligated into vector pCA24N as previously described (Guo et al., 2014). The recombinant plasmid was verified by DNA sequencing using primer pair pCA24N-F/pCA24N-R and introduced into E. coli K12. All the primers used in this study are listed in Supplementary Table S1.

## Quantitative Real-time Reverse-Transcription PCR (qRT-PCR)

Cells were grown at different temperatures and collected at the exponential stage (turbidity ∼1.0 at OD600) or during biofilm formation. To test the effect of the tyrosine on the genes transcription, cells were grown in SWLB medium with or without the L-tyrosine or D-tyrosine at a concentration of 1 mg/mL. Total RNAs was isolated using Tiangen RNA prepPure Cell (Tigangen Biotech, Beijing, China) following the manufacturer's instructions. Complementary DNA was synthesized using the Reverse Transcription System (Promega, Madison, WI, United States) according to manufacturer's instructions. qRT-PCR was performed using the Step One Real-Time PCR system (Life Technologies, Carlsbad, CA, United States) using SYBR Green Mix (Thermo Fisher Scientific, Waltham, MA, United States). Primers for qRT-PCR for rrsG, melA, hsp90, hmgA, fahA, and maiA are listed in Supplementary Table S1. The transcript of the rrsG was used to normalize the total RNAs in different samples. A quantification method based on the relative amount of a target gene versus a reference gene was used (Pfaffl, 2001). Fold change of the target gene at different temperatures (T1 and T2) was calculated as 2ˆ−(Ct target\_T1−C<sup>t</sup> rrsG\_T1)/ 2ˆ−(Ct target\_T2−C<sup>t</sup> rrsG\_T1). The level ofrrsG transcript was used to normalize the gene expression data in different samples.

#### Heat Stress Assay

For wild-type SM9913 and melA mutant strains, cells were grown in SWLB medium and treated at 45◦C for 30 min in the exponential stage (turbidity ∼1.0 at OD600). Cell viability was determined by serial dilutions using 3.4% NaCl solution and plated on SWLB agar plates (Donegan et al., 1991). For E. coli K12 with an empty vector or expressing the melA product from SM9913, cells were grown in LB medium supplemented with 30 µg/mL chloramphenicol and 0.5 mM IPTG and treated at 65◦C for 10 min in the exponential stage (turbidity ∼1.0 at OD600). Cell viability was determined by serial dilutions using 0.8% NaCl solution and plated on LB agar plates.

## RESULTS

## SM9913 Forms Biofilm

Attached biofilm formation was examined using the 96-well polystyrene plate assay, which partially mimics an aquatic environment in which bacteria are attached to the surface of the particles (Lee et al., 2008). For clarity, we hereafter refer to this type of biofilm as the "attached biofilm." SM9913 grew at a wide range of temperatures from 4 to 37◦C (see growth curve in SWLB at 15, 25, and 37◦C in Supplementary Figure S1). Since the optimal growth temperature of SM9913 is 15◦C (Qin et al., 2011), and the doubling time of SM9913 was over 30 h in SWLB at 4◦C, 15◦C was chosen as the optimal temperature in this study. Attached biofilm of SM9913 was analyzed in a nutrient-enriched medium (sea water Luria-Bertani; SWLB), in which it formed attached biofilm after 1 day of incubation at 4, 15, 25, and 37◦C. Results showed that SM9913 formed the most abundant attached biofilm when cultured at 4◦C (**Figure 1A**). In SWLB or a nutrientless marine medium 2216E, SM9913 formed an initial attached biofilm within 1 day, which continued to grow until day 5 at the optimal growth temperature (**Figure 1B**). Moreover, SM9913 formed 1.9 ± 0.1-fold more attached biofilm in 2216E compared to SWLB after 5 days of incubation (**Figure 1B**), suggesting lower nutrient levels led to the formation of more attached biofilm. Static culturing of SM9913 in aerobic conditions can also lead to the formation of "pellicle" on the surface of static liquid medium, resembling the biofilm that formed at the liquid–air interface; for clarity, this type of biofilm is henceforth referred to as "pellicle." Visible pellicle was observed after 1 day incubation of SM9913 at 15◦C in SWLB, continued to grow until day 6 before gradually dispersed over time (**Figure 1C**). In contrast to attached biofilm formed at the solid–liquid interface, SM9913 formed pellicle in SWLB but not in 2216E over a period of 2 weeks, indicating that higher nutrient levels promoted pellicle formation (**Figure 1C**). Therefore, SM9913 has the capacity to form biofilm at both the solid–liquid and air–liquid interfaces, with lower temperatures and lower nutrient levels promoting the formation of attached biofilm and higher nutrient levels promoting the formation of pellicles at the air–liquid interface.

