sign in sign up
<<
Home , Acidimicrobiaceae, Bacterioplankton, Creatininase, Cyanobacteria, Duganella, Hydrogenophilaceae

INTERACTIONS OF AND CO-OCCURRING

MICROORGANISMS DURING CYANOBACTERIAL HARMFUL ALGAL

BLOOMS

A dissertation submitted

to Kent State University in partial

fulfillment of the requirements for the

degree of Doctor of Philosophy

by

Kai Wang

May 2021

© Copyright

All rights reserved

Except for previously published materials Dissertation written by

Kai Wang

B.S., Sichuan Agricultural University, 2012

M.S., University of China, 2015

Ph.D., Kent State University, 2021

Approved by

Xiaozhen Mou, Ph.D. , Chair, Doctoral Dissertation Committee

Helen Piontkivska, Ph.D. , Members, Doctoral Dissertation Committee

David Costello, Ph.D.

Hanbin Mao, Ph.D.

Joseph Ortiz, Ph.D.

Accepted by

Laura G. Leff, Ph.D. , Chair, Department of Biological Sciences

Mandy Munro-Stasiuk, Ph.D., Interim Dean, College of Arts and Sciences TABLE OF CONTENTS

TABLE OF CONTENTS ...... iii

LIST OF FIGURES ...... v

LIST OF TABLES ...... ix

ACKNOWLEDGEMENTS ...... x

I. GENERAL INTRODUCTION ...... 1

REFERENCES ...... 14

II. CYANOBACTERIAL BLOOMS ALTER THE RELATIVE IMPORTANCE OF

NEUTRAL AND SELECTIVE PROCESSES IN ASSEMBLING FRESHWATER

BACTERIOPLANKTON COMMUNITY ...... 25

PREFACE ...... 25

ABSTRACT ...... 25

INTRODUCTION ...... 26

MATERIALS AND METHODS ...... 29

RESULTS ...... 36

DISCUSSION ...... 44

CONCLUSIONS ...... 49

REFERENCES ...... 51

III. CO-OCCURRING REGULATE THE SUCCESSION OF

CYANOBACTERIAL HARMFUL ALGAL BLOOMS ...... 83

PREFACE ...... 83

iii ABSTRACT ...... 83

INTRODUCTION ...... 84

MATERIALS AND METHODS ...... 87

RESULTS ...... 93

DISCUSSION ...... 102

CONCLUSIONS ...... 106

REFERENCES ...... 108

IV. COORDINATED DIEL EXPRESSION OF CYANOBACTERIA AND

THEIR ...... 140

PREFACE ...... 140

ABSTRACT ...... 140

INTRODUCTION ...... 141

MATERIALS AND METHODS ...... 143

RESULTS ...... 149

DISCUSSION ...... 156

CONCLUSIONS ...... 160

REFERENCES ...... 161

V. SUMMARY ...... 177

REFERENCES ...... 183

iv LIST OF FIGURES

Figure 1 Average abundances of estimated microbial communities...... 60

Figure 2 Time-lag regression analysis of microbial communities...... 61

Figure 3 Fit of the neutral model...... 62

Figure 4 (a) The estimated migration rate, (b) Neutral and non-neutral partitions of the communities, (c) List of non-neutral bacterioplankton genera...... 63

Figure 5 Relationship between the measured environmental parameters and bacterioplankton communities based on (a) all samples and (b) only non-bloom samples, (c) Heatmap of

Pearson’s correlation coefficient values...... 64

Figure 6 (a) The co-occurrence patterns among bacterioplankton communities, cyanobacterial, planktonic , and abundances, (b-g) Ternary plots showing relative abundance from modules 1-6 in the three different periods...... 65

Figure 7 Rarefaction curves of operational taxonomic unit at 97% sequence similarity level.

...... 66

Figure 8 Temporal distribution pattern of bacterioplankton alpha diversity indexes...... 67

Figure 9 Time-lag regression analysis of bacterioplankton communities at (a) , (b) class, (c) order, (d) family, and (e) genus levels...... 68

Figure 10 The relative contribution of measured environmental parameters on (a) cyanobacterial, (b) planktonic algae, and (c) zooplankton communities...... 69

Figure 11 Distance-based linear redundancy visualizing the relative contribution of measured environmental parameters on (a) non-neutral OTUs, and (b) neutral OTUs...... 70

v Figure 12 The distributions of degree for the microbial community co-occurrence network

(green) and Erdös-Rényi random network (orange)...... 71

Figure 13 Zi-Pi plot showing the distribution of bacterioplankton OTUs based on their topological roles...... 72

Figure 14 Network visualizing significant correlations between measured biotic and abiotic variables...... 73

Figure 15 Temporal dynamics of (a) cyanobacterial genera and (b) their associated microbiome , (c) Correlation analysis of cyanobacterial abundances, (d) PCA analysis of cyanobacterial metatranscriptomes...... 116

Figure 16 Environmental variables during sampling period in Harsha Lake...... 117

Figure 17 Differentially expressed cyanobacterial in of (a) , (b)

Phosphorus, (c) , (d) Extracellular , and (e) Secondary metabolites.

...... 118

Figure 18 WGCNA networks of (a) , (b) , (c) , (d) non- cyanobacterial, and (e) gene expression matrices...... 119

Figure 19 (a) The WGCNA approach directly links (a) Ana1 and (c) NC8 module structures to Anabaena ; sPLS regression was used to predict Anabaena biomass...... 120

Figure 20 Temporal change of averaged expression levels of genes of the enriched GO terms within the (a) Ana1, and (c) NC8 modules. The value importance in projection values...... 121

Figure 21 (a) The WGCNA approach directly links (a) Mic4, (c) Mic6, (e) Mic7, (g) NC2, and (i) CP9 module structures to Microcystis biomass; sPLS regression was used to predict

vi Microcystis biomass...... 122

Figure 22 Temporal change of averaged expression levels of genes of the enriched GO terms within the (a) Mic4, Mic6, Mic7, (c) NC2 modules, and (e) CP9 module. The value importance in projection values...... 123

Figure 23 The WGCNA approach directly links (a) Pla12, (c) NC5, and (e) CP8 module structures to Planktothrix biomass; sPLS regression was used to predict Planktothrix biomass...... 124

Figure 24 Temporal change of averaged expression levels of genes of the enriched GO terms within the (a) Pla12, (c) NC5 modules, and CP8 module. The value importance in projection values...... 125

Figure 25 A schematic diagram showing potential interactions between cyanobacteria and their microbiome...... 126

Figure 26 Measured physicochemical variables of Lake Erie and microcosm samples...... 169

Figure 27 The expression pattern of Microcystis genes between the Microcystis (MCY) and whole water (WW) samples...... 170

Figure 28 Averaged dDEGs related to (a) N and P ; (b) Fe metabolism; (c) vitamin

biosynthesis; (d) stress response and H2O2 depletion; and (e) synthesis...... 171

Figure 29 The expression pattern of microbiome genes between the microbiome (MIB) and whole water (WW) samples...... 172

Figure 30 Averaged dDEGs of the microbiome related to (a) N metabolism; (b) Organic nitrogen metabolism; (c) P metabolism; (d) Iron metabolism and methionine synthesis; and

vii (e) Stress response and H2O2 depletion...... 173

Figure 31 Microbiome communities identified in the MIBs and WWs...... 174

Figure 32 Correlations between Microcystis genes and microbiome genes...... 175

Figure 33 A schematic diagram showing potential interactions between Microcystis and its microbiome communities...... 176

viii LIST OF TABLES

Table 1 List of indicator found before, during and after the bloom periods...... 74

Table 2 Topological properties of the entire co-occurrence networks of microbial communities and their associated random networks...... 79

Table 3 Lists of module hubs and connectors in co-occurrence network...... 80

Table 4 Lists of keystone species in co-occurrence network...... 82

Table 5 The summary of genes within ABR-Ana1 module...... 127

Table 6 The summary of genes within ABR-NC8 module...... 129

Table 7 The summary of genes within MBR-Mic modules...... 130

Table 8 The summary of genes within MBR-NC2 module...... 132

Table 9 The summary of genes within MBR-CP9 module...... 134

Table 10 The summary of genes within PBR-Pla12 module...... 135

Table 11 The summary of genes within PBR-NC5 module...... 137

Table 12 The summary of genes within PBR-CP8 module...... 139

ix ACKNOWLEDGEMENTS

First and foremost, I would like to express my deep and sincere gratitude to my advisor,

Dr. Xiaozhen Mou, for giving me the opportunity to do research and providing individual guidance throughout this research. I am extremely grateful for what she has offered me.

I would also like to thank my committee members, Drs. Helen Piontkivska, David

Costello, Hanbin Mao, and Joseph Ortiz, for serving in my advisory committee and giving me great support and valuable advice during my study. Besides, I greatly appreciate my collaborators, Dr. Jingrang Lu at the United States Environmental Protection Agency and

Mandy Razzano at the Ohio Environmental Protection Agency, for their support and advice to lead my step forward. I am also grateful for the funding support from the National Science

Foundation Grants (CBET 1605161to X.M), Safe and Sustainable Water Resources (SSWR:

4.01D, 4.3.1 and 4.3.3 to J.L), and Kent State University.

I appreciate my lab members, friends, and some undergraduates for their kind help and support towards the completion of my research. Special thanks to Emily Kitchen and Youchul

Jeon. I show my sincere thanks to all the faculties and staff at the Department of Biological

Sciences for giving wonderful courses and helping me out through the study in class and research. I express my special thanks to Antonio D'angona and Joshua Talbott for their support in the server setup.

Lastly, I would like to thank all of my family members for their endless love, caring, and sacrifices for educating and preparing me for my future.

x CHAPTER I

GENERAL INTRODUCTION

Cyanobacteria are the earth’s oldest (∼3.5 billion ago) -producing that can obtain their energy through (Paerl and Otten, 2013). It is believed that the photosynthetic activities of cyanobacteria converted the early reducing into an oxidizing state, which dramatically changed the composition and diversity of forms on Earth (Schopf, 2000). Cyanobacteria can grow rapidly under suitable

conditions (i.e., high nutrients, warm water temperature, and elevated CO2 levels) and form

dense blooms (Huisman et al., 2018). Temperature and CO2 concentration are also expected to rise in the future, which will result in enhanced water stratification and alterations in weather patterns (i.e., seasonal or interannual) (Kosten et al., 2012). All these conditions are thought to increase the , intensity, and duration of cyanobacterial blooms (Huisman et al., 2018).

Cyanobacteria live with a diversity of other microbial communities in water environments

(Hmelo et al., 2012). However, little is known about the assembly mechanism of microbial communities under the pressure of blooms and the reciprocal impacts between cyanobacteria and the co-occurring microorganisms. This dissertation aimed to study the interactions between cyanobacteria and their co-occurring microorganisms and examine how these interactions regulate community assembly and resource cycling (i.e., C, N, P, vitamins, and iron).

1 The traits of bloom-forming cyanobacteria

Many cyanobacteria can form dense blooms, and these bloom-forming cyanobacteria

may carry different physiological traits, such as N2 fixation (Zehr, 2011), gas vesicle synthesis

(Pfeifer, 2012), CO2-concentrating mechanism (Price et al., 2008), and toxic secondary metabolites production (i.e., ) (Sivonen and Jones, 1999). Some of these traits provide cyanobacteria with a distinct competitive advantage over other and eukaryotic algae, which tends to favor their dominance and enables the development of cyanobacterial harmful algal blooms (CyanoHABs) (Huisman et al., 2018).

Some bloom-forming cyanobacteria, i.e., genera Anabaena, ,

Cylindrospermopsis, , and , can fix atmospheric N2 that is not directly available to many other phytoplankton (Schindler et al., 2008). The N-fixing process gives cyanobacteria a competitive advantage in N-limited water environments, where they can develop blooms if other nutrient elements (i.e., phosphorus (P) and iron) are ample. It is

generally assumed that N-limited conditions favor the growth of N2-fixing cyanobacteria,

whereas, when N supply is plentiful, non-N2-fixing cyanobacteria would outcompete N2-fixing taxa (Paerl and Otten, 2016). However, many freshwater CyanoHABs showed an opposite

pattern, with N2-fixers dominate when N was replete and then transited into non-N2-fixing

genera under low N concentrations (Beversdorf et al., 2013; Lu et al., 2019).

Buoyancy, which allows cyanobacterial cells to move up and down in the water column, is another adaptive advantage of cyanobacteria (Walsby, 1994; Pfeifer, 2012). The movement speed of buoyant cyanobacteria increases with size (Visser et al., 1997), with larger

2 colonies of cyanobacteria move up and down more rapidly than smaller ones (Visser et al.,

1997). Diel vertical movement enables cyanobacteria to have better access to light at the surface during the day and absorb nutrients in deeper waters at night (Villareal et al., 2003).

Cyanobacteria derive their buoyancy ability from producing gas vesicles that are hollow protein structures filled with gases (Walsby, 1994; Pfeifer, 2012). Buoyant cyanobacteria can

form dense surface blooms and intercept the influx of light and atmospheric CO2 (Huisman et al., 2004).

Cyanobacteria use CO2 as the carbon source for photosynthesis and growth. During the

bloom period, dense cyanobacterial blooms can rapidly deplete water CO2 in the surface layer and raise water pH, and thus shifting the equilibrium of inorganic carbon in the water towards

- 2- bicarbonate (HCO3 ) and carbonate (CO3 ) (Price et al., 2008; Kosten et al., 2012). Over their evolutionary history, cyanobacteria have experienced a changing gaseous environment from

CO2 rich to O2 rich. This has imposed evolutionary pressure on cyanobacteria to evolve adaptive strategies for efficiently obtaining inorganic carbon for photosynthesis (Price et al.,

2008). Specifically, cyanobacteria have developed CO2-concentrating mechanisms (CCMs)

that can concentrate CO2 up to 1000-fold around the of the carbon-fixing

(Rubisco), which enables Rubisco to utilize CO2 more efficiently (Price et al., 2008;

Burnap et al., 2015). CCMs include five different inorganic carbon uptake systems (two for

- CO2 and three for HCO3 ), which have different inorganic carbon affinities (Price et al., 2008;

Burnap et al., 2015). The combination of these five uptake systems provides cyanobacteria flexibility in utilizing the carbon in response to inorganic carbon availability in the water

3 environment (Sandrini et al., 2014; Sandrini et al., 2016).

Some bloom-forming cyanobacteria can also produce a variety of secondary metabolites (i.e., cyanotoxins) that are toxic to both invertebrates and vertebrates via a variety of mechanisms (Carmichael, 2001; Metcalf and Good, 2002; Merel et al., 2013). Cyanotoxins are widely produced by a number of bloom-forming cyanobacterial taxa, including Anabaena,

Aphanizomenon, , Dolichospermum, Microcystis, Nodularia, , and Planktonthrix (Rantala et al., 2004; Pearson et al., 2010). Cyanotoxins present chronic and acute hazards to humans, such as affecting the nervous system (), the liver

(hepatotoxins), or the skin (dermatoxins) (Sivonen and Jones, 1999). The most frequently detected and abundant cyanotoxins during CyanoHABs are microcystins (MCs), which have a wide range of toxic effects (Huisman et al., 2018). The ecological significance of producing cyanotoxins has not been fully resolved. Some studies showed that cyanotoxins might function as deterrents against grazing (Lemaire et al., 2012; Jiang et al., 2016), however, there is also evidence showed that cyanotoxins (i.e., MCs) might be evolved before the origin of their predators (i.e., and Cladocerans) (Rantala et al., 2004). A recent study has suggested that MCs may play roles in protecting the carbon-fixing enzymes and other cyanobacterial proteins from (Zilliges et al., 2011).

Environmental factors that affect cyanobacterial blooms

Environmental changes have been shown to play important roles in promoting

CyanoHABs worldwide (Paerl and Otten, 2013). For example, anthropogenic activities (i.e., agricultural, urban, and industrial) has significantly increased the inputs of N and P

4 () into the rivers and lakes, which has promoted CyanoHABs expansion and persistence since the 1960s (Huisman et al., 2005; Huisman et al., 2018). Since then, measures to reduce nutrient loading have been implemented to improve water quality and prevent or mitigate CyanoHABs (Sellner et al., 1993; Sevilla et al., 2008). P has traditionally been considered the major nutrient that limits the cyanobacterial biomass accumulation in freshwater (Schindler et al., 1975). P enrichment may favor the formation of CyanoHABs,

especially for N2-fixing cyanobacterial genera that can fulfill their own N needs by the N-fixing process (Gallon, 1992; Downing et al., 2001). However, in recent decades, N inputs have gradually increased due to the application of N-containing fertilizers, human and agricultural wastes, stormwater runoff, groundwater discharge, and atmospheric deposition (Shelford et al.,

2012). Consequently, the N/P ratio is currently rising in aquatic ecosystems (Shapiro, 1990;

Shelford et al., 2012), which was suggested to increase the proliferation of non-N2-fixing cyanobacteria, such as Microcystis, Planktothrix, and Oscillatoria (Paerl and Otten, 2013).

Moreover, increased N loading may favor the production of the N-rich cyanotoxins (i.e., microcystins) (Van de Waal et al., 2009). N and P co-limitation was frequently observed in freshwater environments (Paerl et al., 2011; Paerl et al., 2014). For example, combined enrichment with N and P led to higher total biomass than either N or P alone (Paerl et al., 2011;

Paerl et al., 2014). Iron also showed significant roles in regulating bloom formation, especially when N and P are not limited (Xu et al., 2013). In addition, iron is an important co-factor for , which may play significant roles in affecting the dominance of N-fixing species

(Larson et al., 2018).

5 In addition to nutrients, high surface water temperatures promote the proliferation of cyanobacteria in several ways (Joehnk et al., 2008; Paerl and Huisman, 2008; Kosten et al.,

2012). Many bloom-forming cyanobacteria have higher growth rates at warm temperatures

(Nalewajko and Murphy, 2001; Thomas and Litchman, 2016). For example, Microcystis showed a higher growth rate (1.6 divisions day-1) over Anabaena (1.25 divisions day-1) at higher water temperature (28-32 °C) (Nalewajko and Murphy, 2001). In addition, cyanobacteria seem to grow faster in warm temperatures than eukaryotic phytoplankton (Visser et al., 2016). Warming surface water also leads to a more stable water stratification (less vertical mixing) (Johnk et al., 2008; Paerl and Huisman, 2008; Kosten et al., 2012), which provides ideal conditions for buoyant cyanobacteria to float upwards to use light and shade other cyanobacteria and eukaryotic algae (Walsby et al., 1997; Huisman et al., 2004). Furthermore, temperature and nutrients often have synergistic effects on cyanobacterial growth (Kosten et al., 2012; Rigosi et al., 2014), which implies that nutrient loading may need to be reduced even more to control CyanoHABs in a future warmer climate (Kosten et al., 2012).

Mathematical models and laboratory experiments showed that rising atmospheric CO2 levels are likely to increase the frequency, intensity, and duration of CyanoHABs in the future

(Verspagen et al., 2014; Ji et al., 2017). Rising CO2 levels can increase the CO2 influx into the surface layer that can be used by surface-dwelling cyanobacteria and other phytoplankton

(Visser et al., 2016). In addition, cyanobacteria are thought to have the better ability over other

phytoplankton to rapidly adapt to the rising CO2 concentrations due to the evolved CCM mechanisms (Hutchins et al., 2013; Sandrini et al., 2014).

6 The interactions between cyanobacteria and co-occurring microorganisms

In aquatic systems, CyanoHABs can cause significant changes in bacterial communities.

Some are attached to cyanobacterial cells (Ploug et al., 2011; Hmelo et al., 2012), whereas others develop free-living populations and inhabit the area surrounding cyanobacterial cells (Brauer et al., 2015). Cyanobacteria can interact with these free-living bacteria but also maintain attached bacteria on their cellular surface.

Cyanobacteria-bacteria interactions are often highly complex including both cooperative and competitive relationships (Amin et al., 2012). Their relationships are mainly based on resource exchange (Seymour et al., 2017). Aquatic heterotrophic bacteria benefit from associations with cyanobacteria that produce dissolved organic compounds, including typically highly labile dissolved organic carbon (Larsson and Hagström, 1979), and more complex products (i.e., ) (Teeling et al., 2012). Cyanobacteria can either benefit (Croft et al., 2005; Amin et al., 2009) or suffer (Gumbo et al. 2008) from the presence of heterotrophic bacteria. For example, cyanobacteria can obtain nutrients (i.e., N, P, iron) via heterotrophic bacteria remineralization (Legendre and Rassoulzadegan, 1995; Amin et al., 2009) but they may also suffer nutrient competition with heterotrophic bacteria or may even be killed via lysis

(Gumbo et al. 2008).