#### Pyomelanin Production Is Induced at Higher Temperatures

Wild-type SM9913 strain produced a dark brown pigment when grown in SWLB or 2216E at higher temperatures after prolonged incubation (e.g., 37◦C for 2 days) (**Figure 2A**, only the culture grown in SWLB was shown here). To explore the molecular

basis of the occurrence of the dark brown pigment at higher temperatures, cells grown at 37 and 15◦C were collected and the protein profiles were assessed by SDS-PAGE. A total of 10 differentially produced bands of various sizes were selected for mass spectrometry analysis (**Figure 2B** and Supplementary Table S2). Among them, two of the proteins induced at 37◦C were identified to be 4-hydroxyphenylpyruvate dioxygenase (4-HPPD) (band #7, ∼38 kDa) (Supplementary Figure S2) and heat shock protein 90 (HSP90) (band #3, ∼70 kDa). HSP90 is a chaperone protein that help other protein folding properly, and stabilizes proteins against heat damage (Rutherford and Lindquist, 1998). 4-HPPD is encoded by melA gene which catalyzes the conversion of 4-hydroxyphenylpyruvate into HGA during the catabolism of L-tyrosine. HGA can be auto-oxidized, and polymerized to form pyomelanin (Moran, 2005). The 4-HPPD protein of SM9913 (PSM\_A0972) shares 72% similarity (100% coverage) with the 4- HPPD protein of Shewanella oneidensis (Supplementary Figure S3), which is a key protein participate pyomelanin biosynthesis in S. oneidensis (Turick et al., 2009). To further verify that the pigment induced at 37◦C was pyomelanin, we performed FTIR to analyze the structure of the pigment which has been described previously for pyomelanin (Bilinska, 1996). Results showed that the mainly FTIR bands at the waves of 3, 282, 2, 926, 1, 519 cm−<sup>1</sup> as well as the fingerprint regions between 1, 000 and 500 cm−<sup>1</sup> shared highly similarity between the pigment extracts of SM9913 to the FTIR scans of pyomelanin as previously reported from Shewanella algae (Turick et al., 2002) and Aspergillus fumigatus (Schmaler-Ripcke et al., 2009) (Supplementary Figure S4). In addition, synthetic pyomelanin produced by using commercially purchased HGA (Tokyo Chemical Industry, Co., Ltd, Japan) to synthesize pyomelanin via in vitro auto-oxidation was analyzed by FTIR and they also shared high similarity with the pigment extracted from SM9913 (Supplementary Figure S5).

Consistent with increased protein expression of 4-HPPD and HSP90 at 37◦C, the transcription of melA and hsp90 were induced upon the temperature upshift by sixfold and ninefold, respectively (**Figure 2C**). In order to investigate the physiologic role of melA in SM9913, the melA coding region was removed from the SM9913 genome using the fusion PCR method which we developed for genetic manipulation of Pseudoalteromonas strains (Wang et al., 2015). Deletion of melA gene was verified by PCR followed by DNA sequencing (Supplementary Figure S6A). As expected, no visible dark brown pigment was produced in the 1melA strain after prolonged incubation at 37◦C for up to 3 days (Supplementary Figure S6B). The level of pyomelanin in the culture was also quantified by measuring the absorbance of the supernatant at 400 nm, a method employed previously to measure the pyomelanin production in bacterial culture (Turick et al., 2002). As shown in **Figure 2D**, the 1melA strain produced a reduced level of pyomelanin in the culture than that of the wild-type strain when cultured at 37◦C, confirming that melA expression was critical for pyomelanin production at higher temperatures. In addition, we performed the complementation study by constructing the pBBR1MCS-Cm-melA plasmid to express the melA gene using its native promoter. As expected, the introduction of pBBR1MCS-Cm-melA into the 1melA strain restored the pigmentation at 37◦C, while the introduction of the empty plasmid pBBR1MCS-Cm could not do so (**Figure 2E**). Collectively, these results demonstrate that the accumulation of the melA product increased the pyomelanin production in SM9913 at higher temperatures.