Cyanobacteria-associated bacterial communities are often restricted to only a handful of phyla, including , , (Alpha-, Beta-, and

Gamma-), , and (Berg et al., 2009; Woodhouse et al., 2016;

Liu et al., 2019; Wang et al., 2020). However, at lower levels (i.e., genus and species),

7 cyanobacteria-associated microbial communities varied over temporal scales with dynamic shifts in cyanobacterial species and bloom phases (Allgaier and Grossart, 2006; Engström-Öst et al. 2013). The results showed that the variation of cyanobacteria-associated microbial communities has been linked to changes in the dissolved organic matter released during the different phases of CyanoHABs (Engström-Öst et al. 2013). Such microbial community shifts may in turn have feedback effects on cyanobacterial growth, bloom development, and prolongation, and stimulate changes in the quality and quantity of compounds that will be released by the cyanobacteria (Sarmento and Gasol 2012).

Cyanobacteria also a diversity of (i.e., ), which can cause high cyanobacterial mortality; these “cyanobacteria grazers” are often host-specific (Gerphagnon et al., 2015) and can shift cyanobacterial community structures by selectively removing certain cyanobacterial taxa. Cyanophages may also play a different role in helping the dispersion of filamentous cyanobacteria. For example, the lysed filamentous cyanobacteria are split into smaller but viable filament fragments, which may help disperse the cyanobacteria in the aquatic environments (Pollard and Young 2010). Cyanophages constitute a major evolutionary driving force in the host , promoting (Kaplan 2016). On the other hand, cyanobacteria hosts have evolved defensive mechanisms against cyanophage , including the antivirus genes (Makarova et al., 2011) and CRISPR-Cas systems (Kuno et al.,

2014).

Introduction to the major techniques used in this study

16S rRNA gene sequencing: The 16S ribosomal RNA (rRNA) gene is highly

8 conserved between different bacterial species and can be used as a reliable molecular clock

(Weisburg et al., 1991). In addition to highly conserved regions, the 16S rRNA gene contains hypervariable regions (V1-V9) that can provide species-specific resolution for the identification of bacteria (Gray et al., 1984). The next-generation sequencing (NGS) technique has revolutionized genomic research and has become increasingly rapid, sensitive, and cost- effective. 16S rRNA amplicon sequencing has been used extensively in the identification and classification of bacteria and (Patel et al., 2001). 16S rRNA gene sequencing provides in-depth information about bacterial communities without culturing. For 16S rRNA gene sequencing, DNA is first extracted from membrane filters used to collect bacteria in the water, specific variable regions (i.e., V3-V4) of 16S rRNA gene are amplified, sequenced, and then identification of generated reads is based on similarity search against public reference databases of 16S rRNA gene sequences (i.e., SILVA database). A 16S rRNA gene sequencing was performed in the present study to examine the bacterial community compositions following the progress of CyanoHABs.

Metatranscriptomic sequencing: Metatranscriptomes refer to the total content of gene transcripts in a natural community (i.e., water), considered as a unique entity, at a given point of sampling (Bashiardes et al., 2016). Unlike 16S amplicon and metagenomic sequencing that focus on identifying community composition and their potential function based on DNA sequences, metatranscriptomic sequencing of community RNAs is used to study the active function profile within a community and to explore their gene expression patterns over time or at different conditions (Bashiardes et al., 2016). In addition, the metatranscriptomic method

9 can be used to examine differences in the active functions of microbial communities which may be identical at the composition level (Bashiardes et al., 2016). This improves our understanding of the structure, active function, and adaptive mechanisms of complex microbial communities. Furthermore, unlike quantitative reverse transcription PCR (RT-qPCR) to quantify the expression of known genes of interest, metatranscriptomics sequencing method also discovers previously unknown transcripts directly from the sequence data (Shakya et al.,

2019). In this dissertation research, metatranscriptomic sequencing was performed to study the gene expression of cyanobacteria and the co-occurring microorganisms over a seasonal and day-night period to examine their potential interactions.

Research objectives and dissertation outlines

The general objective of this dissertation was to examine interactions between cyanobacteria and their co-occurring microorganisms. These interactions might help us understand the significance of these interactions in regulating community assembly and cyanobacterial species succession. Specifically, I aimed to understand how CyanoHABs affect the assembly of the co-occurring microorganisms, how the co-occurring microorganisms influence the development and succession of CyanoHABs, and the metabolic activities of cyanobacteria and the co-occurring microorganisms during diel cycles. Following Chapter I

(this chapter), I reported my research findings in three chapters and provided a summary of the overall findings in Chapter V. A brief summary of each chapter of my dissertation is provided below.

Chapter I: General introduction

10 In this chapter, I gave a brief introduction of our current knowledge on cyanobacteria, the factors affecting CyanoHABs, and the interactions of cyanobacteria and the co-occurring microorganisms. I also described the hypothesis of each chapter here.

Chapter II: Cyanobacterial blooms alter the relative importance of neutral and selective processes in assembling freshwater bacterioplankton community

As common biological disturbances, CyanoHABs represent a threat to the water quality and function (Paerl and Fulton 2006). The strength of ecological selection has been found to vary with different cyanobacterial succession states (e.g., bloom vs. non-bloom)

(Berry et al., 2017). In this chapter, I hypothesized that neutral processes may be less important during the bloom period due to the highly selective pressure caused by cyanobacterial bloom.

Our results suggest that neutral processes play significant roles in assembling bacterioplankton communities over a cyanobacterial bloom succession and its relative importance was weakened by biotic pressures (interspecific interactions) during the bloom period. Our results also indicate that among biotic factors, cyanobacteria had greater impacts on bacterioplankton community assembly than planktonic algae and zooplankton.

Chapter III: Co-occurring microorganisms regulate the succession of cyanobacterial harmful algal blooms

It is generally assumed that N-limited conditions would favor the growth of N2-fixing

cyanobacteria, whereas, when N supply is ample, non-N2-fixing cyanobacteria would

outcompete slow-growing N2-fixing taxa (Paerl and Otten, 2016). However, many blooms have

been reported to start with N2-fixers when N was replete and then transited into non-N2-fixing

11 genera under low N concentrations (Beversdorf et al., 2013; Lu et al., 2019). In this chapter, I hypothesized that the resource exchange between cyanobacteria and their associated microbiome may favor the growth of certain cyanobacterial taxa (measured by biomass), which in turn regulates cyanobacterial bloom development and succession. Our results suggest that besides environmental conditions and inherent traits of specific cyanobacterial species, the development and succession of CyanoHABs are regulated by co-occurring microorganisms.

Specifically, the co-occurring microorganisms can alleviate the nutrient limitation of cyanobacteria by remineralizing organic compounds.

Chapter IV: Coordinated diel gene expression of cyanobacteria and their microbiome

Circadian rhythms in cyanobacterial metabolisms have been well recognized. However, whether this programmed activity of cyanobacteria could elicit a coordinated diel gene expression in their co-occurring microorganisms (microbiome) and how the response of the microbiome, in turn, impact cyanobacterial metabolism is unknown. In this chapter, I hypothesized that cyanobacteria can elicit coordinated diel expression in their microbiome and microbiome organisms may, in turn, impact the diel metabolism of cyanobacteria, due to tight interactions between cyanobacteria and microbiome organisms. Our results suggest that the

diel fluxes of OC and vitamin B12 in Microcystis largely impact the diel expression of microbiome genes. Meanwhile, the microbiome communities may support the growth of

Microcystis by supplying them with recycled nutrients but compete with Microcystis for iron.

Chapter V: Summary

In this chapter, I synthesized the overall findings of my studies and discussed the results

12 of my dissertation in a broader context.

13 REFERENCES

Allgaier, M., & Grossart, H. P. (2006). Seasonal dynamics and phylogenetic diversity of free-

living and particle-associated bacterial communities in four lakes in northeastern Germany.

Aquatic , 45(2), 115-128.

Amin, S. A., Green, D. H., Hart, M. C., Küpper, F. C., Sunda, W. G., & Carrano, C. J. (2009).

Photolysis of iron- chelates promotes bacterial-algal mutualism. Proceedings

of the National Academy of Sciences, 106(40), 17071-17076.

Amin, S. A., Parker, M. S., & Armbrust, E. V. (2012). Interactions between and

bacteria. and Molecular Biology Reviews, 76(3), 667-684.

Bashiardes, S., Zilberman-Schapira, G., & Elinav, E. (2016). Use of metatranscriptomics in

microbiome research. Bioinformatics and biology insights, 10, BBI-S34610.

Berg, K. A., Lyra, C., Sivonen, K., Paulin, L., Suomalainen, S., Tuomi, P., & Rapala, J. (2009).

High diversity of cultivable heterotrophic bacteria in association with cyanobacterial water

blooms. The ISME journal, 3(3), 314-325.

Beversdorf, L. J., Miller, T. R., & McMahon, K. D. (2013). The role of in

cyanobacterial bloom in a temperate, eutrophic lake. PloS one, 8(2), e56103.

Brauer, V. S., Stomp, M., Bouvier, T., Fouilland, E., Leboulanger, C., Confurius-Guns, V., ... &

Huisman, J. (2015). Competition and facilitation between the marine nitrogen-fixing

cyanobacterium and its associated bacterial community. Frontiers in

microbiology, 5, 795.

Burnap, R. L., Hagemann, M., & Kaplan, A. (2015). Regulation of CO2 concentrating

14 mechanism in cyanobacteria. Life, 5(1), 348-371.

Carmichael, W. W. (2001). Health effects of -producing cyanobacteria: “The

CyanoHABs”. Human and ecological risk assessment: An International Journal, 7(5),

1393-1407.

Croft, M. T., Lawrence, A. D., Raux-Deery, E., Warren, M. J., & Smith, A. G. (2005). Algae

acquire vitamin B12 through a symbiotic relationship with bacteria. Nature, 438(7064), 90-

93.

Downing, J. A., Watson, S. B., & McCauley, E. (2001). Predicting cyanobacteria dominance in

lakes. Canadian journal of fisheries and aquatic sciences, 58(10), 1905-1908.

Engström-Öst, J., Autio, R., Setälä, O., Sopanen, S., & Suikkanen, S. (2013).

community dynamics during decay of a cyanobacteria bloom: a mesocosm experiment.

Hydrobiologia, 701(1), 25-35.

Gallon, J. R. (1992). Reconciling the incompatible: N2 fixation and O2. Tansley Review No.

44. New Phytol, 122, 571-609.

Gerphagnon, M., Macarthur, D. J., Latour, D., Gachon, C. M., Van Ogtrop, F., Gleason, F. H.,

& Sime‐Ngando, T. (2015). Microbial players involved in the decline of filamentous and

colonial cyanobacterial blooms with a focus on fungal . Environmental

Microbiology, 17(8), 2573-2587.

Gray, M. W., Sankoff, D., & Cedergren, R. J. (1984). On the evolutionary descent of organisms

and : a global phylogeny based on a highly conserved structural core in small

subunit ribosomal RNA. Nucleic Acids Research, 12(14), 5837-5852.

15 Grossart, H. P., Levold, F., Allgaier, M., Simon, M., & Brinkhoff, T. (2005). Marine

species harbour distinct bacterial communities. Environmental Microbiology, 7(6), 860-

873.

Gumbo, R. J., Ross, G., & Cloete, E. T. (2008). Biological control of Microcystis dominated

harmful algal blooms. African Journal of Biotechnology, 7(25).

Hmelo, L. R., Van Mooy, B. A. S., & Mincer, T. J. (2012). Characterization of bacterial

epibionts on the cyanobacterium . Aquatic microbial ecology, 67(1), 1-14.

Huisman, J., Sharples, J., Stroom, J. M., Visser, P. M., Kardinaal, W. E. A., Verspagen, J. M.,

& Sommeijer, B. (2004). Changes in turbulent mixing shift competition for light between

phytoplankton species. Ecology, 85(11), 2960-2970.

Huisman, J., Codd, G. A., Paerl, H. W., Ibelings, B. W., Verspagen, J. M., & Visser, P. M.

(2018). Cyanobacterial blooms. Nature Reviews Microbiology, 16(8), 471-483.

Hutchins, D. A., Fu, F. X., Webb, E. A., Walworth, N., & Tagliabue, A. (2013). Taxon-specific

response of marine nitrogen fixers to elevated concentrations. Nature

Geoscience, 6(9), 790-795.

Ji, X., Verspagen, J. M., Stomp, M., & Huisman, J. (2017). Competition between cyanobacteria

and at low versus elevated CO2: who will win, and why?. Journal of

experimental , 68(14), 3815-3828.

Jiang, X., Gao, H., Zhang, L., Liang, H., & Zhu, X. (2016). Rapid of tolerance to

toxic Microcystis in two cladoceran grazers. Scientific reports, 6(1), 1-7.

Joehnk, K. D., Huisman, J. E. F., Sharples, J., Sommeijer, B. E. N., Visser, P. M., & Stroom, J.

16 M. (2008). Summer heatwaves promote blooms of harmful cyanobacteria. Global change

biology, 14(3), 495-512.

Kaplan, A. (2016). Cyanophages: starving the host to recruit resources. Current Biology,

26(12), R511-R513.

Kosten, S., Huszar, V. L., Bécares, E., Costa, L. S., van Donk, E., Hansson, L. A., ... & Scheffer,

M. (2012). Warmer climates boost cyanobacterial dominance in shallow lakes. Global

Change Biology, 18(1), 118-126.

Kuno, S., Sako, Y., & Yoshida, T. (2014). Diversification of CRISPR within coexisting

genotypes in a natural population of the bloom-forming cyanobacterium Microcystis

aeruginosa. Microbiology, 160(5), 903-916.

Larsson, U., & Hagström, A. (1979). Phytoplankton exudate release as an energy source for

the growth of pelagic bacteria. , 52(3), 199-206.

Larson, C. A., Mirza, B., Rodrigues, J. L. M., & Passy, S. I. (2018). Iron limitation effects on

nitrogen-fixing organisms with possible implications for cyanobacterial blooms. FEMS

microbiology ecology, 94(5), fiy046.

Legendre, L., & Rassoulzadegan, F. (1995). Plankton and nutrient dynamics in marine waters.

Ophelia, 41(1), 153-172.

Lemaire, V., Brusciotti, S., van Gremberghe, I., Vyverman, W., Vanoverbeke, J., & De Meester,

L. (2012). Genotype×genotype interactions between the toxic cyanobacterium Microcystis

and its grazer, the waterflea Daphnia. Evolutionary applications, 5(2), 168-182.

Liu, M., Liu, L., Chen, H., Yu, Z., Yang, J. R., Xue, Y., ... & Yang, J. (2019). Community

17 dynamics of free-living and particle-attached bacteria following a reservoir Microcystis

bloom. Science of the Total Environment, 660, 501-511.

Lu, J., Zhu, B., Struewing, I., Xu, N., & Duan, S. (2019). Nitrogen-phosphorus-associated

metabolic activities during the development of a cyanobacterial bloom revealed by

metatranscriptomics. Scientific reports, 9(1), 1-11.

Makarova, K. S., Wolf, Y. I., Snir, S., & Koonin, E. V. (2011). Defense islands in bacterial and

archaeal and prediction of novel defense systems. Journal of bacteriology,

193(21), 6039-6056.

Merel, S., Walker, D., Chicana, R., Snyder, S., Baurès, E., & Thomas, O. (2013). State of

knowledge and concerns on cyanobacterial blooms and cyanotoxins. Environment

international, 59, 303-327.

Metcalf, J. S., & Codd, G. A. (2012). Cyanotoxins. In Ecology of Cyanobacteria II (pp. 651-

675). Springer, Dordrecht.

Nalewajko, C., & Murphy, T. P. (2001). Effects of temperature, and availability of nitrogen and

phosphorus on the abundance of Anabaena and Microcystis in Lake Biwa, Japan: an

experimental approach. Limnology, 2(1), 45-48.

Paerl, H. W., & Huisman, J. (2008). Blooms like it hot. Science, 320(5872), 57-58.

Paerl, H. W., Hall, N. S., & Calandrino, E. S. (2011). Controlling harmful cyanobacterial

blooms in a world experiencing anthropogenic and climatic-induced change. Science of

the total environment, 409(10), 1739-1745.

Paerl, H. W., & Otten, T. G. (2013). Harmful cyanobacterial blooms: causes, consequences,

18 and controls. Microbial ecology, 65(4), 995-1010.

Paerl, H. W., Xu, H., Hall, N. S., Zhu, G., Qin, B., Wu, Y., ... & Joyner, A. R. (2014). Controlling

cyanobacterial blooms in hypertrophic Lake Taihu, China: will nitrogen reductions cause

replacement of non-N2 fixing by N2 fixing taxa?. PloS one, 9(11), e113123.

Paerl, H. W., & Otten, T. G. (2016). Duelling ‘CyanoHABs’: unravelling the environmental

drivers controlling dominance and succession among diazotrophic and non‐N2‐fixing

harmful cyanobacteria. Environmental microbiology, 18(2), 316-324.

Parveen, B., Ravet, V., Djediat, C., Mary, I., Quiblier, C., Debroas, D., & Humbert, J. F. (2013).

Bacterial communities associated with Microcystis colonies differ from free‐living

communities living in the same ecosystem. Environmental Microbiology Reports, 5(5),

716-724.

Patel, J. B. (2001). 16S rRNA gene sequencing for bacterial pathogen identification in the

clinical laboratory. Molecular diagnosis, 6(4), 313-321.

Pearson, L., Mihali, T., Moffitt, M., Kellmann, R., & Neilan, B. (2010). On the chemistry,

toxicology and genetics of the cyanobacterial , , , and

. Marine drugs, 8(5), 1650-1680.

Pfeifer, F. (2012). Distribution, formation and regulation of gas vesicles. Nature Reviews

Microbiology, 10(10), 705-715.

Ploug, H., Adam, B., Musat, N., Kalvelage, T., Lavik, G., Wolf-Gladrow, D., & Kuypers, M.

M. (2011). Carbon, nitrogen and O2 fluxes associated with the cyanobacterium Nodularia

spumigena in the Baltic . The ISME Journal, 5(9), 1549-1558.

19 Pollard, P. C., & Young, L. M. (2010). Lake viruses lyse cyanobacteria, Cylindrospermopsis

raciborskii, enhances filamentous-host dispersal in Australia. Acta Oecologica, 36(1), 114-

119.

Price, G. D., Badger, M. R., Woodger, F. J., & Long, B. M. (2008). Advances in understanding

the cyanobacterial CO2-concentrating-mechanism (CCM): functional components, Ci

transporters, diversity, genetic regulation and prospects for engineering into . Journal

of experimental botany, 59(7), 1441-1461.

Rantala, A., Fewer, D. P., Hisbergues, M., Rouhiainen, L., Vaitomaa, J., Börner, T., & Sivonen,

K. (2004). Phylogenetic evidence for the early evolution of microcystin synthesis.

Proceedings of the National Academy of Sciences, 101(2), 568-573.

Riemann, L., & Winding, A. (2001). Community dynamics of free-living and particle-

associated bacterial assemblages during a bloom. Microbial

ecology, 42(3), 274-285.

Rigosi, A., Carey, C. C., Ibelings, B. W., & Brookes, J. D. (2014). The interaction between

climate warming and eutrophication to promote cyanobacteria is dependent on trophic

state and varies among taxa. Limnology and Oceanography, 59(1), 99-114.

Sandrini, G., Matthijs, H. C., Verspagen, J. M., Muyzer, G., & Huisman, J. (2014). Genetic

diversity of inorganic carbon uptake systems causes variation in CO2 response of the

cyanobacterium Microcystis. The ISME journal, 8(3), 589-600.

Sandrini, G., Ji, X., Verspagen, J. M., Tann, R. P., Slot, P. C., Luimstra, V. M., ... & Huisman,

J. (2016). Rapid adaptation of harmful cyanobacteria to rising CO2. Proceedings of the

20 National Academy of Sciences, 113(33), 9315-9320.

Sarmento, H., & Gasol, J. M. (2012). Use of phytoplankton‐derived dissolved organic carbon

by different types of bacterioplankton. Environmental Microbiology, 14(9), 2348-2360.

Schindler, D. W. (1975). Whole-lake eutrophication experiments with phosphorus, nitrogen

and carbon: With 8 figures and 1 table in the text. Internationale Vereinigung für

theoretische und angewandte Limnologie: Verhandlungen, 19(4), 3221-3231.

Schopf, J. W. (2000). The record: tracing the roots of the cyanobacterial lineage. In The

ecology of cyanobacteria (pp. 13-35). Springer, Dordrecht.

Sellner, K. G., Brownlee, D. C., Bundy, M. H., Brownlee, S. G., & Braun, K. (1993).

Zooplankton grazing in a Potomac River cyanobacteria bloom. Estuaries, 16(4), 859-872.

Seymour, J. R., Amin, S. A., Raina, J. B., & Stocker, R. (2017). Zooming in on the :

the ecological interface for phytoplankton-bacteria relationships. Nature microbiology,

2(7), 1-12.

Sevilla, E., Martin‐Luna, B., Vela, L., Bes, M. T., Fillat, M. F., & Peleato, M. L. (2008). Iron

availability affects mcyD expression and microcystin-LR synthesis in Microcystis

aeruginosa PCC7806. Environmental microbiology, 10(10), 2476-2483.

Shao, K., Zhang, L., Wang, Y., Yao, X., Tang, X., Qin, B., & Gao, G. (2014). The responses of

the taxa composition of particle-attached bacterial community to the decomposition of

Microcystis blooms. Science of the total environment, 488, 236-242.

Shapiro, J. (1990). Current beliefs regarding dominance by blue-greens: the case for the

importance of CO2 and pH. Internationale Vereinigung für theoretische und angewandte

21 Limnologie: Verhandlungen, 24(1), 38-54.

Shakya, M., Lo, C. C., & Chain, P. S. (2019). Advances and challenges in metatranscriptomic

analysis. Frontiers in genetics, 10, 904.

Shelford, E. J., Middelboe, M., Møller, E. F., & Suttle, C. A. (2012). -driven nitrogen

cycling enhances phytoplankton growth. Aquatic microbial ecology, 66(1), 41-46.

Sivonen, K., & Jones, G. (1999). Cyanobacterial toxins. Toxic cyanobacteria in water: a guide

to their public health consequences, monitoring and management, 1, 43-112.

Teeling, H., Fuchs, B. M., Becher, D., Klockow, C., Gardebrecht, A., Bennke, C. M., ... &

Amann, R. (2012). -controlled succession of marine bacterioplankton

populations induced by a phytoplankton bloom. Science, 336(6081), 608-611.

Thomas, M. K., & Litchman, E. (2016). Effects of temperature and nitrogen availability on the

growth of invasive and native cyanobacteria. Hydrobiologia, 763(1), 357-369.

Van de Waal, D. B., Verspagen, J. M., Lürling, M., Van Donk, E., Visser, P. M., & Huisman, J.

(2009). The ecological stoichiometry of toxins produced by harmful cyanobacteria: an

experimental test of the carbon-nutrient balance hypothesis. Ecology letters, 12(12), 1326-

1335.

Verspagen, J. M., Van de Waal, D. B., Finke, J. F., Visser, P. M., Van Donk, E., & Huisman, J.

(2014). Rising CO2 levels will intensify phytoplankton blooms in eutrophic and

hypertrophic lakes. PloS one, 9(8), e104325.

Visser, P. M., Verspagen, J. M., Sandrini, G., Stal, L. J., Matthijs, H. C., Davis, T. W., ... &

Huisman, J. (2016). How rising CO2 and global warming may stimulate harmful

22 cyanobacterial blooms. Harmful Algae, 54, 145-159.

Villareal, T. A., & Carpenter, E. J. (2003). Buoyancy regulation and the potential for vertical

migration in the oceanic cyanobacterium Trichodesmium. Microbial ecology, 45(1), 1-10.

Walsby, A. E. (1985). The permeability of to the gases nitrogen and oxygen.

Proceedings of the Royal society of London. Series B. Biological sciences, 226(1244),

345-366.

Walsby, A. E., Hayes, P. K., Boje, R., & Stal, L. J. (1997). The selective advantage of buoyancy

provided by gas vesicles for planktonic cyanobacteria in the Baltic Sea. New Phytologist,

136(3), 407-417.

Wang, K., Razzano, M., & Mou, X. (2020). Cyanobacterial blooms alter the relative

importance of neutral and selective processes in assembling freshwater bacterioplankton

community. Science of The Total Environment, 706, 135724.

Weisburg, W. G., Barns, S. M., Pelletier, D. A., & Lane, D. J. (1991). 16S ribosomal DNA

amplification for phylogenetic study. Journal of bacteriology, 173(2), 697-703.

Woodhouse, J. N., Kinsela, A. S., Collins, R. N., Bowling, L. C., Honeyman, G. L., Holliday,

J. K., & Neilan, B. A. (2016). Microbial communities reflect temporal changes in

cyanobacterial composition in a shallow ephemeral freshwater lake. The ISME journal,

10(6), 1337-1351.

Xu, H., Zhu, G., Qin, B., & Paerl, H. W. (2013). Growth response of Microcystis spp. to iron

enrichment in different regions of Lake Taihu, China. Hydrobiologia, 700(1), 187-202.

Yang, C., Wang, Q., Simon, P. N., Liu, J., Liu, L., Dai, X., ... & Luo, F. (2017). Distinct network

23 interactions in particle-associated and free-living bacterial communities during a

Microcystis aeruginosa bloom in a plateau lake. Frontiers in microbiology, 8, 1202.

Zehr, J. P. (2011). Nitrogen fixation by marine cyanobacteria. Trends in microbiology, 19(4),

162-173.

Zheng, Q., Lin, W., Wang, Y., Li, Y., He, C., Shen, Y., ... & Jiao, N. (2020). Highly enriched N‐

containing organic molecules of lysates and their rapid transformation by

heterotrophic bacteria. Limnology and Oceanography.

Zilliges, Y., Kehr, J. C., Meissner, S., Ishida, K., Mikkat, S., Hagemann, M., ... & Dittmann, E.

(2011). The cyanobacterial hepatotoxin microcystin binds to proteins and increases the

fitness of Microcystis under oxidative stress conditions. PloS one, 6(3), e17615.

24 CHAPTER II

CYANOBACTERIAL BLOOMS ALTER THE RELATIVE IMPORTANCE OF NEUTRAL

AND SELECTIVE PROCESSES IN ASSEMBLING FRESHWATER

BACTERIOPLANKTON COMMUNITY

PREFACE

This chapter was published in the Science of The Total Environment journal under the title “Cyanobacterial blooms alter the relative importance of neutral and selective processes in assembling freshwater bacterioplankton community” Volume 706: 135724. Reprinted here with permission of the publisher. The author list is as follows: Kai Wang, Mandy Razzano, and

Xiaozhen Mou. Kai Wang performed all of the experiments, did data analysis, and wrote this manuscript. Mandy Razzano helped collect samples, measure phytoplankton biomass and physicochemical variables. Xiaozhen Mou, who is also my dissertation advisor, helped me with experiment design, data analysis, and writing of this manuscript.

ABSTRACT

CyanoHABs have substantial impacts on the functioning and sustainability of freshwater ecosystems by restricting light penetration, depleting dissolved oxygen, and producing various toxins. This study combined physicochemical variable measurements, 16S

25 rRNA gene sequencing, and microscopy observations to examine mechanisms that govern the assembly of bacterioplankton communities following the progress of CyanoHABs in a freshwater reservoir. Throughout the sampling season, bacterioplankton distribution patterns were well predicted by a neutral model, which assumes passive dispersal and ecological drift as the predominant mechanisms for community assembly. The neutral model consistently explained the distribution of over 67% of bacterioplankton OTUs and its fit was weaker during the bloom stage (R2 = 0.322) than the before- (R2 = 0.549) and after-bloom stages (R2 = 0.535).

Variations of environmental factors, acting as selective pressures, explained shifts of non- neutral OTUs (above/under neutral prediction) (63.9%) better than neutral OTUs (34.5%). Co- occurrence network analysis organized microbial communities into modules and revealed strong positive correlations between bacterioplankton and cyanobacteria than with planktonic algae and zooplankton. Overall, our results suggest that neutral processes play significant roles in assembling bacterioplankton communities over a cyanobacterial bloom succession and its relative importance was weakened by biotic pressures (interspecific interactions) during the bloom period. Our results also indicate that among biotic factors, cyanobacteria had greater impacts on bacterioplankton community assembly than planktonic algae and zooplankton.

INTRODUCTION

Microorganisms are highly diverse and abundant, and they play important roles in environmental processes (Worden et al., 2015), industrial production (Wagner et al., 2002), and human health (Tremaroli and Bäckhed, 2012). Microorganisms realize their ecological

26 functions largely in the form of a community rather than as individuals (Lepp et al., 2004).

Thus, elucidating mechanisms that shape microbial community composition is essential for understanding their ecological functions in a variety of systems (Cira et al., 2018).

A vast number of studies have demonstrated that microbial community structures were shaped mainly by selective processes including abiotic (environmental conditions) and biotic

(interspecific and intraspecific interactions) factors (Burns et al., 2016; Xue et al., 2018). The so-called selective theory assumes that ecological traits differ among community members, which allows the differentiation of niches across all species within a community (Stegen et al.,

2012). While the selective theory well explained microbial community assembly in a variety of systems, it often fails to explain taxa co-occurrence in environments that host highly diverse communities (Hérault, 2007; Jabot et al., 2008; Jeraldo et al., 2012).

On the other hand, neutral theory has been found particularly useful in explaining microbial community with immense diversity, where specific ecological traits of each taxon are difficult to measure (Woodcock et al., 2007; Ofiţeru et al., 2010; Venkataraman et al., 2015).

The neutral theory assumes that species have no difference in ecological fitness and taxa distribution is predominately determined by neutral processes, such as passive dispersal and ecological drift (Chase and Myers, 2011; Burns et al., 2016).

It is broadly accepted that selective and neutral processes concurrently impact microbial community assembly (Cira et al., 2018). However, the relative importance between neutral and selective processes may alter both temporally and spatially. For example, selective processes were found overweigh neutral processes in determining subsurface microbial community

27 structures, and the relative importance of one another was maximized at both ends of an environmental variation gradient (Stegen et al., 2012). In addition, the transition between selective and neutral processes was found to depend on the fraction of immigrating organisms and the magnitude of fitness differences among community members in synthetic microbial communities (Cira et al., 2018).

CyanoHABs represent a common biological disturbance to freshwater environments and can bring substantial impacts on water quality, microbial diversity, and ecosystem function

(Fogg, 1969; Paerl, 1988). As common biological disturbances, CyanoHABs represent a threat to the water quality and ecosystem function (Paerl and Fulton 2006). The strength of ecological selection has been found to vary with different cyanobacterial succession states (e.g., bloom vs. non-bloom) (Berry et al., 2017). However, little is known about the importance of ecological selection relative to neutral processes in governing bacterioplankton community composition (BCC) at different stages of cyanobacterial blooms. We hypothesized that neutral processes may be less important during the bloom period due to the highly selective pressure caused by cyanobacterial bloom.

Barberton Reservoir (41°04' N, 81°37' W) was selected as a sampling site to test our hypothesis. Barberton Reservoir serves as a primary source for the city of

Barberton Ohio, USA. This area (196 acres) has a temperate climate with an annual mean rain of 38 inches, an annual mean snow of 41 inches, and an annual mean high and low temperatures of 15.3 °C and 5.5 °C, respectively (https://www.bestplaces.net/climate/city/ohio/barberton).

Based on Ohio EPA monitoring data, cyanobacterial blooms occurred annually in the Barberton

28 reservoir in recent years (Ohio EPA).

MATERIALS AND METHODS

Sample collection and processing

Surface water samples (0.5 m) were collected at three sites on 9 occasions across a 6- month period between May and October 2017. Water temperature (Temp), dissolved oxygen

(DO), pH, electrical conductivity (EC), and turbidity (formazin turbidity unit, FTU) were measured in situ using a multi-parameter sonde (YSI, 650 MDS, OH, USA).

Planktonic algae samples were collected from an integrated water column to a depth that is twice the Secchi depth using a Van Dorn bottle. Samples were preserved with Lugol’s solution (0.7 mL of stock solution per 100 mL sample). Zooplankton samples were collected with an 80 µm Wisconsin plankton net from the entire water column and preserved with 70% ethanol.

Raw surface water samples were immediately placed in designated prelabeled sample containers after collection. Then, samples (2 L) for DNA extraction were transported in an icebox back to the Kent State University within one hour and samples (2 L) for chemical analyses were sent overnight to the Ohio Environmental Protection Agency lab. Once arrived in the lab, water samples (1.5 L) were sequentially filtered through 3.0- and 0.2-μm-pore-size membrane filters (Pall Life Sciences, Ann Arbor, MI, USA). Cells that were collected on 0.2-

μm-pore-size membrane filters were frozen immediately and stored at -80 °C until DNA extraction.

29 The filtrates (50 mL) were collected in sterile conical centrifuge tubes and frozen at -

20 °C until analyses of dissolved organic carbon (DOC), dissolved nitrogen (DN), plus

+ (NOx), ammonium (NH4 ), and soluble reactive phosphorus (SRP). Another 1.8 mL of water samples that passed through 3-μm-pore-size membrane filters were preserved in triplicate with freshly made paraformaldehyde (1 % final concentration) and stored at room temperature until enumeration of bacterioplankton cells.

Chemical analyses

Nutrients were measured according to procedures described in standard EPA methods

+ (https://www.epa.state.oh.us/ddagw/labcert). Briefly, concentrations of NOx, NH4 , and SRP were determined based on the automated hydrazine reduction method (Ohio EPA Method

4500), phenate method (Ohio EPA method 4500B), and colorimetry method (USEPA method

365.1), respectively. Dissolved organic carbon (DOC) and dissolved organic nitrogen (DON)

concentrations were determined with a TOC/TN analyzer (TOC-VCPN; Shimadzu Corp., Tokyo,

Japan) after samples were acidified overnight with 2M HCl to remove inorganic carbon. Total organic carbon (TOC) and total nitrogen (TN) were measured using unprocessed raw water samples following Ohio EPA Method 335.2, and USEPA Method 351.2, respectively.

- Bicarbonate (HCO3 ) was measured following the Ohio EPA method 2320B, a titration method using pH endpoints. Total phosphorus (TP) concentrations were determined using unprocessed raw water samples according to USEPA Method 365.4. Water samples (200 mL) for a (Chl a) analysis were covered with black plastic bags and transported to the lab in an icebox. Samples were filtered onto pre-combusted (450 oC for 4 hours) glass fiber filters

30 (GF/F, Whatman International Ltd., Maidstone, England) in subdued light. (Chl a) concentrations were measured by spectrophotometry according to procedures described in

Ohio EPA Method 445.0.

Bacterioplankton cell abundance measurement

Bacterioplankton cells were counted following a flow cytometry protocol described previously (Mou et al., 2011). Briefly, preserved samples (1.8 mL) were stained with 200 μL of SYBR Green I (1:5000 dilution of the commercial stock; Molecular Probes Inc. Eugene,

Oregon, USA), and then incubated for 20 min in the dark at room temperature. Internal standards, i.e., 20 μL beads (5.14-μm-diameter; Spherotech Inc., Lake Forest, IL, USA), were then added into each sample with a final concentration of 1×105 beads mL-1. Samples were then examined using a BD-FACS Aria CaliburTM flow cytometer (BD, Franklin Lakes, NJ,

USA) with a 488 nm argon laser.

Phytoplankton and zooplankton identification and enumeration

Planktonic algae and cyanobacteria were microscopically identified and quantified by

BSA Environmental Services (Beachwood, Ohio). Cell abundance data were then converted to biovolumes for each taxon using standard cell sizes published in the “Biovolume Calculator”

(https://www.water.vic.gov.au/data/assets/excel_doc/ 0027/65592/BIOVOLUME-

CALCULATOR.XLSX). Zooplankton cells were enumerated with the aid of a compound microscope to the family level based on published keys (Witty, 2004; Haney et al., 2013).

PCR and sequencing

DNA was extracted from cells collected on 0.2-μm-pore-size membrane filters using

31 PowerSoil DNA extraction kits (Mo Bio Laboratories, Inc., Carlsbad, CA, USA) according to the manufacturer’s instructions. The V3-V4 regions of 16S rRNA genes were PCR amplified using primers 341F (5’-CCTAYGGGRBGCASCAG-3’) and 806R (5’-

GGACTACNNGGGTATCTAAT-3’) (Muyzer et al., 1993). PCR program consisted of an initial denaturation at 96 oC for 10 min, followed by 30 cycles of denaturation at 94 oC for 30 s, annealing at 55 oC for 10 s, and extension at 72 oC for 45 s, and a final extension at 72 oC for

10 min. Amplicons were purified using a Qiagen gel extraction kit (Qiagen, Chatsworth, CA,

USA), and then quantified with Quanti-iT PicoGreen dsDNA assay kit (Life Technologies,

Carlsbad, CA, USA). Purified PCR amplicons were sequenced by the HiSeq2500 platform at the Novogene Corporation Inc (Beijing, China). Raw sequences were deposited in NCBI short read archive database under BioProject number PRJNA517173.

Bioinformatic analysis

Sequence processing was performed using QIIME version 1.9.0 (Caporaso et al., 2010).

Reads with an average Phred score < 30, read length < 200 bp, containing ambiguous bases or homopolymers of length > 8 bp were removed. Then chimeric sequences were identified by uchime (Edgar et al., 2011) and removed from the sequence pool. The UCLUST pipeline was used to pick up operational taxonomic units (OTUs) at 97% sequence identity level (Edgar,

2010). OTU singletons were removed from the sequence pool to minimize impacts from sequencing errors and prevent overestimation of bacterioplankton diversity. Representative sequences of OTUs (the longest sequence within each OTU group) were annotated using the

RDP classifier based on the SILVA database version 128 (Quast et al., 2012). OTUs that were

32 affiliated with cyanobacteria and were removed from sequence libraries. Sequence libraries were then normalized to the same depth, i.e., the number of sequences, by random resampling each sequencing library based on the size of the smallest library using QIIME.

Alpha diversity (i.e., Shannon index and Chao 1) were also estimated using QIIME.

Statistical analyses

Statistical analyses were performed using the vegan package (Oksanen et al., 2011) in

R version 3.5.1 unless otherwise mentioned. Differences in biotic (bacterioplankton, cyanobacteria, planktonic algae, and zooplankton data) and abiotic (physicochemical variables) variables were examined using analysis of variance (ANOVA). Bacterioplankton community composition was examined based on 16S rRNA gene amplicon sequencing data

(cyanobacterial sequences were excluded), while community composition of cyanobacterial, planktonic algae, and zooplankton cell abundances and community composition were based on microscopic observations.

Temporal dynamics of bacterioplankton, cyanobacterial, planktonic algae, and zooplankton community structures and environmental conditions were examined by time-lag analyses (Collins et al., 2000). Briefly, Bray-Curtis dissimilarity was computed for community compositions, while Euclidean dissimilarity was computed for environmental variables. These two sets of dissimilarity values were then plotted against all-time lags (Collins et al., 2000).

Indicator OTUs were used to explain the temporal distribution pattern in bacterioplankton diversity (Xue et al., 2018). Indicator OTUs of bacterioplankton at before-, during- and after- cyanobacterial bloom periods were identified using the indicspecies package (Cáceres et al.,

33 2009). OTUs were designated as indicators when indicator values > 0.7 and P-value < 0.05 based on 1000 permutations.

Neutral community model analysis

A neutral community model analysis was performed to examine the potential contribution of neutral processes to the taxonomic assembly of bacterioplankton communities

(exclude cyanobacteria). Specifically, the Sloan neutral model was used to predict the relationship between the observed frequency of occurrence of OTUs and their relative abundance (Sloan et al., 2006). This model assumes that abundant taxa are widespread because they are more likely to disperse by chance, while rare taxa will be lost with time due to dispersal limitation. The migration rate (m) was calculated to estimate the dispersal limitation of bacterioplankton (Burns et al., 2016), which was determined based on non-linear least-square fitting using minpack.lm package (Elzhov et al., 2010). The overall fitting of the neutral model was assessed by calculating the generalized R-squared (1-the sum of squares of residuals/the total sum of squares) (Östman et al., 2010). R code used to fit the neutral model was modified from a previous paper (Burns et al., 2016).

We further compared the taxonomic differences of OTUs that did and did not fit the neutral model. Firstly, samples belonging to the same sampling time group (before, during, or after bloom) were pooled, and then OTUs from this pool were assigned into three subgroups depending on whether they fit the model (above, below, or within the model) according to 95% confidence interval. Each subgroup was then considered as a distinct community sample for further analysis. To further examine the variation in phylogenetic composition, pairwise

34 unweighted UniFrac distances were calculated among neutral and non-neutral subgroups using the GUniFrac package in R. Phylogenetic tree was constructed according to phylogenetic sampling theory described in O’Dwyer et al. (2012) using the picante package (Kembel et al.,

2010) in R. Non-metric multidimensional scaling was performed to visualize difference among neutral and non-neutral subgroups based on UniFrac distances.

Correlations between environmental variables and bacterioplankton communities

A step-wise distance-based redundancy analysis (dbRDA) was performed to examine potential contributions of environmental variables to observed variations within the bacterioplankton communities (based on all or just the non-bloom samples), neutrally distributed bacterioplankton OTUs, non-neutrally distributed bacterioplankton OTUs, and cyanobacterial, planktonic algae, zooplankton cell counts. Mantel tests were also conducted to examine individual environmental variables correlated with the dynamics of cyanobacterial, planktonic algae, and zooplankton communities.