## Pyomelanin Production Increases during Pellicle Formation

During the formation of pellicles in SM9913, an increased amount of brownish pigment at later-stage was observed (**Figure 1C**). We therefore hypothesized that pyomelanin production is increased during the formation of pellicles in SM9913. In order to test this hypothesis, supernatants were collected during the formation of pellicle and quantified by measuring the absorbance at 400 nm. The OD<sup>400</sup> values of the wild-type strain was increased during the culturing, while the 1melA strain produced a reduced level of pigment (**Figure 3A**), suggesting an increased production of pyomelanin in the culture of SM9913. We further performed qRT-PCR to analyze the melA transcription during pellicle formation. Notably, when cultured at the optimal temperature of 15◦C, the transcription of melA was increased 72 ± 8-fold at the initial stage of pellicle formation (day 1), and 36 ± 3-fold at the later stage of pellicle formation (day 5)

SM9913 at 37◦C compared to the condition at 15◦C quantified by qRT-PCR. Overnight cultures were diluted to OD600∼0.1 and re-grown until OD600∼1 at 15◦C. Cells were collected before and after transfer to 37◦C for 25 min. (D) Pigment production of SM9913 and 1melA strains cultured at 37◦C. (E) Pigment production of SM9913 and 1melA strains with exogenous SM9913 melA gene expression (1melA/pBBR1MCS-Cm-melA) or with an empty plasmid (1melA/pBBR1MCS-Cm) at 37◦C for 2 days measured at OD400. Data are the average of three independent cultures, and one standard deviation is shown in (C–E).

culture at 15◦C in SWLB. (B) Fold changes of the melA, hmgA, maiA, and fahA transcription in biofilm cells (day 1 or day 5) compared to planktonic cells (OD600∼ 1) in SWLB at 15◦C. Data are the average of three independent cultures, and one standard deviation is shown in (A,B).

compared to the cells growing planktonically (**Figure 3B**). Similar results were obtained during pellicle formation at 4◦C, with the transcription of melA significantly induced at day 5 at the onset of forming visible pellicle (Supplementary Figure S7). Therefore, pyomelanin production was increased during pellicle formation of SM9913.

of melA in E. coli K12 was induced with 0.5 mM IPTG. Data are the average of three independent cultures, and one standard deviation is shown in (A,B).

## Pyomelanin Formation Increases Heat Resistance

We next tested whether the formation of pyomelanin protects SM9913 from stress. When treated at 45◦C for 30 min, the cell viability of 1melA strain decreased 63 ± 1-fold compared to wild-type strain (**Figure 4A**). To examine whether ectopic expression of the melA product can provide protection for other bacterial cells during heat stress, the melA gene from SM9913 genome was cloned and expressed in E. coli K12 strain (**Table 1**). E. coli K12 carries a tyrosine aminotransferase which is required for the initial conversion of tyrosine to 4-hydroxyphenylpyruvate (Gelfand and Rudo, 1977), but it lacks the gene/genes responsible for conversion of 4-hydroxyphenylpyruvate to HGA which is required for the synthesis of pyomelanin. The ectopic expression of the melA product in E. coli K12 was checked by SDS-PAGE and a protein band with a similar size of 4-HPPD was produced as expected (Supplementary Figure S8A). When melA was induced by adding IPTG in E. coli K12, a dark brown pigment was observed in the culture supernatant (Supplementary Figure S8B). Moreover, when treated at 65◦C for 10 min, the cell viability of E. coli K-12 increased 119 ± 1-fold when melA was induced (**Figure 4B**). Thus, the synthesis of pyomelanin also plays a protective role against heat damage in E. coli.