Network analysis

Network analysis was performed to explore the relationships between members within microbial communities based on all or just the non-bloom samples. Briefly, pair-wise correlations of bacterioplankton communities, cyanobacterial, planktonic algae, and zooplankton communities were calculated using the maximal information coefficient (MIC) analysis. This allowed detections of linear and non-linear associations between variables. To reduce complexity, only bacterioplankton OTUs that presented in five or more samples and contained more than 20 sequences were retained for the construction of networks. Pairwise

35 MIC values were calculated using the MINE package (Reshef et al., 2011). False discovery rates (FDR) of MIC values were determined using the Benjamini-Hochberg method

(Benjamini et al., 1995). A correlation was regarded as significant when FDR < 5% (Wang et al., 2018). In the present study, this FDR criterion was equivalent to MIC > 0.41. The matrix of MIC values > 0.41 was then used to visualize the network associations using Gephi version

0.9.2 (Bastian et al., 2017). Node-level topological properties (i.e., modularity, degree, network diameter, clustering coefficient, and path length) were further calculated using the igraph package (Csardi et al., 2006). In addition, 1000 Erdös-Réyni random networks, which had an identical number of nodes and edges, were generated and compared with the whole network

(Erds and Rényi, 1960) to examine network properties.

Keystone species were defined as taxa that other species largely depend upon, therefore the community structure would drastically change upon their removal (Berry et al., 2014). In co-occurrence networks, keystone species were identified based on a high degree (> 100) and low betweenness centrality values (< 5000) (Ma et al., 2016). Network hubs and connectors were identified according to taxa connectivity as described by Guimerà and Amaral (Guimera et al., 2005).

RESULTS

Abundance and structural dynamics of cyanobacterial, planktonic algae and zooplankton

During the sampling period, the total cyanobacterial biovolume (6.1-21.3 mm3 L-1) showed an increasing trend from May to August and then decreased to the initial level in

36 October (Fig. 1a). Severe cyanobacterial bloom appeared only in August samples (Fig. 1a), based on guidelines (biovolume >10 mm3 L-1) of the United States Environmental Protection

Agency (USEPA). The bloom was contributed by four cyanobacterial genera including

Planktolyngbya, Aphanizomenon, Pseudanabaena, and Cylindrospermopsis. Subsequently, samples were divided into three groups based on sampling time and the levels of cyanobacterial biovolume: before-bloom samples (May-July), during-bloom samples (August), and after- bloom samples (September-October).

A total of 22 morphologically distinguishable cyanobacterial genera were identified microscopically, with , Planktolyngbya, Aphanizomenon, Pseudanabaena, and

Cylindrospermopsis as the most abundant genera according to cell abundance (Fig. 1b).

Members of these five major cyanobacterial genera accounted for > 90% of the total cyanobacteria and their distribution varied temporally. Chroococcus was predominant in May

(5.7 ´ 108 cell/L), while the Planktolyngbya (4.6 ´ 108 cell/L), Aphanizomenon (2.3 ´ 108 cell/L), Pseudanabaena (2.1 ´ 108 cell/L), and Cylindrospermopsis (3.7 ´ 107 cell/L) peaked their cell abundances in August (one-way ANOVA; P < 0.05). When considering all 22 identified cyanobacterial genera, a significant directional change of community structure was also observed by time-lag regression analysis (Fig. 2a).

The total biovolume of planktonic algae followed the same temporal pattern as cyanobacteria (0.3-16.7 mm3 L-1; Fig. 1c), with peak values occurred in August (Fig. 1c). A total of 92 morphologically distinguishable algal genera were identified microscopically. Major genera (accounted for > 90% of the total community) included Cyclotella, and Aulacoseira of

37 Diatoms, Carteria and Chlamydomonas of Green algae, Plagioselmis and Cryptomonas of

Cryptophyta and Euglena of Euglenophyta (Fig. 1d). Aulacoseira, Cryptomonas, and Euglena were predominant during-bloom period, while Chlamydomonas and Plagioselmis were more abundant after-bloom period (one-way ANOVA, P < 0.05). However, a directional change was not identified by time-lag analysis for planktonic algal communities when considering all identified 92 algal genera (one-way ANOVA, P >0.05; Fig. 2b).

Zooplankton community showed a different temporal distribution pattern from those of cyanobacterial and planktonic algae communities. The number of zooplankton ranged between

13 and 106 individuals L-1and the peak values of zooplankton cells observed in late June and

July (Fig. 1e), early than the peak time of cyanobacteria and planktonic algae communities

(Figs. 1a and 1c). A total of 13 morphologically distinguishable zooplankton families were identified microscopically (Fig. 1f). The dominant zooplankton families were Brachionus,

Conochilus, Daphnia, Diaptomidae, Eubosmina, Microcyclops, and Tylotrocha (Fig. 1f).

Eubosmina and Microcyclops were more abundant in May and early June, Daphnia was more abundant in late June, while the rest four major families were more abundant in July (one-way

ANOVA, P < 0.05). Time-lag regression analysis also exhibited a significant directional change through sampling periods based on all identified zooplankton families (Fig. 2c).

Abundance and structural dynamics of bacterioplankton community

Cell counts of bacterioplankton ranged between 7.6 ´ 106 and 10.7 ´ 106 cells/mL during the sampling period (Fig. 1g). The temporal variation of bacterioplankton abundance was distinct from those of cyanobacteria, planktonic algae, or zooplankton (Fig. 1g). The

38 maximum bacterioplankton cell number was delayed to September when cyanobacteria number abundance dropped from its peak in August (Fig. 1g).

Sequencing analysis of the V3-V4 regions of the 16 rRNA genes identified 3196 bacterioplankton OTUs at 0.03-distance threshold. Rarefaction curves almost approached plateaus for all samples, reflecting good recovery rate at the OTU level (Fig. 7). Shannon index ranged between 5.2-6.5 and showed a decreasing trend from May to October (Fig. 8a). Chao 1 index ranged between 477-507 and showed no significant change throughout sampling periods

(Fig. 8b). Neither index exhibited significant differences among bloom succession states

(before-, during-, and after-bloom) (one-way ANOVA, P > 0.05).

Taxonomy annotations revealed a reservoir bacterioplankton community dominated by

Actinobacteria, Bacteroidetes and Proteobacteria of the Alpha-, Beta-, and Gamma lineages; these taxa collectively account for > 90% of the total sequences. Taxonomic composition of bacterioplankton communities had a distinct temporal distribution pattern from those of planktonic algae and zooplankton communities (Fig. 1h). Bacterioplankton communities were dominated by 15 genera, which collectively accounted for > 80% of the total sequences.

Pseudomonas, Aquabacterium and an uncultured SAR11 genus were enriched during cyanobacterial blooms, whereas other bacterioplankton genera had higher relative abundance either before or after cyanobacterial blooms (p < 0.05; Fig. 1h).

Like for cyanobacteria and zooplankton, time-lag regression analysis identified a significant directional change of bacterioplankton communities (OTUs) (Fig. 2d). However, the regression slope of bacterioplankton communities was gentler compared with those of

39 cyanobacteria, and zooplankton communities (Fig. 2a and 2c). Consistently, at broader taxonomic levels (i.e., phylum, class, order, family and genus, Fig. 9), bacterioplankton communities showed insignificant or gentler regression slopes compared with those of cyanobacterial and zooplankton communities.

Indicator OTUs for before- (39 OTUs), during- (56 OTUs) and after-bloom (21 OTUs) periods were identified, and they represented significantly different bacterioplankton genera

(Table 1). Verrucomicrobiaceae, Caulobacteraceae, and Luteolibacter indicator OTUs were most abundant in before-bloom period (P < 0.05). Indicator OTUs of Flavobacterium,

Neisseriaceae and one uncultured Proteobacteria genus were dominant in the during bloom period (P < 0.05). Indicator OTUs of bacteriumGLA1 and an unidentified genus of

Alphaproteobacteria and Parcubacteria were predominant in the after-bloom period (P < 0.05).

Importance of neutral processes on bacterioplankton community assembly

Overall, the neutral model well explained the observed frequency of most bacterioplankton OTUs. Over three quarters (78.4%), or 2346, of total OTUs (2992) had observed frequencies falling within the 95% confidence intervals of the neutral model (Fig. 3a).

Only a small number of bacterioplankton OTUs occurred more (408 OTUs, 13.6%) or less

(237 OTUs, 8.0%) frequently than predicted by the neutral model during the entire sampling period (Figs. 3b-d). The fraction of neutral OTUs (before: 84%; during: 67%; after: 85%) altered across bloom development periods and was the lowest during the bloom period. The fit of the neutral model for bacterioplankton communities varied over cyanobacteria succession periods and was the weakest at the during-bloom period (R2 = 0.322, Fig. 3c).

40 Migration rates were calculated to estimate bacterioplankton dispersal ability across cyanobacterial bloom succession. The results showed that migration rates of bacterioplankton were higher in the before- (m = 0.248) and after- (m = 0.257) bloom periods than the during- bloom period (m = 0.122) (Fig. 4a), showing that bacterioplankton communities became increasingly dispersal limited during the bloom period.

Bacterioplankton OTUs with observed frequencies above, below or within 95% confidence intervals of neutral model prediction were designated as overrepresented, underrepresented, and neutral OTUs, respectively. The above three types of OTUs consisted of phylogenetically distinct groups. Compared with neutral OTUs, over- or under-represented

OTUs was more closely clustered phylogenetically (Fig. 4b) and their taxonomic composition largely remained during different stages of bloom succession. Overrepresented bacterioplankton OTUs were consistently dominated by Aeromonas, Legionella, OM27 clade,

Methylomonas, and uncultured_Rickettsiales OTUs (P < 0.01) across the entire bloom succession. Meanwhile, underrepresented bacterioplankton OTUs were consistently dominated by Lachnospira, Phaselicystis, Hyphomicrobium, uncultured and uncultured Caulobacter (P < 0.01) (Fig. 4c).

Importance of selective processes on bacterioplankton communities

A dbRDA analysis of all samples demonstrated that temporal distribution of bacterioplankton community could be largely explained by Temp, EC, Chl a, DON, TOC, and

- HCO3 , which together accounted for 25.4% of the observed variations (Fig. 5a). A second dbRDA analysis, utilizing only non-bloom samples, identified the same subset of

41 environmental variables could explain higher of the observed bacterioplankton variations (Fig.

5b, 40.2%) compared to using all (bloom and non-bloom) samples. The temporal distribution of other microbial communities, i.e, cyanobacterial, planktonic algae, and zooplankton communities, could also be explained by a similar subset of environmental variables (Fig. 10).

+ - These included Temp, EC, pH, NH4 , NOx, DON, SRP, TOC, or HCO3 , which together accounted for 21.6-52.8% of the observed variations (Fig. 10). Results of direct Mantel test further confirmed that each one of these variables was significantly correlated with variations of bacterioplankton, cyanobacterial, planktonic algae and zooplankton communities (Fig. 5c).

When examining neutral OTUs and over-/under-represented OTUs separately, environmental variables appeared to explain shifts of over-/under-represented OTUs (63.9%; Fig. 11a) better than neutral OTUs (34.5%; Fig. 11b). Consistently, the Mantel test found that environmental variables were correlated more strongly with non-neutral communities (both over- and under- represented OTUs) than neutral communities.

Microbial community network was performed to examine interactions within bacterioplankton communities and among communities of all studied microorganisms including cyanobacteria, planktonic algae, and zooplankton (Fig. 6). The produced co- occurrence network consisted of 1313 OTU nodes, 11472 positive edges and 2568 negative edges (Fig. 6a). This network had a characteristic of a scale-free, non-random and modular network topography (power-law distribution: R2=0.873; Fig. 12). Consistently, when compared with an Erdös-Réyni random network, the obtained network (Fig. 6a) exhibited greater values of modularity, average clustering coefficient, and average path length (Table 2).

42 The obtained network grouped microbial communities into 6 major modules, which collectively accounted for more than 90% of the taxa (Figs. 6b-g). Ternary plots showed that, except for Module 2, structures of individual modules were strongly determined by taxonomic relatedness and exhibited clear temporal distribution patterns (Figs. 6b-g). Generally, module

1 OTUs were dominated by Bacteroidetes and Cyanobacteria in the before- and during-bloom periods. Modules 3 OTUs were dominated by Bacteroidetes and Betaproteobacteria in the before-bloom period. Module 4 OTUs were dominated by and

Gammaproteobacteria in the during-bloom period. Module 5 OTUs were dominated by

Deltaproteobacteria and in the before- and after-bloom periods.

Module 6 OTUs were dominated by in the during-bloom period.

A second network was built to investigate the potential correlation between module

- + OTUs and environmental conditions across bloom successions. HCO3 , EC, TOC, Temp, NH4 ,

and NOx appeared as important drivers of network connections (Fig. 13). Zooplankton and planktonic algae had lower degree values with both biotic and abiotic variables compared with cyanobacterial communities (P < 0.05, Fig. 13). Another network analysis, utilizing only non- bloom samples, revealed that the numbers of relationships (73 positive and 18 negative) between bacterioplankton and cyanobacteria was lower than those generated in a network analysis based on all samples (189 positive and 59 negative; Fig. 14).

Highly linked OTUs within (hub) or between (connector) modules were identified based on their connectivity within the microbial community network. A total of 23 module hubs (i.e., highly linked OTUs within their own module) and 17 connectors, (i.e., highly linked

43 OTUs between different modules) were further identified from network based on the value of connectivity (Fig. 13). They primarily belonged to module 1 (2 hubs, 8 connectors), 2 (6 hubs,

2 connectors), 3 (0 hubs, 4 connectors) and 4 (15 hubs, 3 connectors). Majority of the hubs (15 out of 23) and connectors (13 out of 17) was neutral OTUs (Table 3). The taxonomic compositions were different between network hubs and connectors. Hub OTUs were mainly affiliated with Alphaproteobacteria, Gammaproteobacteria, and Actinobacteria, while the connector OTUs were mainly members of Betaproteobacteria and Flavobacteriia.

A total of 11 OTUs were assigned as keystone species according to the degree (> 100) and betweenness centrality values (< 5000). These keystone species were OTUs affiliated with

Sphingobacteriales (1 OTU), Rhizobiaceae (1 OTU), Acetobacteraceae (1 OTUs), Opitutus

(1 OTU), (2 OTUs), and Verrucomicrobiaceae (5 OTUs) (Table 4). Except for one, all identified keystone species appeared to be underrepresented OTUs (Table 4).

DISCUSSION

Cyanobacterial blooms represent a series of complex ecological disturbances to freshwater systems, threating water quality and ecosystem function (Paerl and Fulton 2006).

The strength of this ecological disturbance varies with different cyanobacterial succession states and may, in turn, alter natural microbial community structure (Berry et al., 2017). Our results supported the hypothesis that the relative importance of neutral and selective processes changes across bloom succession states (i.e., bloom vs non-bloom) in bacterioplankton community assembly.

44 The importance of neutral processes to BCC

To quantify the relative importance of neutral and selective processes is a central challenge to studies on mechanisms of microbial community assembly (Xue et al., 2018). The neutral theory assumes that microbial communities are shaped by passive dispersal and ecological drift (Chase and Myers, 2011; Burns et al., 2016). In the present study, the neutral model well explained the bacterioplankton distribution patterns by only incorporating the effects of random dispersal (Fig. 3a). In addition, most of the identified network hubs and connectors were within neutral model prediction (Table 3), suggesting the importance of neutral process in maintaining the stability of microbial community structures. During the bloom period, the ability of the neutral model to predict the distribution of bacterioplankton

OTUs decreased compared to before and after bloom periods (Fig. 3), indicating that neutral processes became less important. This was probably due to cyanobacterial bloom, which caused perturbation (Moustaka-Gouni et al., 2006) and led to increased selective pressures

(abiotic or biotic) to BCCs in the freshwater system. In addition, the decrease in the fit of the neutral model was accompanied by a decrease in the estimated migration rate (Fig. 3a), which suggests that blooms of cyanobacteria may also reduce the dispersal ability of bacterioplankton.

This was in line with a previous study, which showed that reductions in immigration and dispersal ability can lessen the relative importance of neutral processes in regulating microbial communities (Cira et al., 2018).

Within each bloom succession period, there were a number of bacterioplankton OTUs that were not randomly distributed (above or below neutral model prediction), implying that

45 they are ecologically different from bacterioplankton OTUs within the neutral model prediction

(Burns et al., 2011; Xue et al., 2018). These over- and under-represented bacterioplankton

OTUs are likely taxa that are specifically selected by and against the environment, respectively

(Burns et al., 2016; Cira et al., 2018). Consistently, environmental factors were found to have greater impacts on non-neutral OTUs than neutral OTUs (Fig. 5). Further, the over- and under- represented bacterioplankton OTUs of the neutral model prediction (Fig. 4c) were closely phylogenetically clustered (Fig. 4b), suggesting that the environmental conditions select bacterioplankton taxa based on a specific set of phylogenetically conserved features (Stegen et al., 2012). This is in contrast to neutrally distributed OTUs, which exhibited greater phylogenetic variation across bloom succession than non-neutral bacterioplankton OTUs (Fig.

4b). This pattern indicated that the neutrally distributed bacterioplankton OTUs are less likely to be specifically adapted to the environment and their distributions in the environment are mainly the result of random dispersal and ecological drift (Stegen et al., 2012).

The importance of environmental factors to BCC

Impacts of environmental conditions and species interactions, two important components of selective pressures, on microbial community composition have been repeatedly observed in freshwater systems (Woodhouse et al., 2016; Xue et al., 2018). Our study observed that the explained BCC variations by environmental factors decreased when bloom samples were included (Figs. 5a and b), indicating a less important role of environmental factors in shaping BCCs during the bloom stage than the non-bloom stages.

Different indicator OTUs were identified across three bloom succession stages (Table

46 1). This indicates that shifts of bacterioplankton communities are likely resulted from differential responses of bacterioplankton taxa to environmental heterogeneity (multiple niches) during cyanobacterial bloom events. Moreover, all of the identified keystone species (except one) were non-neutral OTUs, further suggesting that selective processes might also play important roles in selecting bacterioplankton taxa that have fundamental impacts on the structure of microbial communities.

The importance of interspecific interactions to BCC

Microbial interactions are also dominant drivers in shaping the dynamics of community composition (Ju and Zhang, 2015). Our results found a co-occurrence network with non- random and module structure properties, which have been reported in several previous ecological networks of microbial communities (Woodhouse et al., 2016; Zhao et al., 2016; Hu et al., 2017). This pattern emphasizes the influence of interspecific interactions on bacterioplankton community assembly (Ju et al., 2014; Konopka et al., 2015) and indicates that closely connected bacterioplankton OTUs are also ecologically related, which in turn reflects their common niche preference or the mutualistic correlations (Ju et al., 2014).

In addition, a network analysis that used only non-bloom samples revealed a lower number of relationships (91 relations) between bacterioplankton and cyanobacteria than that used all of the samples (248 relations, Fig. 13). This suggests that the bloom event increased the relative importance of biotic factors in regulating BCCs. This complex interspecific interactions between bacterioplankton and cyanobacteria likely reduce the sensitivity of BCCs to environmental stimuli (Konopka et al., 2015). This is because complex interaction networks

47 can serve as an ecological buffer against the changing environment (Konopka et al., 2015); for example, metabolic pathways can be achieved alternatively by engaging different groups of interacting functional agents (Konopka et al., 2015).

More positive relationships were identified than negative ones among microorganisms within the observed co-occurrence network, suggesting that interspecific cooperation might contribute to the stability and resilience of the microbial communities (Ju et al., 2014; Xue et al., 2018). In addition, rare taxa had more positive interactions than abundant taxa, which has been reported in previous studies (Liu et al., 2014; Xue et al., 2018; Liu et al., 2019). Rare taxa provided a high diversity, which could increase the functional redundancy (Caron and

Countway, 2009) and provide a biological buffer to microbial communities against the environmental disturbance (Yachi and Loreau, 1999).

The composition of phytoplankton community governs the quality and quantity of dissolved organic carbon (Kent et al., 2006), which in turn influence the structure and activities of bacterioplankton communities (Cole, 1982; Liu et al., 2014). Clear shifts of planktonic algae and cyanobacterial community compositions were observed in our study (Fig. 1), which might play important roles in shaping bacterioplankton community assembly. Zooplankton exhibited lower degree values of connection with bacterioplankton and cyanobacterial communities in the co-occurrence network (Fig. 13), suggesting that they had weaker relationships with bacterioplankton and cyanobacterial communities, therefore play less important roles in shaping bacterioplankton communities. This is in contrast with previous studies, which suggest that cyanobacterial biomass had stronger correlations with eukaryotic communities (Xue et al.,

48 2018; Liu et al., 2019). This discrepancy could be partly caused by differences in analytical methods on eukaryotic communities in our (microscopy) and others (sequencing) studies (Xue et al., 2018; Liu et al., 2019). The high abundance of cyanobacteria under nutrient-rich conditions may also overwhelm the effects of less abundant zooplankton (Paerl and Otten,

2013).

Integrating results from neutral community model, RDA and network analyses, our results emphasize that neutral and selective (abiotic and biotic) processes can act concurrently to regulate the assembly of bacterioplankton communities. However, the relative importance between neutral and selective processes in shaping BCCs may vary over different bloom succession stages. During the bloom stage, interspecific interactions (biotic factors of selective processes) may be more important in shaping BCC variations.

Our study focused on free-living bacterioplankton (FLB), which are thought to play more important roles in utilizing DOM than particle-associated bacteria (PAB) (Karner and

Herndl, 1992) and account for most of the total bacterial production during a bloom (Middelboe et al., 1995). PAB has been shown to respond differently to cyanobacterial bloom from the FLB

(Liu et al., 2014). Therefore, there might be a different mechanism underlying PAB community assembly. Nevertheless, our study can expand current understandings of ecological mechanisms for FLB community assembly in changing environments.

CONCLUSIONS

In conclusion, our results demonstrated that cyanobacterial bloom significantly altered

49 the bacterioplankton community composition without affecting the alpha diversity. During successions of cyanobacterial blooms, neutral process might be the foundation of bacterioplankton community assembly, and its relative importance varied depending on the abundance and dispersal ability of immigrant species. Our results also suggest that high and abundance of cyanobacteria may overwhelm the effects of planktonic algae and zooplankton on bacterioplankton community assembly.

50 REFERENCES

Bagatini, I. L., Eiler, A., Bertilsson, S., Klaveness, D., Tessarolli, L. P., & Vieira, A. A. H.

(2014). Host-specificity and dynamics in bacterial communities associated with bloom-

forming freshwater phytoplankton. PloS one, 9(1), e85950.

Bastian, M., Heymann, S., & Jacomy, M. (2009, March). Gephi: an open source software for

exploring and manipulating networks. In Proceedings of the International AAAI

Conference on Web and Social Media (Vol. 3, No. 1).

Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and

powerful approach to multiple testing. Journal of the Royal statistical society: series B

(Methodological), 57(1), 289-300.

Berry, D., & Widder, S. (2014). Deciphering microbial interactions and detecting keystone

species with co-occurrence networks. Frontiers in microbiology, 5, 219.

Berry, M. A., Davis, T. W., Cory, R. M., Duhaime, M. B., Johengen, T. H., Kling, G. W., ... &

Denef, V. J. (2017). Cyanobacterial harmful algal blooms are a biological disturbance to

Western Lake Erie bacterial communities. Environmental microbiology, 19(3), 1149-

1162.

Burns, J. H., & Strauss, S. Y. (2011). More closely related species are more ecologically

similar in an experimental test. Proceedings of the National Academy of Sciences,

108(13), 5302-5307.

Burns, A. R., Stephens, W. Z., Stagaman, K., Wong, S., Rawls, J. F., Guillemin, K., &

Bohannan, B. J. (2016). Contribution of neutral processes to the assembly of gut

51 microbial communities in the zebrafish over host development. The ISME journal, 10(3),

655-664.

Caporaso, J. G., Kuczynski, J., Stombaugh, J., Bittinger, K., Bushman, F. D., Costello, E.

K., ... & Knight, R. (2010). QIIME allows analysis of high-throughput community

sequencing data. Nature methods, 7(5), 335-336.

Cáceres, M. D., & Legendre, P. (2009). Associations between species and groups of sites:

indices and statistical inference. Ecology, 90(12), 3566-3574.

Chase, J. M., & Myers, J. A. (2011). Disentangling the importance of ecological niches from

stochastic processes across scales. Philosophical transactions of the Royal Society B:

Biological sciences, 366(1576), 2351-2363.

Cira, N. J., Pearce, M. T., & Quake, S. R. (2018). Neutral and selective dynamics in a

synthetic microbial community. Proceedings of the National Academy of Sciences,

115(42), E9842-E9848.

Collins, S. L., Micheli, F., & Hartt, L. (2000). A method to determine rates and patterns of

variability in ecological communities. Oikos, 91(2), 285-293.

Csardi, G., & Nepusz, T. (2006). The igraph software package for complex network research.

InterJournal, complex systems, 1695(5), 1-9.

Edgar, R. C. (2010). Search and clustering orders of magnitude faster than BLAST.

Bioinformatics, 26(19), 2460-2461.

Edgar, R. C., Haas, B. J., Clemente, J. C., Quince, C., & Knight, R. (2011). UCHIME

improves sensitivity and speed of chimera detection. Bioinformatics, 27(16), 2194-2200.

52 Elzhov, T. V., Mullen, K. M., Spiess, A. N., & Bolker, B. (2010). R interface to the

Levenberg-Marquardt nonlinear least-squares algorithm found in MINPACK. Plus

support for bounds, 1-2.

Erdős, P., & Rényi, A. (1960). On the evolution of random graphs. Publ. Math. Inst. Hung.

Acad. Sci, 5(1), 17-60.

Fogg, G. E. (1969). The Leeuwenhoek Lecture, 1968-The physiology of an algal nuisance.

Proceedings of the Royal Society of London. Series B. Biological Sciences, 173(1031),

175-189.

Guimera, R., & Amaral, L. A. N. (2005). Functional cartography of complex metabolic

networks. nature, 433(7028), 895-900.

Haney, J. F., Aliberti, M., Allan, E., Allard, S., Bauer, D., Beagen, W., ... & Doan, U. (2013).

An image-based key to the zooplankton of North America. University of New Hampshire

Center for Freshwater Biology.

Hérault, B. (2007). Reconciling niche and neutrality through the Emergent Group approach.

Perspectives in Ecology, Evolution and Systematics, 9(2), 71-78.

Hu, A., Ju, F., Hou, L., Li, J., Yang, X., Wang, H., ... & Yu, C. P. (2017). Strong impact of

anthropogenic contamination on the co‐occurrence patterns of a riverine microbial

community. Environmental microbiology, 19(12), 4993-5009.

Jabot, F., Etienne, R. S., & Chave, J. (2008). Reconciling neutral community models and

environmental filtering: theory and an empirical test. Oikos, 117(9), 1308-1320.

Jeraldo, P., Sipos, M., Chia, N., Brulc, J. M., Dhillon, A. S., Konkel, M. E., ... & Goldenfeld,

53 N. (2012). Quantification of the relative roles of niche and neutral processes in

structuring gastrointestinal . Proceedings of the National Academy of

Sciences, 109(25), 9692-9698.

Ju, F., Xia, Y., Guo, F., Wang, Z., & Zhang, T. (2014). Taxonomic relatedness shapes bacterial

assembly in activated sludge of globally distributed wastewater treatment plants.

Environmental microbiology, 16(8), 2421-2432.

Ju, F., & Zhang, T. (2015). Bacterial assembly and temporal dynamics in activated sludge of a

full-scale municipal wastewater treatment plant. The ISME journal, 9(3), 683-695.

Karner, M., & Herndl, G. J. (1992). Extracellular enzymatic activity and secondary

production in free-living and marine-snow-associated bacteria. Marine biology, 113(2),

341-347.

Kembel, S. W., Cowan, P. D., Helmus, M. R., Cornwell, W. K., Morlon, H., Ackerly, D. D., ...

& Webb, C. O. (2010). Picante: R tools for integrating phylogenies and ecology.

Bioinformatics, 26(11), 1463-1464.

Konopka, A., Lindemann, S., & Fredrickson, J. (2015). Dynamics in microbial communities:

unraveling mechanisms to identify principles. The ISME journal, 9(7), 1488-1495.

Kraft, N. J., Adler, P. B., Godoy, O., James, E. C., Fuller, S., & Levine, J. M. (2015).

Community assembly, coexistence and the environmental filtering metaphor. Functional

ecology, 29(5), 592-599.

Lepp, P. W., Brinig, M. M., Ouverney, C. C., Palm, K., Armitage, G. C., & Relman, D. A.

(2004). Methanogenic Archaea and human periodontal disease. Proceedings of the

54 national academy of sciences, 101(16), 6176-6181.

Liu, L., Chen, H., Liu, M., Yang, J. R., Xiao, P., Wilkinson, D. M., & Yang, J. (2019).

Response of the eukaryotic plankton community to the cyanobacterial biomass cycle

over 6 years in two subtropical reservoirs. The ISME journal, 13(9), 2196-2208.

Liu, L., Yang, J., Lv, H., & Yu, Z. (2014). Synchronous dynamics and correlations between

bacteria and phytoplankton in a subtropical drinking water reservoir. FEMS

microbiology ecology, 90(1), 126-138.

Liu, M., Liu, L., Chen, H., Yu, Z., Yang, J. R., Xue, Y., ... & Yang, J. (2019). Community

dynamics of free-living and particle-attached bacteria following a reservoir Microcystis

bloom. Science of the Total Environment, 660, 501-511.

Ma, B., Wang, H., Dsouza, M., Lou, J., He, Y., Dai, Z., ... & Gilbert, J. A. (2016). Geographic

patterns of co-occurrence network topological features for at continental

scale in eastern China. The ISME journal, 10(8), 1891-1901.

Middelboe, M., Søndergaard, M., Letarte, Y., & Borch, N. H. (1995). Attached and free-

living bacteria: production and polymer hydrolysis during a diatom bloom. Microbial

Ecology, 29(3), 231-248.

Mou, X., Vila‐Costa, M., Sun, S., Zhao, W., Sharma, S., & Moran, M. A. (2011).

Metatranscriptomic signature of exogenous polyamine utilization by coastal

bacterioplankton. Environmental microbiology reports, 3(6), 798-806.

Moustaka-Gouni, M., Vardaka, E., Michaloudi, E., Kormas, K. A., Tryfon, E., Mihalatou,

H., ... & Lanaras, T. (2006). Plankton web structure in a eutrophic polymictic lake

55 with a history of toxic cyanobacterial blooms. Limnology and Oceanography, 51(1part2),

715-727.

Muyzer, G., De Waal, E. C., & Uitterlinden, A. G. (1993). Profiling of complex microbial

populations by denaturing gradient gel electrophoresis analysis of polymerase chain

reaction-amplified genes coding for 16S rRNA. Applied and environmental

microbiology, 59(3), 695-700.

O'Dwyer, J. P., Kembel, S. W., & Green, J. L. (2012). Phylogenetic diversity theory sheds

light on the structure of microbial communities. PLoS Comput Biol, 8(12), e1002832.

Ofiţeru, I. D., Lunn, M., Curtis, T. P., Wells, G. F., Criddle, C. S., Francis, C. A., & Sloan, W.

T. (2010). Combined niche and neutral effects in a microbial wastewater treatment

community. Proceedings of the National Academy of Sciences, 107(35), 15345-15350.

Oksanen, J. (2011). Multivariate analysis of ecological communities in R: vegan tutorial. R

package version, 1(7), 1-43.

Östman, Ö., Drakare, S., Kritzberg, E. S., Langenheder, S., Logue, J. B., & Lindström, E. S.

(2010). Regional invariance among microbial communities. Ecology letters, 13(1), 118-

127.

Paerl, H. W. (1988). Nuisance phytoplankton blooms in coastal, estuarine, and inland waters

1. Limnology and Oceanography, 33(4part2), 823-843.

Paerl, H. W., & Fulton, R. S. (2006). Ecology of harmful cyanobacteria. In Ecology of

harmful algae (pp. 95-109). Springer, Berlin, Heidelberg.

Paerl, H. W., & Otten, T. G. (2013). Harmful cyanobacterial blooms: causes, consequences,

56 and controls. Microbial ecology, 65(4), 995-1010.

Quast, C., Pruesse, E., Yilmaz, P., Gerken, J., Schweer, T., Yarza, P., ... & Glöckner, F. O.

(2012). The SILVA ribosomal RNA gene database project: improved data processing and

web-based tools. Nucleic acids research, 41(D1), D590-D596.

Reshef, D. N., Reshef, Y. A., Finucane, H. K., Grossman, S. R., McVean, G., Turnbaugh, P.

J., ... & Sabeti, P. C. (2011). Detecting novel associations in large data sets. science,

334(6062), 1518-1524.

Sloan, W. T., Lunn, M., Woodcock, S., Head, I. M., Nee, S., & Curtis, T. P. (2006).

Quantifying the roles of immigration and chance in shaping community

structure. Environmental microbiology, 8(4), 732-740.

Stegen, J. C., Lin, X., Konopka, A. E., & Fredrickson, J. K. (2012). Stochastic and

deterministic assembly processes in subsurface microbial communities. The ISME

journal, 6(9), 1653-1664.

Tremaroli, V., & Bäckhed, F. (2012). Functional interactions between the and

host metabolism. Nature, 489(7415), 242-249.

Venkataraman, A., Bassis, C. M., Beck, J. M., Young, V. B., Curtis, J. L., Huffnagle, G. B., &

Schmidt, T. M. (2015). Application of a neutral community model to assess structuring of

the human lung microbiome. MBio, 6(1).

Wagner, M., Loy, A., Nogueira, R., Purkhold, U., Lee, N., & Daims, H. (2002). Microbial

community composition and function in wastewater treatment plants. Antonie Van

Leeuwenhoek, 81(1), 665-680.

57 Wang, K., Zou, L., Lu, X., & Mou, X. (2018). Organic carbon source and salinity shape

sediment bacterial composition in two China marginal and their major tributaries.

Science of the Total Environment, 633, 1510-1517.

Witty, L.M., 2004. Practical guide to identifying freshwater crustacean zooplankton.

Woodcock, S., Van Der Gast, C. J., Bell, T., Lunn, M., Curtis, T. P., Head, I. M., & Sloan, W.

T. (2007). Neutral assembly of bacterial communities. FEMS microbiology ecology,

62(2), 171-180.

Woodhouse, J. N., Kinsela, A. S., Collins, R. N., Bowling, L. C., Honeyman, G. L., Holliday,

J. K., & Neilan, B. A. (2016). Microbial communities reflect temporal changes in

cyanobacterial composition in a shallow ephemeral freshwater lake. The ISME journal,

10(6), 1337-1351.

Worden, A. Z., Follows, M. J., Giovannoni, S. J., Wilken, S., Zimmerman, A. E., & Keeling,

P. J. (2015). Rethinking the marine carbon cycle: factoring in the multifarious lifestyles

of microbes. Science, 347(6223).

Xing, P., Zheng, J., Li, H., & Liu, Q. (2013). genotypes involved in methane

formation during anaerobic decomposition of Microcystis blooms at different

temperatures. World Journal of Microbiology and Biotechnology, 29(2), 373-377.

Xue, Y., Chen, H., Yang, J. R., Liu, M., Huang, B., & Yang, J. (2018). Distinct patterns and

processes of abundant and rare eukaryotic plankton communities following a reservoir

cyanobacterial bloom. The ISME journal, 12(9), 2263-2277.

Zhao, D., Shen, F., Zeng, J., Huang, R., Yu, Z., & Wu, Q. L. (2016). Network analysis reveals

58 seasonal variation of co-occurrence correlations between cyanobacteria and other bacterioplankton. Science of the Total Environment, 573, 817-825.

59 25 a Before During After b 20 SporichthyaceaChroococcuse;uncultured 4 SediminibacterAphanocapsaium biovolume

/L) 15 Candidatus planktolunaPlanktolyngbya 3 VerrucomicrobiaceaCylindrospee;unculturedrmopsis 10 Pseudarcicella 2 (mm Pseudanabaena 5 Opitutae vadinHA64;uncultured d Limnohabitans 0 FlavobacteriumChlamydomonas 0 Cyanobacteria hgcI clade Plagioselmis 20 SRA11 clade;unculturedCyclotella c Ramlibacter Carteria −2 /L) 3 15 Pseudomonas Euglena AquabacteriumCryptomonas CL500.29 marineAulacosei groupra −4 10 Brevundimonas 5 f Daphnia

Planktonic algae Planktonic Eubosmina biovolume (mm biovolume 0 Microcyclops Diaphanosoma 120 e Brachionus Conochilus 100 Tylotrocha cell 80 h ;uncultured 4 60 Sediminibacterium 40 Candidatus planktoluna Verrucomicrobiaceae;uncultured

counts counts (cells/L) 20 2

Zooplankton Pseudarcicella 0 Opitutae vadinHA64;uncultured Limnohabitans 0 12 g hgcI clade ) 6 11 SRA11 clade;uncultured 10

counts Ramlibacter 10 −2 ☓ Pseudomonas

cell 9 Aquabacterium CL500.29 marine group −4 8 Brevundimonas

(Cells/mL 7

Bacterial May Jun Jul Aug Sep Oct 6

May Jun Jun Jul Jul Aug Aug Sep Oct 07 22 17 25 23 08 28 25 17

Figure 1 Average abundances of estimated microbial communities. (a) estimated cyanobacterial cell biovolume, the x-axis showed the sampling events, (b) individual cyanobacterial cell abundance, (c) planktonic algae biovolume, (d) individual planktonic algae cell abundance at the genus level, (e) total zooplankton cell abundance, (f) individual zooplankton cell abundance at the family level, (g) total bacterioplankton cell abundance, and

(h) relative abundance of bacterioplankton (16S) at the genus level across the sampling period as measured at three sites.

60 1 1.0 (a) Cyanobacteria R² = 0.200, P < 0.05 0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2 (b) Planktonic algae R² = 0.051, P > 0.05

distance 0 0.0 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 1.0 1 (c) Zooplankton (d) Bacterioplankton R² = 0.604, P < 0.01 R² = 0.116, P < 0.05 0.8 0.8 Dissimilarity 0.6 0.6

0.4 0.4

0.2 0.2

0.0 0 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14

Time lag (square root of day)

Figure 2 Time-lag regression analysis of microbial communities. (a) cyanobacterial cell abundance, (b) planktonic algae cell abundance, (c) zooplankton cell abundance, and (d) bacterioplankton communities (16S).

61 1.00 1.00 a) All samples b) Before bloom R2 = 0.714 R2 = 0.549

0.75 0.75

0.50 0.50

0.25 Above prediction 0.25 Blow prediction

0.00 0.00

1.00 c) During bloom 1.00 d) After bloom R2 = 0.322 R2 = 0.535

Occurrence frequency 0.75 0.75

0.50 0.50

0.25 0.25

0.00 0.00 -14 -12 -10 -8 -6 -4 -2 -14 -12 -10 -8 -6 -4 -2 Log (mean relative abundance)

Figure 3 Fit of the neutral model. The predicted occurrence frequencies for bacterioplankton communities (OTUs) representing (a) all samples, (b) before bloom samples, (c) during bloom samples, and (d) after bloom samples, respectively. OTUs that occur more frequently than predicted by the model are shown in green, OTUs that occur less frequently than predicted are shown in orange, and OTUs that occur within the prediction are shown in black.

Dashed lines represent 95% confidence intervals around the model prediction (blue line).

62 a c 0.4 1000 BacteroidesBacteroides 1000 0.3 LachnospiLachnospirara OM27cladeOM27clade 800800 0.2 LegionellaLegionella uncultureduncultured Ric Rickettsialeskettsiales bacte bacteriumrium 600600

Migration rate Migration 0.1 uncultureduncultured Betaproteobacte Betaproteobacteriaria HyphomicrobiumHyphomicrobium 0.0 uncultureduncultured Caulobacter Caulobacter 400400 May Jun Jul Aug Sep Oct PhaselicystisPhaselicystis b FlFlavaobactevobacteriumrium 200200 1.0 Stress: 0.02 uncultureduncultured Fl Flavaobactevobacteriumrium uncultureduncultured Altere Altererythrobacterrythrobacter 00 PseudomonasPseudomonas

0.5 uncultureduncultured Stenotrophomonas Stenotrophomonas [Eubacte[Eubacterium]rium] coprostanoligenes coprostanoligenes group group [Ruminococcus][Ruminococcus] torques torques group group OpitutusOpitutus 0.0 SphaerotilusSphaerotilus NMDS2 uncultureduncultured Ambiguoustaxa Ambiguoustaxa AeromonasAeromonas

0.5 Methylomonas - Methylomonas Above prediction SaprospiSaprospirara Within prediction LautropiaLautropia

1.0 Below prediction ZymomonasZymomonas - Above prediction Below prediction -1.0 -0.5 0.0 0.5 1.0 BeBe DuDu AfterAfter BeBe DuDu AfterAfter forefore ringring forefore ringring NMDS1

Figure 4 (a) The estimated migration rate for bacterioplankton communities (OTUs), (b)

Neutral and non-neutral partitions of the bacterioplankton communities are compositionally and phylogenetically distinct. Non-metric multidimensional scaling ordination based on

UniFrac distances, (c) List of non-neutral bacterioplankton genera and had a relative abundance of > 1% of the total sequence (color gradient represents the number of sequences).