## Genes that Involved in L-Tyrosine Catabolism Are Differentially Regulated

Pyomelanin was produced by the accumulation of HGA starting from the L-tyrosine degradation (Kotob et al., 1995; Serre et al., 1999). Genomic analysis suggests that all of the four genes that participate in the L-tyrosine catabolism (as defined in the KEGG database<sup>1</sup> ) are present in the genome of SM9913. For

<sup>1</sup>http://www.kegg.jp/

many bacteria with one chromosome, the genes that participate in the L-tyrosine catabolism were located in the chromosome as one single transcriptional unit, such as in Pseudomonas aeruginosa PAO1 and P. aeruginosa PA14 (Rodriguez-Rojas et al., 2009). However, in SM9913, the genes involved in the catabolism of L-tyrosine are located on two chromosomes, respectively. Gene fahA, located 197 bp upstream of melA on chromosome I, encodes fumarylacetoacetase which converts 4-fumarylacetoacetate into acetoacetate and fumarate. Two other genes, hmgA (PSM\_B0404) and maiA (PSM\_B0403), are located on chromosome II, with maiA lying 7 bp upstream of hmgA. Gene hmgA encodes for homogentisate 1, 2 dioxygenase, which converts HGA to 4-maleylacetoacetate. Gene maiA encodes maleylacetoacetate isomerase, which converts 4 maleylacetoacetate to 4-fumarylacetoacetate (Flydal et al., 2012) (**Figure 5A**). To test whether L-tyrosine was required to the formation of pyomelanin, L-tyrosine was added to the SWLB agar plates and pyomelanin production was investigated. As expected, the addition of L-tyrosine increased the pyomelanin production in the wild-type strain at 37◦C, while did not induce the formation of pyomelanin after prolonged incubation at 15◦C (**Figure 5B**). qRT-PCR results showed that the addition of L-tyrosine increased the transcripts of all of the four genes, and the transcription of melA was slightly lower than that of hmgA and maiA (**Figure 5C**). Importantly, fahA was the most highly induced gene among these four genes at 15◦C (**Figure 5C**). High expression of fahA product can convert more 4-fumarylacetoacetate into acetoacetate and fumarate. Acetoacetate and fumarate could therefore be utilized in the TCA cycle and contribute to the generation of ATP. However, under heat stress and during biofilm formation condition in which increased pyomelanin production was detected, the four genes were all induced and the level of melA transcription which is responsible for the synthesis of HGA was significantly higher induced than those of the hmgA and maiA which are responsible for the degradation of HGA (**Figures 2C**, **3B**).

Collectively, these results demonstrated that the formation of pyomelanin in SM9913 is mainly caused by an increased expression of melA gene which is responsible for the synthesis of HGA as comparing to the genes that are responsible for the degradation of HGA during heat stress and during biofilm formation.

## DISCUSSION

In this study, we found that SM9913 has the ability to form attached biofilm on the inorganic surface and to form pellicles growing at the air–liquid interface. Biofilm formation may benefit bacteria and other biofilm-forming organisms to survive in different marine habitats, including the deep-sea sediment (Hall-Stoodley et al., 2004). Some Pseudoalteromonas species are known to form biofilms (Mai-Prochnow et al., 2004; Huang et al., 2007; Iijima et al., 2009), and the pigmented Pseudoalteromonas spp. has been shown to exhibit antifouling and antibacterial capabilities (Egan et al., 2002; Bowman, 2007). In this study, we showed that the deep-sea non-pigmented bacteria Pseudoalteromonas sp. SM9913 produced pyomelanin pigment only at elevated temperatures or during biofilm formation. We also have demonstrated here that melA gene which is responsible for the synthesis of HGA was induced by a high temperature and during biofilm formation. The pyomelanin production was abolished when the melA gene was deleted. Additionally, deletion of melA in SM9913 reduced survival upon heat stress. Ectopic expression of the melA gene of SM9913 in E. coli K-12 which does not have an endogenous melA gene was also able to provide protection to E. coli cells during heat stress.

The physiological function of pyomelanin has been previously explored in the clinically important pathogens. For example, in Legionella pneumophila, the secreted pyomelanin conferred a ferric reductase activity which played an important role in iron uptake, thus enhancing the growth under iron-limiting conditions (Chatfield and Cianciotto, 2007; Zheng et al., 2013). In Pseudomonas aeruginosa and Vibrio cholerae, pyomelanin producing or hyperproducing variants were frequently isolated and showed increased persistence or virulence (Rodriguez-Rojas et al., 2009; Valeru et al., 2009). In Burkholderia cenocepacia, pyomelanin production helped scavenge free radicals, resulting in the attenuation of the host cell (Keith et al., 2007). Although SM9913 was isolated from the deepsea sediment where the temperature was near the freezingpoint, it retained the ability to survive at a wide range of temperatures. In this study, we found that the pyomelanin was induced during heat stress and it is possible that the process of auto-oxidizing HGA to pyomelanin might help to scavenge free radicals, which are often generated during heat stress, and to protect the cells from heat damage (Benov and Fridovich, 1995; Mols and Abee, 2011). The increased expression of melA gene is responsible for the synthesis of pyomelanin at elevated temperatures or during biofilm formation; however, we observed that the melA mutant strain still showed a small increase by the absorbance at 400 nm during heat stress and during biofilm formation (**Figures 2D**, **3A**). It is possible that the other type of pigments may be produced under heat stress and during pellicle formation through an unknown pathway, which showed absorbance at 400 nm in the culture supernatants. Whether the production of pyomelanin or other unknown pigments can be induced in other stress conditions needs further investigation. Meanwhile, the physiological function of pyomelanin in extreme marine environment also needs to be explored.