63 a b

1.0 Bacterioplankton Temp Bacterioplankton All samples 1.0 TOC Non-bloom samples Chl a HCO3 0.5 0.5 DON EC (8.3%) 0.0 DON EC 0.0 (13.1%) RDA2 0.5 - HCO Chl a 3 RDA2 0.5 - TOC

1.0 Temp - 1.0 - -2.0 -1.0 0.0 1.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 RDA1 (27.1%) RDA1 (17.1%)

May June July August September October - 3 + 4 x a NH Temp NO FTU DON DOC SRP

c TP pH EC HCO DO TN Chl 1.0 TOC

TN 0.8

TP 0.6 DON

+ 0.4 NH4 Bacterioplankton NOx 0.2 DOC 0 Temp Cyanobacteria pH -0.2

SRP -0.4 P values EC

0.01 - 0.05 - -0.6 HCO3 <0.01 Mantel’s r Planktonic algae DO -0.8 0.25 - 0.5 FTU > 0.5 Zooplankton -1.0 Chl a Figure 5 Distance-based redundancy visualizing of the relationship between the measured environmental parameters and bacterioplankton community composition (16S rRNA) based on (a) all samples and (b) only non-bloom samples, (c) Heatmap of Pearson’s correlation coefficient values (color gradient) calculated based on correlation matrix of environmental variables. Curved lines denote correlations between bacterioplankton communities, cyanobacterial, planktonic algae and zooplankton cell abundances and individual environmental variables that were revealed by partial mantel tests. Width of edges corresponds to the values of Mantel’s r. Edge color denotes the statistical significance (P values) based on 999 permutations. Only edges with P values < 0.05 were shown.

64 During During (a) 100 100 (b) Module 1 (c) Module 2

80 20 80 20

60 40 60 40

40 60 40 60

20 80 20 80

100 100 Before 100 80 60 40 20 After Before 100 80 60 40 20 After

During During 100 100 (d) Module 3 (e) Module 4 80 20 80 20

60 40 60 40

Module 1 (19.8%) Module 2 (19.5%) Module 3 (16.7%) 40 60 40 60 Module 4 (15.4%) Module 5 (12.7%) Module 6 (6.6%)

Others (9.3%) 20 80 20 80

100 100 100 80 60 40 20 Before After Before 100 80 60 40 20 After

Ternary plots legend During During 100 100 Taxa Degree (f) Module 5 (g) Module 6 20 20 Actinobacteria Alphaproteobacteria 25 80 80 50 75 Bacteroidetes Betaproteobacteria 100 60 40 60 40 Firmicutes

Verrucomicrobia Gammaproteobacteria 40 60 40 60 Cyanobacteria/Algae Others 20 80 20 80 Zooplankton

100 100 100 80 60 40 20 100 80 60 40 20 Before AfterBefore After

Figure 6 (a) The co-occurrence patterns among bacterioplankton communities (OTUs), cyanobacterial, planktonic algae, and zooplankton cell abundances revealed by network analysis. The nodes were colored according to different types of modularity classes. The size of each node is proportional to the number of connections (i.e., degree), (b-g) Ternary plots showing relative abundance from modules 1-6 in the three different periods. Each circle represents one individual OTU. For each OTU, abundance was averaged over all samples at each period.

65

Figure 7 Rarefaction curves of operational taxonomic unit (OTU) at 97% sequence similarity level.

66 a 6.4

6.2

6.0

index 5.8

5.6

Shannon 5.4

5.2 y a ul ul ug ug un un J J J J A b 505 A 7− 8− 17− 25− 17−Oct 22− 28− 25−Sep 23−M 500

495

490 Chao1 485

480

Mayy Jun Jun Jul Jul a ul ul Aug Aug Sep Oct ug ug un un J J J J A 17 25 A 23 07 22 08 28 25 17 7− 8− 17− 25− 17−Oct 22− 28− 25−Sep 23−M Figure 8 Temporal distribution pattern of bacterioplankton alpha diversity indexes.

67 1 1 Phylum Family

0.8 0.8 R² = 0.116, P < 0.05 R² = 0.050, P > 0.05

0.6 0.6

0.4 0.4

0.2 0.2

0 0 1 1 Class Genus

0.8 0.8 R² = 0.144, P < 0.05 R² = 0.120, P < 0.05 0.6 0.6 distance 0.4 0.4

0.2 0.2 Dissimilarity 0 0 0 2 4 6 8 10 12 14 1 Order

0.8 R² = 0.112, P < 0.05

0.6

0.4

0.2

0 0 2 4 6 8 10 12 14 Time lag (square root of day)

Figure 9 Time-lag regression analysis of bacterioplankton communities (16S) at (a) phylum,

(b) class, (c) order, (d) family, and (e) genus levels.

68 (a) Cyanobacteria

1.0 HCO3 1.0 DON EC

0.5 DON SRP 0.5 Temp (7.3%) (17.0%) 0.0

NOx 0.0 NOx 0.5

pH RDA2 RDA2 - +

0.5 NH4 -

1.0 Temp

- EC

1.0 HCO3 - (b) Planktonic algae 1.5 - -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -2.0 -1.0 0.0 1.0 RDA1 (27.0%) RDA1 (14.3%)

(c Zooplankton 1.0 0.5 CCcyano CCbact EC May June July (19.6%) 0.0

CCalgae RDA2 August September October 0.5 - 1.0 - Temp

-1.5 -1.0 -0.5 0.0 0.5 1.0 RDA1 (33.2%)

Figure 10 Distance-based linear redundancy visualizing the relative contribution of measured environmental parameters on (a) cyanobacterial, (b) planktonic algae, and (c) zooplankton communities.

69 (c) Non-neutral OTUs (d) Neutral OTUs May Temp TOC HCO3 0.5 Jun 0.5 DON DON pH EC Jul 0.0 (20.2%)

NOx (11.5%) NOx

0.0 EC pH Aug 0.5 RDA2 RDA2 - HCO3 Sep TOC Temp 0.5 - 1.0

- Oct

-1.0 -0.5 0.0 0.5 1.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 RDA1 (43.7%) RDA1 (23.0%) Figure 11 Distance-based linear redundancy visualizing the relative contribution of measured environmental parameters on (a) non-neutral OTUs, and (b) neutral OTUs.

70

Figure 12 The distributions of degree for the microbial community co-occurrence network

(green) and Erdös-Rényi random network (orange).

71

Figure 13 Zi-Pi plot showing the distribution of bacterioplankton OTUs based on their topological roles. Each symbol represents an OTU. The topological role of each OTU was determined according to the scatter plot of within-module connectivity (Zi) and among- module connectivity (Pi). Orange: module hubs; Green: connectors.

72 Zooplankton (1.57%) Planktonic algae (1%) Abiotic factors

Temp Cyanobacteria (18.52%) + NH4 Alphaproteobacteria (13.53%)

Chl a Betaproteobacteria (13.53%)

Deltaproteobacteria (10.83%)

NOx Bacteroidetes (10.68%)

Verrucomicrobia (8.4%)

C/N Gammaproteobacteria (5.84%) TN HCO EC TOC 3 Actinobacteria (5.7%)

Firmicutes (1.71%)

Others (7.43%)

Figure 14 Network visualizing significant correlations between measured biotic and abiotic variables. The size of each node is proportional to the number of connections (i.e., degree).

(To reduce complexity, correlations just between bacterioplankton communities were removed).

73 Table 1 List of indicator species found before, during and after the bloom periods.

r P Family Genus Species Before bloom (n = 39) 1042 0.876 0.038 Oxalobacteraceae Duganella Ambiguous_taxa 1083 0.866 0.012 Saprospiraceae NA NA 1180 0.821 0.008 Sporichthyaceae hgcIclade Ambiguous_taxa 1232 0.868 0.009 uncultured_bacterium 1238 0.827 0.035 Oligoflexaceae NA NA 1246 0.8 0.047 Hyphomicrobiaceae Devosia Ambiguous_taxa 174 0.972 0.003 Caulobacteraceae Brevundimonas NA 1755 0.821 0.008 Flavobacteriaceae Chryseobacterium NA 2039 0.884 0.01 Oxalobacteraceae Duganella NA 2084 0.812 0.028 Oceanospirillaceae Pseudohongiella 214 0.904 0.01 0319-6G20 uncultured_bacterium uncultured_bacterium 2151 0.915 0.001 Verrucomicrobiaceae uncultured_ NA 2271 0.88 0.008 uncultured_bacterium uncultured_bacterium uncultured_bacterium 233 0.919 0.019 Verrucomicrobiaceae Luteolibacter uncultured_bacterium 2423 0.907 0.002 Chitinophagaceae Dinghuibacter uncultured_Sphingobacteriumsp. 250 0.941 0.01 Fibrobacteraceae uncultured_ uncultured_bacterium 29 0.952 0.002 Verrucomicrobiaceae uncultured_ Ambiguous_taxa 302 0.947 0.025 Verrucomicrobiaceae Prosthecobacter uncultured_bacterium 309 0.959 0.005 uncultured_ uncultured_Thiobacillussp. 345 0.934 0.002 Chitinophagaceae Lacibacter Ambiguous_taxa 406 0.875 0.048 Limnobacter Ambiguous_taxa 407 0.911 0.018 Verrucomicrobiaceae Haloferula uncultured_bacterium

74 414 0.919 0.05 Chitinophagaceae uncultured_ uncultured_bacterium 564 0.857 0.013 PHOS-HE51 uncultured_Sphingobacterialesbacterium uncultured_Sphingobacterialesbacterium 581 0.84 0.026 Chitinophagaceae uncultured_bacterium uncultured_bacterium 604 0.896 0.026 Chthoniobacteraceae Chthoniobacter uncultured_bacterium 626 0.898 0.008 Cytophagaceae Emticicia Ambiguous_taxa 629 0.871 0.015 Verrucomicrobiaceae Luteolibacter uncultured_bacterium 630 0.89 0.011 Sphingobacteriaceae Mucilaginibacter uncultured_bacterium 671 0.854 0.019 Sphingobacteriaceae NA 681 0.837 0.02 Roseiflexaceae Roseiflexus uncultured_bacterium 722 0.886 0.032 Oxalobacteraceae Duganella Ambiguous_taxa 773 0.872 0.016 Flavobacteriaceae Flavobacterium Ambiguous_taxa 785 0.869 0.028 NS11-12marinegroup uncultured_bacterium uncultured_bacterium 837 0.836 0.025 Opitutaceae Opitutus uncultured_bacterium 845 0.837 0.015 Bdellovibrionaceae Bdellovibrio uncultured_bacterium 856 0.912 0.009 Verrucomicrobiaceae uncultured_ Ambiguous_taxa 911 0.891 0.006 Sphingobacteriaceae Pedobacter Ambiguous_taxa 923 0.837 0.004 uncultured_ uncultured_deltaproteobacterium uncultured_deltaproteobacterium During bloom (n = 56) 1012 0.837 0.004 LWSR-14 uncultured_bacterium uncultured_bacterium 1016 0.842 0.001 Planococcaceae Planomicrobium uncultured_bacterium 1026 0.807 0.011 Methylococcaceae Methyloparacoccus Ambiguous_taxa 1047 0.804 0.004 NA NA NA 1080 0.816 0.016 Propionibacteriaceae uncultured_ uncultured_Actinomycetalesbacterium 1086 0.836 0.007 NA NA NA 1099 0.833 0.007 NA NA NA 1143 0.904 0.004 NA NA NA

75 1208 0.856 0.003 Rickettsiaceae uncultured_ uncultured_alphaproteobacterium 1217 0.865 0.005 RickettsialesIncertaeSedis uncultured_ uncultured_bacterium 1227 0.864 0.001 Legionellaceae Legionella uncultured_bacterium 1239 0.863 0.002 Legionellaceae Legionella uncultured_bacterium 1254 0.816 0.002 uncultured_proteobacterium uncultured_proteobacterium uncultured_proteobacterium 1259 0.845 0.04 pLW-20 uncultured_soilbacterium uncultured_soilbacterium 1399 0.816 0.005 NS9marinegroup uncultured_bacterium uncultured_bacterium 1437 0.809 0.009 NA NA NA 1456 0.805 0.004 Oligoflexaceae Oligoflexus NA 1475 0.816 0.002 Bdellovibrionaceae OM27clade uncultured_bacterium 1543 0.913 0.001 NA NA NA 1717 0.804 0.003 NA NA NA 1725 0.913 0.001 Nitrosomonadaceae uncultured_ NA 1868 0.902 0.001 uncultured_ wastewatermetagenome wastewatermetagenome 2221 0.816 0.001 uncultured_ uncultured_Bacteroidetesbacterium uncultured_Bacteroidetesbacterium 2381 0.825 0.005 Spongiibacteraceae BD1-7clade uncultured_bacterium 239 0.948 0.008 NA NA NA 292 0.894 0.028 Flavobacteriaceae Flavobacterium Flavobacteriumsp.JJ011 2999 0.953 0.001 RhodospirillalesIncertaeSedis Reyranella Ambiguous_taxa 3015 0.916 0.009 Comamonadaceae NA NA 3027 0.812 0.024 Comamonadaceae Ideonella NA 303 0.959 0.002 Neisseriaceae uncultured_ uncultured_betaproteobacterium 434 0.874 0.034 Oxalobacteraceae NA NA 466 0.833 0.028 P3OB-42 uncultured_deltaproteobacterium uncultured_deltaproteobacterium 469 0.892 0.01 Bdellovibrionaceae OM27clade uncultured_bacterium 488 0.861 0.025 Legionellaceae Legionella uncultured_bacterium

76 536 0.927 0.004 0319-6G20 NA NA 580 0.875 0.004 NA NA NA 589 0.903 0.012 Planctomycetaceae uncultured_ uncultured_bacterium 591 0.924 0.003 0319-6G20 uncultured_bacterium uncultured_bacterium 656 0.91 0.011 Bdellovibrionaceae Bdellovibrio uncultured_bacterium 665 0.861 0.006 uncultured_soilbacterium uncultured_soilbacterium uncultured_soilbacterium 721 0.964 0.001 Bdellovibrionaceae OM27clade uncultured_bacterium 725 0.859 0.019 NA NA NA 753 0.84 0.017 Neisseriaceae uncultured_ uncultured_betaproteobacterium 826 0.861 0.018 Moraxellaceae Fluviicoccus uncultured_bacterium 831 0.951 0.019 Oligoflexaceae Oligoflexus uncultured_bacterium 842 0.874 0.008 UnknownFamily CandidatusMethylacidiphilum NA 847 0.879 0.02 Nitrosomonadaceae uncultured_ NA 857 0.942 0.001 NA NA NA 887 0.836 0.004 Simkaniaceae CandidatusRenichlamydia uncultured_CandidatusRhabdochlamydiasp. 891 0.827 0.01 RickettsialesIncertaeSedis CandidatusCaptivus uncultured_bacterium 910 0.848 0.005 NA NA 957 0.847 0.002 uncultured_bacterium uncultured_bacterium uncultured_bacterium 970 0.916 0.002 NA NA NA 972 0.848 0.001 NA NA NA 975 0.845 0.039 Oligoflexaceae uncultured_bacterium uncultured_bacterium 982 0.872 0.015 Burkholderiaceae Limnobacter uncultured_bacterium After bloom (n = 21) 1183 0.861 0.002 NA NA NA 1229 0.921 0.001 Alcaligenaceae Sutterella uncultured_organism 1230 0.869 0.008 FamilyI NA NA

77 1362 0.811 0.004 Neisseriaceae uncultured_ uncultured_betaproteobacterium 1534 0.801 0.006 FamilyI NA 1609 0.911 0.003 Sphingobacteriaceae uncultured_bacterium 1743 0.921 0.003 Caulobacteraceae Brevundimonas NA 1845 0.816 0.002 uncultured_bacterium uncultured_bacterium uncultured_bacterium 2286 0.915 0.029 Sphingomonadaceae Blastomonas Ambiguous_taxa 268 0.976 0.001 Oligoflexaceae bacteriumGLA1 bacteriumGLA1 3195 0.871 0.009 Flavobacteriaceae Flavobacterium Ambiguous_taxa 396 0.973 0.002 uncultured_bacterium uncultured_bacterium uncultured_bacterium 534 0.985 0.002 Mitochondria NA NA 541 0.963 0.001 NA NA NA 554 0.903 0.003 Chitinophagaceae uncultured_ uncultured_bacterium 575 0.906 0.002 FamilyI Ambiguous_taxa 643 0.956 0.001 LWSR-14 uncultured_bacterium uncultured_bacterium 651 0.867 0.006 Moraxellaceae Acinetobacter Acinetobactercalcoaceticus 702 0.898 0.002 uncultured_bacterium uncultured_bacterium uncultured_bacterium 817 0.895 0.002 Verrucomicrobiaceae Akkermansia uncultured_bacterium 853 0.918 0.001 CL500-29marinegroup uncultured_actinobacterium

78 Table 2 Topological properties of the entire co-occurrence networks of microbial communities and their associated random networks.

Real networks Random network Average Average Average Average Small- Network Average Nodesa Edgesb Modularityc clustering path Modularityc clustering path word diametere degreeg coefficientd lengthf coefficientd lengthf coefficienth Whole 1313 14040 0.589 0.414 11 3.811 22.535 0.1748 0.015 2.675 26.9 aNumber of OTUs with the correlation r > 0.7 or r < -0.7 and statistical significance (P-value < 0.01); bNumber of strong and significant correlations between nodes; cModularity > 0.4 suggests that the network has module structure characteristics; dHow nodes are embedded in their neighborhood, and the degree to which nodes tend to cluster together; eThe maximum distance between all possible pairs of nodes; fThe average number of steps along the shortest paths for all possible pairs of network nodes; gNode connectivity showing how many connections (on average) each node has to the other nodes in the network; hSmall-word coefficient > 1 indicates “small-world” properties, that is, high interconnectivity and high efficiency.

79 Table 3 Lists of module hubs and connectors in co-occurrence network.

OTU p_score z_score type NCM prediction Module 364 0.734 0.818 connector above 1 1165 0.660 -0.064 connector above 1 319 0.667 -0.068 connector above 3 2212 0.024 3.019 module_hub above 2 831 0.667 -0.866 connector below 4 466 0.087 2.977 module_hub below 4 123 0.048 2.787 module_hub below 4 366 0.046 2.882 module_hub below 4 665 0.042 3.357 module_hub below 4 255 0.041 3.452 module_hub below 4 580 0.036 4.022 module_hub below 4 488 0.000 3.642 module_hub below 4 431 0.722 -0.847 connector within 1 113 0.720 -0.954 connector within 1 952 0.694 0.818 connector within 1 2552 0.659 1.646 connector within 1 780 0.656 -0.954 connector within 1 444 0.653 -0.940 connector within 1 888 0.711 0.519 connector within 2 2130 0.665 -0.834 connector within 2 63 0.662 -0.254 connector within 3 385 0.720 -0.954 connector within 3

80 31 0.663 -0.940 connector within 3 146 0.742 1.038 connector within 4 160 0.667 0.730 connector within 4 2730 0.190 2.766 module_hub within 1 88 0.111 2.918 module_hub within 1 23 0.065 3.220 module_hub within 2 125 0.049 2.817 module_hub within 2 37 0.045 3.119 module_hub within 2 148 0.042 3.473 module_hub within 2 56 0.000 2.716 module_hub within 2 753 0.121 3.167 module_hub within 4 481 0.091 2.787 module_hub within 4 1239 0.087 2.977 module_hub within 4 27 0.081 3.262 module_hub within 4 151 0.046 2.882 module_hub within 4 814 0.045 2.977 module_hub within 4 701 0.000 2.787 module_hub within 4 106 0.000 3.072 module_hub within 4

81 Table 4 Lists of keystone species in co-occurrence network.

OTU Degree Betweeness centrality NCM prediction Family Genus 186 113 2202 within NS11-12marinegroup uncultured 709 100 3493 below Rhizobiaceae Rhizobium 547 118 4802 below Acetobacteraceae Roseococcus 2039 123 4608 below Oxalobacteraceae Duganella 722 110 3059 below Oxalobacteraceae Duganella 214 102 2649 below Opitutaceae Opitutus 1837 108 3415 below Verrucomicrobiaceae Brevifollis 407 118 3117 below Verrucomicrobiaceae Haloferula 629 113 4588 below Verrucomicrobiaceae Luteolibacter 29 105 2106 below Verrucomicrobiaceae uncultured 848 100 3780 below Verrucomicrobiaceae uncultured

82 CHAPTER III

CO-OCCURRING MICROORGANISMS REGULATE THE SUCCESSION OF

CYANOBACTERIAL HARMFUL ALGAL BLOOMS

PREFACE

This chapter was submitted to the Environmental Pollution journal under the title “Co- occurring microorganisms regulate the succession of cyanobacterial harmful algal blooms”.

The author list is as follows: Kai Wang, Huansheng Cao, Ian Struewing, Joel Allen, Jingrang

Lu, Xiaozhen Mou. Kai Wang did all of the data analysis and wrote this manuscript. Huansheng

Cao helped analyze RNA-seq data. Ian Struewing and Joel Allen helped measure nutrient- related variables. Jingrang Lu helped collect samples and sequencing library preparation.

Xiaozhen Mou helped me with data analysis and writing of this manuscript.

ABSTRACT

CyanoHABs have been found to transmit from N2 fixer-dominated to non-N2 fixer- dominated in many freshwater environments when the supply of N decreases. To elucidate the mechanisms underlying such “counter-intuitive” CyanoHAB species succession, metatranscriptomes (biotic data) and water quality-related variables (abiotic data) were analyzed weekly during a bloom season in Harsha Lake, a multipurpose lake that serves as a

83 drinking water source and recreational ground. Our results showed that CyanoHABs in Harsh

Lake started with N2-fixing Anabaena in June (ANA stage) when N is high and transitioned to

non-N2-fixing Microcystis- and Planktothrix-dominated in July (MIC-PLA stage) when N became limited (low TN/TP). Meanwhile, the concentrations of cyanotoxins (i.e., microcystins) were significantly higher in the MIC-PLA stage. Water quality results revealed that N species

(i.e., TN, TN/TP) and water temperature were significantly correlated with cyanobacterial biomass. Expression levels of several C- and N-processing-related cyanobacterial genes were highly predictive of the biomass of their species. More importantly, the biomasses of Microcystis and Planktothrix were also significantly associated with expressions of microbial genes (mostly from heterotrophic bacteria) related to processing organic substrates

(alkaline phosphatase, peptidase, -active enzymes) and cyanophage genes.

Collectively, our results suggest that besides environmental conditions and inherent traits of specific cyanobacterial species, the development and succession of CyanoHABs are regulated by co-occurring microorganisms. Specifically, the co-occurring microorganisms can alleviate the nutrient limitation of cyanobacteria by remineralizing organic compounds.

INTRODUCTION

CyanoHABs are occurring with increasing frequencies and impacted areas globally as a result of eutrophication and global warming (Huisman et al., 2018). Recognized as important environmental hazards, CyanoHABs deteriorate the water quality of various aquatic environments by increasing water turbidity (Scheffer et al., 1993), depleting dissolved oxygen

84 (Rabalais et al., 2010), and producing a variety of toxic secondary metabolites (Carmichael,

2001). Many cyanobacterial genera can cause CyanoHABs in freshwater ecosystems; these

include N2-fixing (diazotrophic) taxa, such as Anabaena, Cylindrospermopsis, and Nodularia

(Schindler et al., 2008), and non-N2-fixing taxa, such as Microcystis, Planktothrix, and

Oscillatoria (Paerl and Otten, 2013).

Nitrogen availability is a known regulator for CyanoHAB community structure (Lu et

al., 2019). It is generally assumed that N limited condition favors the growth of N2-fixing

cyanobacteria, whereas, when N supply is ample, non-N2-fixing cyanobacteria would

outcompete slow-growing N2-fixing taxa (Paerl and Otten, 2016). However, many blooms have

been reported to start with N2-fixers when N was replete and then transited into non-N2-fixing genera under low N concentrations (Beversdorf et al., 2013; Lu et al., 2019). This “opposite

scenario” has been attributed to direct/indirect transfers of N from N2-fixers to non-N2-fixing cyanobacteria (Beversdorf et al., 2013; Lu et al., 2019).

Many recent studies have suggested the important role of microorganisms in regulating CyanoHABs (Brauer et al., 2015; Woodhouse et al., 2016; Wang et al., 2020).

Nutrient recycling by the co-occurring microorganisms (i.e., the microbiome of CyanoHABs) has been shown to supply a large proportion of nutrients to cyanobacteria when the allochthonous supply is low (Christie-Oleza et al., 2017). Moreover, microbiome communities have also been found to provide key limiting micronutrients (i.e., iron and vitamins) to cyanobacteria (Croft et al., 2005; Amin et al., 2015; Xie et al., 2016).

CyanoHABs also host a diversity of cyanophages; these “cyanobacteria grazers” are

85 often -specific (Gerphagnon et al., 2015) and can shift cyanobacterial community structures by selectively removing certain cyanobacterial taxa. On the other hand, cyanobacteria have evolved defensive mechanisms against cyanophage infection, such as the antivirus genes (Makarova et al., 2011; Rohrlack et al., 2013) and CRISPR-Cas systems (Kuno et al., 2014).

The succession of cyanobacterial taxa may be also due to species-specific responses of cyanobacteria to seasonal changes of temperature and light (Oberhaus et al., 2007; Paerl et al.,

2011). For example, warm temperatures (28-32°C) tend to favor the growth of Microcystis (1.6 divisions day-1) over Anabaena (1.25 divisions day-1) (Nalewajko and Murphy, 2001).

Planktothrix can thrive over a broader temperature range and maintain a high growth rate at lower light intensity than the aforementioned cyanobacterial genera (Oberhaus et al., 2007).

Although the impacts of individual environmental and biotic factors on cyanobacterial growth are becoming clear (Huisman et al., 2018), their interplay in driving cyanobacterial species transitions during CyanoHABs remains largely unknown. Elucidating mechanisms governing CyanoHAB development and species succession is critical to understand the ecology of CyanoHABs, which knowledge is essential to guide wise management strategies that can help to prevent and/or mitigate CyanoHABs pollution. To address this knowledge gap, metatranscriptomes of cyanobacteria and their microbiome as well as in situ environmental variables were examined weekly over a four-month period (June to September) in Harsha Lake, a water reservoir in Ohio, USA that experienced CyanoHABs in the past decades. We hypothesized that the resource exchange between cyanobacteria and their associated

86 microbiome may favor the growth of certain cyanobacterial taxa, which in turn regulates cyanobacterial bloom development and succession.

MATERIALS AND METHODS

Sample collection and processing

Surface water samples (~0.5 m depth) were collected weekly from June to September

2015 (15 sampling dates, 28 samples) in Harsha Lake at the site of a drinking plant intake (39.037o, -84.138o) using a plastic water jug, which had been pre-washed with 5% hydrochloric acid. Water temperature (Temp), pH, dissolved oxygen (DO), electrical conductivity (EC), and turbidity were measured in situ using a multi-parameter sonde (YSI,

650 MDS, OH, USA).

Water samples were immediately placed in a cooler with icepacks and transported back to the laboratory. Within approximately 5 hours, 100-200 mL of water sample was filtered using a Durapore polyvinylidene fluoride filter (0.45 μm, MilliPore, Foster City, CA). Cells that were collected on 0.45 μm filters were stored in 1.5-mL Lysing Matrix A tubes (MP biomedicals, Irvine, CA, USA) that contained 600 μL RLT plus solution with RNase inhibitor

(QIAGEN, Chatsworth, CA, USA). Filters were prepared in two replicates for each sampling site and were frozen immediately in liquid nitrogen and stored at -80 °C until RNA extraction.

Filtrates were collected in sterile conical centrifuge tubes and frozen at -20 °C for nutrient analyses. The community structure of cyanobacteria was identified to genus level under a 400× magnification using a Nikon microscope (Nikon Corp., Japan) following taxonomic keys

87 (Thorp and Covich, 2009).

Nutrient analysis

Nutrients were measured according to standard procedures described in the Ohio

Environmental Protection Agency (Ohio EPA) methods

(https://www.epa.state.oh.us/ddagw/labcert) using the Lachat Autoanalyzer (Quickchem 8000,

- - Hach Co, USA). Briefly, nitrate (NO3 ) and nitrite (NO2 ), and soluble reactive phosphorus

(SRP) were measured based on the automated hydrazine reduction method (Ohio EPA Method

4500), and colorimetry method (USEPA method 365.1), respectively. Total organic carbon

(TOC) and total nitrogen (TN) were measured in unfiltered raw water samples following Ohio

EPA Method 335.2, and USEPA Method 351.2, respectively. Total phosphorus (TP) concentrations were determined in unfiltered raw water samples according to USEPA Method

365.4. Microcystins (MCs) were measured in whole water using the MC-ADDA Enzyme-

Linked Immunosorbent Assay (ELISA) kit (Abraxis, Warminster, PA) following the standard freeze-thaw procedures.

RNA extraction and sequencing

RNA was extracted from frozen filters following a procedure described previously (Lu et al., 2019). Briefly, filters were thawed and homogenized with a Mini-Beadbeater-16

(BioSpec Products, Inc., Bartlesville, OK) for 30 seconds twice, and then centrifuged under room temperature at 10,000 g for 3 min. The supernatant was transferred to a new sterile microcentrifuge, and RNA was extracted and purified using an AllPrep DNA/RNA kit

(QIAGEN, Chatsworth, CA, USA) according to the manufacturer’s protocol. The genomic

88 DNA contamination was removed using a TURBO DNA-free kit and according to the manufacturer’s protocol (Invitrogen, Carlsbad, CA). The RNA quality was assessed using an

Agilent 2100 bioanalyzer (Agilent Tech. Inc., Santa Clara, CA). Ribosomal RNA (rRNA) removal, reverse transcription, and library preparation were performed as described previously

(Lu et al., 2019) prior to 300 bp paired-end sequencing on the Illumina MiSeq system. Raw reads were deposited in the National Center for Biotechnology Information (NCBI) short read archive database under accession number SRR12458594-SRR12458621.

RNA-seq data processing

Low-quality reads (Phred score < 15, reads < 50 bp, and adaptor contaminants) were removed using Trimmomatic version 0.39 (Bolger et al., 2014). Ribosomal RNA (rRNA) reads were removed using SortMeRNA version 2.1 (Kopylova et al., 2012) and non-rRNA reads were used in downstream analyses. The taxonomic affiliation was assigned using the Kaiju software version 1.7.2 with default parameters (Menzel et al., 2016). A resampling procedure was performed before running the Kaiju software based on the sample that has the lowest number of sequences to ensure results were comparable between samples. To examine the transcriptomic patterns of cyanobacterial populations, non-rRNA reads were first non- redundantly mapped to three reference cyanobacterial genomes, including NIES 843 (acc. # NC_010296.1), NIVA CYA 126/8 (acc. #

CM002803.1), and Anabaena sp. 90 (acc. # GCA_000312705.1) using STAR software version

2.7.3 (Dobin et al., 2013). These three cyanobacterial species were the most abundant cyanobacteria identified in Harsha Lake (Lu et al., 2019).

89 Sequences that did not map to any of the three cyanobacterial genomes were subsequently mapped to six complete or draft reference genomes of cyanobacterial phages, including Microcystis phages Ma-LMM01 (acc. # GCA_000870225.1), MACPNOA1 (acc. #

GCA_004320245.1), MaMV-DC (acc. # GCA_001505175.1), and Ma_Me-ZS1 (acc. #

GCA_003865555.2), Anabaena phage A-4L (acc. # NC_024358.1), and Planktothrix phage

PaV-LD (acc. # NC_016564.1).

The remaining unmapped non-cyanobacterial reads from bacterioplankton, fungi, and microbial were assembled into contigs using Trinity software version 2.8.5 (Haas et al., 2013). The non-cyanobacterial reads were then mapped back to contigs for quantification using the Bowtie2 alignment tool (Langmead and Salzberg, 2012). Putative coding regions and corresponding encoded proteins of assembled contigs were identified and retrieved using

TransDecoder software (http://transdecoder.github.io/). Assembled contigs and predicted protein-coding sequences were aligned to the SwissProt database using local DIAMOND

BLASTX and BLASTP programs, respectively (Buchfink et al., 2015). The alignment results were further parsed by Trinotate for functional annotation and gene ontology (GO) term assignment. The taxonomic classification of contigs was determined using DIAMOND

(Buchfink et al., 2015) against the NCBI nr database. Transcripts were normalized by calculating transcript per million (TPM) using the RSEM program with default settings (Li et al., 2011).

Statistical analyses

Potential differences in environmental variables between different stages of

90 cyanobacterial blooms (biomass > 10 mg L-1) were examined by one-way ANOVA using the vegan package (Oksanen et al., 2011) in R. The transcript data (TPM values) were square-root- transformed, and a Bray-Curtis similarity matrix was constructed to perform clustering.

Principle component analysis (PCA) based on Bray-Curtis distance was used for visualization of gene expression patterns among samples. Differential gene expression analysis of cyanobacterial gene transcripts between the different clusters (based on PCA results) was performed using the DESeq2 package (Love et al., 2014). The false discovery rate (FDR) values were calculated to correct corresponding P values using the Benjamini-Hochberg algorithm (Benjamini and Hochberg, 1995). Gene transcripts with FDR < 0.05 and an absolute fold change ≥ 1.5 were considered as differentially expressed.

Weighted gene co-expression network analysis (WGCNA) was performed to detect modules (a set of co-expressed genes) that are associated with the variations of cyanobacterial biomass using the R package WGCNA (Langfelder and Horvath, 2008). WGCNA analyses were performed separately on matrices of gene transcripts of cyanobacteria (e.g., Anabaena,

Microcystis, Planktothrix), cyanophages, and other non-cyanobacterial microorganisms.

Expression matrices were all Hellinger-transformed before WGCNA analysis (Guidi et al.,

2006). In WGCNA networks, nodes correspond to gene transcripts, and edges are determined by Pearson pairwise correlations between gene transcripts. WGCNA analysis uses a soft thresholding (p) of the correlation matrix in order to show a scale-free topological network where key nodes are highly connected with others (Zhang et al., 2005). Briefly, pairwise

Pearson correlation coefficients between gene transcripts were calculated. A signed adjacency

91 matrix was calculated by raising the absolute value of these Pearson correlation coefficients to the power of p. Topological overlap measure (TOM) was then calculated based on the obtained adjacency matrix. Modules were identified by performing a hierarchical clustering using a distance matrix based on the TOM values. The first principal component was used to represent each module.

For each module, pairwise Pearson correlation coefficients between environmental variables (including cyanobacterial biomass) and identified principal components were calculated. The corresponding P values were corrected for multiple testing by calculating FDR values (Benjamini and Hochberg, 1995). Gene modules that showed significant correlations with variations of cyanobacterial biomass (Anabaena, Microcystis, or Planktothrix) were selected for downstream analyses.

For each selected module, the module membership is measured by calculating Pearson correlations between the relative abundance of gene transcripts and its first principal component. The module significance is measured by the Pearson correlations between the relative abundance of gene transcripts and cyanobacterial biomass. The results of module membership and module significance were presented as a scatter plot. A correlation between the module membership and module significance indicates the potential correlation of the module structure (topology) and the cyanobacterial biomass (i.e., the more a given gene transcript defines the module topology, the more it is correlated to cyanobacterial biomass).

The sparse partial least square (sPLS) analysis was further performed to assess whether the gene transcripts within the selected module could explain the biomass variation of a

92 cyanobacterial taxon using R package mixOmics (Rohart et al., 2017). The prediction power of the model was assessed by calculating correlation coefficients between the predicted and measured values. The significance of the prediction power was assessed by permuting the data

1,000 times with leave-one-out cross-validation (LOOCV). In addition, the score of value importance in projection (VIP) was calculated for each transcript within the selected module, which estimates the contribution of each transcript in the sPLS regression. Gene transcripts that had VIP scores > 1 were designated as VIP genes and they were considered important for the sPLS prediction of the response variable (e.g., cyanobacterial biomass) (Guidi et al., 2016).

Gene transcripts within each selected module were assumed to have a similar expression pattern across samples and be distinct from those of the other modules (Ghazalpour et al., 2006). Gene ontology (GO) enrichment analysis was performed within individual modules based on the Fisher's exact test using topGO package (Alexa and Rahnenführer, 2009) to identify modules that were enriched with gene products in a specific subcellular compartment or biochemical pathway on the of biological process (BP).

RESULTS

Temporal variations of active communities and environmental variables

A total of 28 metatranscriptomic libraries were sequenced, cyanobacteria accounted for

50.8% (ranging between 13.7-83.8% per library) of the total sequences (Fig. 15a). Active members of cyanobacterial communities went through three distinct successional stages during our study (Fig. 15a). Specifically, Anabaena (10.1-29.0% of the total sequences) dominated in

93 June (designated as the ANA stage) but were outnumbered by Microcystis (6.7-16.1%) and

Planktothrix (14.6-27.8%) from July to early August (the MIC-PLA stage). Then, the relative abundance of cyanobacterial sequences significantly decreased (Anabaena: 0.8-9.5%;

Microcystis: 0.3-1.2%; Planktothrix: 2.9-4.6%) in late August (the Post-bloom stage; Fig. 15a).

The same succession pattern was observed by microscopic measurements of individual cyanobacterial biomass (Fig. 15c). The total cyanobacterial biomass did not change significantly when the ANA stage transitioned to the MIC-PLA stage (one-way ANOVA, P >

0.05) and then significantly decreased when entering the Post-bloom stage (one-way ANOVA,

P < 0.05, Fig. 15d).

The structure of the active members of the cyanobacterial microbiome (i.e., the composition of eukaryotes, bacterioplankton, and cyanophage) shifted along with the cyanobacterial species succession (Fig. 15b). Eukaryotes accounted for 0.38-7.65% of the total sequences and were dominated by phytoplankton Bacillariophyta (0.05-3.2%) and

Chlorophyta (0.3-4.5%), and the relative abundance of phytoplankton was significantly higher in the ANA and Post-bloom stages than the MIC-PLA stage (one-way ANOVA, P < 0.05; Fig.

15b). Zooplankton sequences only accounted for < 0.1% of the total sequences.

Bacterioplankton dominated the microbiome; their relative abundance exhibited a consistently increasing trend throughout the three stages (Fig. 15b). The major taxa of bacterioplankton, i.e.,

Alphaproteobacteria (4.7-21.1%), Betaproteobacteria (2.5-23.9%), Gammaproteobacteria

(4.3-23.1%), and Actinobacteria sequences (2.0-10.6%), had a significantly higher relative abundance in the MIC-PLA and Post-bloom stages than the ANA stage (one-way ANOVA; P

94 < 0.05; Fig. 15b). Viruses only accounted for 0.2-0.9% of the total metatranscriptomic sequences and their relative abundances were not significantly different among the three bloom stages (one-way ANOVA; P > 0.05; Fig. 15b).

Environmental variables also exhibited temporal variations among the three succession stages (Fig. 16). Out of the 13 measured variables, total nitrogen (TN, 757-1960 µg/L), total organic carbon (TOC, 7.04-9.89 mg/L), and dissolved oxygen (DO, 5.2-15.5 mg/L) were significantly higher in the ANA stage compared to the MIC-PLA or Post-bloom stages (one- way ANOVA, P < 0.05, Fig. 16). pH (7.9-9.6) and microcystin (MC) concentrations (0.15-5

µg/L) were significantly higher in the MIC-PLA stage (one-way ANOVA, P < 0.05, Fig. 16).

Turbidity (3.1-19.2) was significantly higher in both ANA and MIC-PLA stages compared to the Post-bloom stage (one-way ANOVA, P < 0.05, Fig. 16). Temperature (Temp, 23.63-

28.59 ℃), electronic conductivity (EC, 208.4-250), total phosphorus (TP, 32.7-86.3 µg/L),

soluble reactive phosphorus (SRP, 23.1-48.6 µg/L), nitrate (NO3, 1.33-515 µg/L), nitrite (NO2,

4.63-41.7 µg/L), and TN/TP (12-48) showed no significant difference among different stages

(one-way ANOVA, P > 0.05, Fig. 16).

Transcriptional responses to variations of Anabaena biomass

Differentially gene expression analyses were performed (pairwise comparisons among

ANA, MIC-PLA, and Post-bloom stages) to identify genes that were responsive to bloom stage transitions in the categories of nutrient metabolism, gas vesicle, secondary metabolites, and extracellular polysaccharides (EPS) productions. Ten differentially expressed genes (DEGs) were identified from Anabaena sequences; all of them were overexpressed in the ANA stage

95 compared to either of the MIC-PLA or the Post-bloom stages (Fig. 17). Four of the above

Anabaena DEGs, were involved in N metabolism (nifBK: N fixation; amtB: ammonium transport; nirA: ferredoxin-nitrite reduction), five were related to gas vesicle synthesis

(gvpA5A6A7FW), and one was involved in EPS production (ANA_C13742, glycosyl related gene; Fig. 17). No MC synthesis gene of Anabaena was overexpressed in the ANA stage compared to either the MIC-PLA or the Post-bloom stages, although direct measurements of MC concentrations were high in the ANA stage (Fig. 16).