Although it has been suggested that the Pseudoalteromonas spp. can be categorized into two clades, we tend to believe that this classification needs refinement, especially for pyomelaninproducing strains. Most of the pigments, such as prodiginines (red) from Pseudoalteromonas rubra (Feher et al., 2008), cyclodigiosin hydrochloride (red) from Pseudoalteromonas denitrificans (Kim et al., 1999), violacein (purple) from Pseudoalteromonas luteoviolacea (Yang et al., 2007), and tambjamines-like alkaloid (yellow) from Pseudoalteromonas tunicata (Franks et al., 2005), are synthesized by specific enzymes that are only present in the genome of their respective strains. However, the L-tyrosine catabolic genes cluster, including melA, hmgA, fahA and maiA genes, that is responsible for the production of pyomelanin are found in all sequenced Pseudoalteromonas strains (data not shown). Thus, we reasoned that Pseudoalteromonas spp. are capable of producing pyomelanin when encounter certain stress condition, which help the bacteria to adapt to changing marine environment. Three Pseudoalteromonas species have been characterized as melaninlike pigments-producing strains, including Pseudoalteromonas nigrifaciens (Baumann et al., 1984), Pseudoalteromonas aliena (Ivanova et al., 2004), and Pseudoalteromonas distinct (Ivanova et al., 2004), but it remains to be determined whether the production of dark brown pigments is due to the accumulation of 4-HPPD or the defective HmgA activity who is responsible for the degradation of HGA. Indeed, strains that overproducing pyomelanin were commonly isolated from patients with cystic fibrosis and bronchiectasis, in which Pseudomonas aeruginosa living in biofilms, and the hyper-production of pyomelanin in these mutants were mostly due to the loss activity of HmgA (Rodriguez-Rojas et al., 2009; Hocquet et al., 2016). Naturally-occurring hyper-pyomelanin producing mutants of Vibrio cholerae due to the mutation of hmgA also have been isolated from different environmental waters or patients (Wang et al., 2011). Moreover, we have recently isolated a pyomelaninproducing variant from the biofilm cells of a non-pigmented strain Pseudoalteromonas lipolytica that are caused by the point mutation in hmgA gene (data not shown here). Thus, a detailed genetic analysis of L-tyrosine catabolic pathway genes in Pseudoalteromonas strains is needed to elucidate the exact mechanism underlying the production of melaninlike compound. Further effort is needed to explore the exact mechanism controlling the L-tyrosine catabolism under different stress conditions in different organisms.

## AUTHOR CONTRIBUTIONS

XW conceived the idea and designed the project. ZZ, XC, PW, YG, XL, and BL carried out the experiments. XW, ZZ, and XC analyzed data. XW, ZZ, and XC wrote the manuscript with input from all other authors.

## FUNDING

The work was supported by National Science Foundation of China (31500093, 31290233, 31270214, and 41606179), National Basic Research Program of China (2013CB955701), and National Science Foundation of Guangdong Province (2014A030310385).

## ACKNOWLEDGMENTS

We thank Research Centre for Life Sciences, University of Science and Technology of China (Hefei, Anhui, China) for Mass Spectrum analysis. XW is awarded the 1000-Youth Elite Program (The Recruitment Program of Global Experts in China).

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.01822/full#supplementary-material

## REFERENCES

fmicb-08-01822 September 21, 2017 Time: 10:30 # 10


**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 Zeng, Cai, Wang, Guo, Liu, Li and Wang. 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.

fmicb-08-01822 September 21, 2017 Time: 10:30 # 11