Weighted gene co-expression network analysis (WGCNA) was further performed to detect co-expressed genes (modules) that were correlated with biomass variations of specific cyanobacterial taxa. Two gene modules that were found significantly correlated with Anabaena biomass by WGCNA. One of these Anabaena biomass responsive (ABR) modules was from the Anabaena (Ana1) expression network (Ana1, 258 genes in total; Fig. 18a) and the other one was from the non-cyanobacteria (NC) expression network (NC8, 59 genes in total; Fig.

18d). The structures (topology) of ABR-Ana1 and -NC8 modules were also correlated significantly with the biomass of Anabaena (r > 0.60, P < 0.05, Figs. 19a-b). WGCNA further

revealed that ABR-Ana1 was also significantly correlated with TN (r = 0.54, P = 0.04) and

TN/TP (r = 0.91, P = 3e-06) (Fig. 18a), while ABR-NC8 was only significantly correlated with

TN (r = 0.62, P = 0.02) (Fig. 18d).

Expressed genes within the ABR-Ana1 module were mainly enriched (FDR < 0.05) in genes of GO (gene ontology) terms for photosynthesis, nucleoside monophosphate biosynthesis, translation, and protein-chromophore linkage (Fig. 20a). The expression level of

96 genes of the enriched GO terms in this module underwent dramatic downregulation from the

ANA stage to the MIC-PLA stage, and then slightly increased in the Post-bloom stage (lasso regression, Fig. 20a). ABR-NC8 module was mainly enriched in GO terms for photosynthetic transport chain and protein-chromophore linkage (Fig. 20c). The expression level of these enriched GO terms showed no significant change between the ANA and MIC-PLA stages and was the lowest in the Post-bloom stage (Fig. 20c).

Sparse partial least square (sPLS) regression analysis was further performed to evaluate how well the expressed genes of the identified ABR modules could explain the biomass variation of Anabaena. sPLS regression revealed that genes of the ABR-Ana1 module predicted 23.4% of the variability of Anabaena biomass (LOOCV-R2 = 0.234, Fig. 19c). In addition, 49 out of the 258 ABR-Ana1 genes were identified as VIP genes (sPLS regression,

VIP score > 1) (Table 5), which had higher prediction power for Anabaena biomass. Five of the VIP genes were affiliated with gas vesicle production, nitrogen fixation, and ammonium transport (Fig. 20b, Table 5). Most of the remaining VIP genes were affiliated with photosynthetic activities, ribosome biogenesis, and translation (Table 5), which were consistent with the results of GO enrichment analysis. The ABR-NC8 module predicted 15.1% (LOOCV-

R2 = 0.151, Fig. 19d) of the variability of Anabaena biomass. Out of the 59 genes in this module,

23 were identified as VIP genes and 8 of the VIP genes were assigned to functional categories including photosynthesis, aspartate oxidase, and ferredoxin (Fig. 20d, Table 6), different from the enriched GO terms (except photosynthetic activities) (Fig. 20c). Taxonomic classification revealed that VIP genes of the ABR-NC8 module were mainly affiliated with both algae

97 (, Bacillariophyta) and bacteria (Firmicutes, Gammaproteobacteria,

Betaproteobacteria, and Actinobacteria) phyla (Table 6).

Transcriptional response to variations of Microcystis biomass

The number of DEGs of Microcystis (27 genes) nearly doubled that of the Anabaena

(Fig. 17) and all of them were overexpressed in the MIC-PLA stage compared to either the

ANA or the Post-bloom stages (Fig. 17). Specifically, seven Microcystis DEGs were affiliated with N transformations, including genes for glutamate metabolism (glnA, gltBDX, glsF), arginine decarboxylation (MAE_46810), and amino acid transport (MAE_32490) (Figs. 17a-1 and 17a-2). Six Micorcystis DEGs were related to P metabolism, including those for alkaline phosphatase (ALP) (MAE_30190), two-component sensor histidine kinase (MAE_02100,

MAE_31720, MAE_03210), creatininase (MAE_62630), and phosphate-starvation induced

ATPase (MAE_43330). (Figs. 17b-1 and 17b-2). Anabaena DEGs did not include any of the microcystin synthesis genes. However, Microcystis DEGs included three microcystin synthesis genes (mcyACE) and their expression level was higher than the other two stages (Fig. 17e-1).

This high representation of mcy genes was coincident with the highest MC concentrations measured in the Harsha lake (Fig. 16). Like those of Anabaena, Microcystis DEGs also contained genes for synthesizing gas vesicles (gvpAI, gvpAII, gvpAIII, gvpCFJN, Fig. 17c) and

EPS (MAE_06090, MAE_27990, MAE_37160, and MAE_12350, Fig. 17d) and they were overexpressed when Microcystis was dominant (in the MIC-PLA stage).

WGCNA analysis identified five Microcystis biomass responsive (MBR) modules (Fig.

18). Three were from the Microcystis network (Mic4, Mic6, and Mic7; 153 genes in total; Fig.

98 18b); one was from the microbiome NC2 (219 genes in total; Fig. 18d), and the last one was from the cyanophage network (CP9; 61 genes in total; Fig. 18e). The structures of all five MBR modules were correlated significantly with Microcystis biomass (0.43 < r < 0.85, P < 0.05,

Figs. 21a, c, e, g, i). MBR modules were significantly correlated with several environmental

- variables, including water temperature (r > 0.51, P > 0.013), NO2 (r > 0.62, P > 0.01) and

- NO3 (r > 0.62, P > 0.01) (Fig. 18).

Expressed genes within the MBR-Mic modules (Fig. 22a) were enriched in different

(except photosynthesis) GO terms from the ABR-Ana1 module and include those affiliated with pigment biosynthesis, translation, homeostasis, ribonucleotide biosynthesis, tRNA aminoacylation, and ncRNA processing. Opposite of the ABR module (Fig. 20a), the expression levels of genes of the enriched GO terms in the MBR-Mic modules consistently showed an increasing trend from the ANA stage to the MIC-PLA stage and then decreased in the Post-bloom stage (Fig. 22a). Expressed genes within the MBR-NC2 modules (Fig. 22c) were enriched in GO terms of water-soluble vitamin biosynthesis, metabolic process, folic acid-containing compound metabolism, and cellular response to DNA damage stimulus. This list of enriched GO terms and the expression pattern of genes of these enriched GO terms both differed from those of the ABR-NC8 module (Fig. 20c).

sPLS regression analysis found that the three MBR-Mic together predicted 57% of

Microcystis biomass variation (LOOCV-R2 = 0.57, Fig. 21b, d, f). From 153 genes of these three modules, 52 were identified as VIP genes. Eight of the VIP genes were affiliated with the

+ + CO2-concentrating mechanism (CCM), CO2 fixation, gas vesicle, and Na /H antiporter (Fig.

99 22b, Table 7). The MBR-NC2 module predicted 57.8% of the Microcystis biomass variability

(LOOCV-R2 = 0.578, Fig. 21h). Within the MBR-NC2 module, 46 out of 218 genes were identified as VIPs and 5 of them were affiliated with CAZymes, alkaline phosphatase protein, and chemotaxis protein (Fig. 22d, Table 8), different from the enriched GO terms in this module

(Fig. 22c). VIP genes of the MBR-NC2 module mainly affiliated with heterotrophic bacteria in the taxa of Aplphaproteobacteria and Actinobacteria (Table 8). The MBR-CP9 module predicted 44.3% (LOOCV-R2 = 0.443, Fig. 21j) of the Microcystis biomass variability. Five out of 61 MBR-CP9 genes were VIPs and their expression a gradually increasing pattern from the

ANA to the Post-bloom stages (Fig. 22e, Table 9). The MBR-CP9 VIPs all appeared to be

Microcystis-specific, including MaMV-DC (accession number [acc. #] GCA_001505175.1),

Ma-LMM01 (acc. # GCA_000870225.1), and MACPNOA1 (acc. # GCA_004320245.1) (Fig.

22f, Table 9).

Transcriptional response to variations of Planktothrix biomass

Sixteen Planktothrix DEGs were identified across the three bloom stages and were all overexpressed in the MIC-PLA stage compared to either the ANA or Post-bloom stages (Fig.

17). Four of the Planktothrix DEGs were N metabolism-related, including genes for glutamate metabolism (gltAD, glnA) and cyanophycin synthesis (cphA) (Fig. 17a). Four Planktothrix

DEGs were related to P metabolism, including those for P-scavenging (pstAC) and two- component sensor histidine kinase (A19Y_0178, A19Y_2590) (Fig. 17b). Three Planktothrix

DEGs were related to the synthesis of secondary metabolites, including (ociAB) and anabaenopeptin (apnA) (Fig. 17e). As has been found for Anabaena and Microcystis,

100 Planktothrix DEGs again included genes for gas vesicle (gvpANG) (Fig. 17c) and EPS

(A19Y_2613, A19Y_1456) productions (Fig. 17d); and all these genes were also overexpressed when Planktothrix was dominant (in the MIC-PLA stage).

WGCNA analyses identified three Planktothrix biomass responsive (PBR) modules

(Fig. 18). One was PBR-Pla12 (136 genes in total; Fig. 18c) from the Planktothrix network; one was PBR-NC5 (184 genes in total; Fig. 18d) from the NC network; and the third one was

PBR-CP8 (47 genes in total; Fig. 18e) from the cyanophage network. The structures of all three

PBR modules were correlated significantly with Planktothrix biomass (0.39 < r < 0.58, P <

0.05, Figs. 23a, c, e). Among the PBR modules, the Pla12 module was significantly associated with water temperature (r = 0.57, P = 0.009) and pH (r = 0.53, P = 0.013) (Fig. 18c), whereas the NC5 (Fig. 18d) and CP8 (Fig. 18e) modules showed no significant correlations with any of the measured 14 environmental variables.

Expressed genes within the PBR-Pla module (Fig. 24a) were enriched in GO terms of

ATP biosynthetic process and cellular protein-containing complex assembly, which were different from those of the MBR-Mic and the ABR-Ana1 modules. The expression level of genes of these enriched GO terms within the PBR-Pla module showed a similar expression pattern as those of the MBR-Mic modules (Fig. 22a) but different from those of the ABR-Ana1 module (Fig. 20a). Expressed genes within the PBR-NC5 module (Fig. 22c) were enriched in completely different GO terms (i.e., glycogen metabolic process, regulation of , and viral life cycle) compared to the ABR-NC8 (Fig. 20c) and the MBR-NC2 modules (Fig. 22c). The expression level of genes of these enriched GO terms in the PBR-NC5

101 module had an increasing trend from the ANA stage to the Post-bloom stage and then kept constant, different from those of the ABR-NC8 and the MBR-NC2 modules.

sPLS regression analysis found that the PBR-Pla12 module predicted 21.1% (LOOCV-

R2 = 0.211, Fig. 23d) of the variability in Planktothrix biomass. Within the PBR-Pla12 module,

62 out of 136 genes were identified as VIP genes and 6 VIP genes were assigned to CCM, cyanophycin synthesis, nitrate reductase, and gas vesicle production (Fig. 24b, Table 10). The

PBR-NC5 module predicted Planktothrix biomass with high accuracy (LOOCV-R2 = 0.70, Fig.

23e). From 185 genes of the PBR-NC5 module, 45 were detected as VIP genes and 7 VIP genes were affiliated with CAZymes, a chemotaxis protein, and peptidases (Fig. 24d, Table 11), which agreed with GO enrichment results (except viral life cycle GO term). VIP genes of the

PBR-NC5 module were mainly assigned to heterotrophic bacteria in the taxa of Actinobacteria and Gammaproteobacteria. The PBR-CP8 module from the cyanophage network only predicted 11.1% (LOOCV-R2 = 0.111, Fig. 23f) of the Planktothrix biomass variability. Within the PBR-CP8 module, 12 out of 47 genes were identified as VIP genes and their expression patterns were different from the MBR-CP9 module and showed gradual increases from the

ANA stage to the MIC-PLA stage, and then kept constant in the Post-bloom stage (Fig. 24e,

Table 12). Among the VIP genes, 2 were affiliated with a Planktothrix-specific phage (i.e.,

PaV-LD, acc. # NC_016564.1) (Fig. 24f).

DISCUSSION

Overall, our results suggest that the inherent genetic traits of cyanobacteria and their

102 close interactions with co-occurring microorganisms on essential processes synergistically govern the transition of CyanoHABs (Fig. 25).

The importance of cyanobacterial traits in bloom succession

The varying physiological and functional traits among cyanobacterial species provide a competitive edge to cyanobacteria over other phytoplankton taxa during resource partitioning and allow them to win dominance and establishing dense blooms in many aquatic environments

(Huisman et al., 2018). Like algae and plants, cyanobacteria use CO2 as the carbon source for

growth. Dissolved CO2 concentration decrease during bloom development (Verspagen et al.,

2014) and thereby increases water pH, which further shifts the equilibrium of inorganic carbon towards bicarbonate and carbonate (Fig. 25) (Verspagen et al., 2014). To outcompete other

phytoplankton for CO2, cyanobacteria have evolved a CCM mechanism that enables cells to take up bicarbonate (Fig. 25) (Price et al., 2008). In the present study, three CCM genes of

Microcystis and Planktothrix were detected with high predictive power for cyanobacterial biomass (high VIP scores) (Fig. 18), suggesting a mechanism by which Microcystis and

Planktothrix achieve a competitive advantage over other phytoplankton under dense bloom conditions.

Like CyanoHABs in other freshwater lakes, MCs production is a major harmful impact on Harsha Lake. MCs can be produced by multiple common freshwater cyanobacterial species, including all three major cyanobacteria we found in Harsha Lake, i.e., Anabaena, Microcystis, and Planktothrix. However, despite the mixed cyanobacterial community, MCs synthesis genes

(e.g., mcyABCD) were highly expressed only by Microcystis during the MIC-PLA stage (Fig.

103 24), and their expression levels followed the same trend as microcystin concentrations. This suggests that Microcystis are the main contributors to MCs in Harsha Lake. Although the ecological roles of these secondary metabolites are not yet resolved, studies have found that

MCs may trigger the upregulation of EPS biosynthesis and subsequently enhance Microcystis colony formation (Gan et al., 2012; Tang et al., 2018). In line with these studies, this study observed a simultaneous increase of gene transcripts involved in the synthesis of MCs and EPS in the MIC-PLA stage (Fig. 25), which helps to explain the establishment of Microcystis blooms.

Cyanobacterial taxa responded differently to temperature (Otten et al., 2015; Paerl et al., 2016). Studies have found that, at warm temperatures, Microcystis (Nalewajko and Murphy,

2001) and Planktothrix (Oberhaus et al., 2007) have higher maximum growth rates than

Anabaena, thus warmer water temperatures would favor the growth of Microcystis and

Planktothrix over Anabaena. This cyanobacterial trait also helps to explain the winning of

Microcystis and Planktothrix in warmer sampling days (Fig. 15a).

The importance of the microbiome in cyanobacterial bloom succession

Close interactions between cyanobacteria and their associated microbiome have been suggested as mutualistic due to their various levels of complementarity (Christie-Oleza et al.,

2017). The nutrient cycling between heterotrophic bacteria and cyanobacteria is vital for a functional system, as are usually carbon- and energy-limited, while the cyanobacteria are often limited by inorganic nutrients (Christie-Oleza et al., 2017).

Consistently, in this study, several microbiome gene modules (both non-cyanobacterial

104 and cyanophages) had significant module-trait relationships with cyanobacterial biomass (Fig.

17d). Microbiome modules that were sensitive to cyanobacterial biomass mainly consisted of heterotrophic bacteria. This indicates that cyanobacterial communities may interact more closely with heterotrophic bacteria than with other microbiome microorganisms. Genes encoding CAZymes (i.e., glycoside ), alkaline phosphatase, and peptidase were mainly involved in the break-down of and complex organic compounds.

Several such genes in the microbiome modules had high VIP values in predicting cyanobacterial biomass (Fig. 25), suggesting microbiome bacteria were actively utilizing organic compounds that likely (partly) were released from cyanobacterial cells.

The cycling of nutrients between cyanobacteria and their associated microbiome was also evidenced by the detection of DEGs involved in the and glutamine:2- oxoglutarate amidotransferase (GS-GOGAT) in Microcystis and Planktothrix metatranscripts

+ at the MIC-PLA stage (Fig. 17a). This indicates that NH4 was supplied to Microcystis and

Planktothrix from the microbiome for the active synthesis of N-rich compound glutamine and glutamate (Fig. 25). In addition, a gene from Planktothrix involved in the synthesis of cyanophycin (cphA) displayed large increases in expression at the MIC-PLA stage (Fig. 17a).

Cyanophycin serves as a reservoir for newly assimilated N when cyanobacteria are exposed to an excess of N in the environment (Stein, 2015); this compound is consumed by cyanobacteria when exogenous N is depleted (Fig. 25) (Harke et al., 2016). However, nutrient measurement showed that at the MIC-PLA stage, N availability (i.e., TN and TN/TP ratios) was significantly decreased from the ANA stage (Fig. 24), reaching N deficiency (Guildford et al., 2000). These

105 observations suggest that N were actively regenerated from heterotrophic bacteria to alleviate the N limitation to cyanobacteria and sustain the growth of non-N-fixing cyanobacterial species

(i.e., Microcystis and Planktothrix in the MIC-PLA stage) during low N supply (Fig. 16).

Cyanophages are important predators of cyanobacteria and may play significant roles in regulating the dynamics of CyanoHABs (Yoshida et al., 2008; Coloma et al., 2017). In the present study, two modules from the cyanophage network showed significant correlations with

Microcysis and Planktothrix (but not Anabaena) biomass. In addition, Microcystis biomass decreased when Microcystis-specific phage gene transcripts increased (Fig. 22). These findings suggest an active role of cyanophage in regulating cyanobacterial shifts.

Overall, our results suggest that the functional profiling information recovered from metatranscriptomics data was helpful to explain the variations of cyanobacterial biomass over different bloom succession stages. However, a lower correlation was identified between the relative and absolute abundances of Microcystis compared with the correlations in Anabaena and Planktothrix. The lower correlation in Micorcystis might be caused by the microscopic counting bias as Microcystis can form colonies which are difficult to separate into individual cells, further affecting Microcystis biomass estimation.

CONCLUSIONS

Activities of cyanobacteria and their microbiome were complementary and closely synchronized to facilitate resource recycling. These interactions may contribute significantly to cyanobacterial bloom development and succession. Varied responses to the availability of

106 CO2/bicarbonate and temperature provide Microcystis and Planktothrix competitive advantages over Anabaena under dense bloom conditions; this also contributes to bloom succession from Anabaena to Microcystis/Planktothrix bloom. Moreover, the activities of cyanophages also likely influence the dynamics of cyanobacterial genotypes.

107 REFERENCES

Alexa, A., & Rahnenführer, J. (2009). Gene set enrichment analysis with topGO. Bioconductor

Improv, 27, 1-26.

Amin, S. A., Hmelo, L. R., Van Tol, H. M., Durham, B. P., Carlson, L. T., Heal, K. R., ... &

Armbrust, E. V. (2015). Interaction and signalling between a cosmopolitan phytoplankton

and associated bacteria. Nature, 522(7554), 98-101.

Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and

powerful approach to multiple testing. Journal of the Royal statistical society: series B

(Methodological), 57(1), 289-300.

Beversdorf, L. J., Miller, T. R., & McMahon, K. D. (2013). The role of nitrogen fixation in

cyanobacterial bloom toxicity in a temperate, eutrophic lake. PloS one, 8(2), e56103.

Bolger, A. M., Lohse, M., & Usadel, B. (2014). Trimmomatic: a flexible trimmer for Illumina

sequence data. Bioinformatics, 30(15), 2114-2120.

Brauer, V. S., Stomp, M., Bouvier, T., Fouilland, E., Leboulanger, C., Confurius-Guns, V., ... &

Huisman, J. (2015). Competition and facilitation between the marine nitrogen-fixing

cyanobacterium Cyanothece and its associated bacterial community. Frontiers in

microbiology, 5, 795.

Buchfink, B., Xie, C., & Huson, D. H. (2015). Fast and sensitive protein alignment using

DIAMOND. Nature methods, 12(1), 59-60.

Carmichael, W. W. (2001). Health effects of toxin-producing cyanobacteria:“The

CyanoHABs”. Human and ecological risk assessment: An International Journal, 7(5),

108 1393-1407.

Christie-Oleza, J. A., Sousoni, D., Lloyd, M., Armengaud, J., & Scanlan, D. J. (2017). Nutrient

recycling facilitates long-term stability of marine microbial –

interactions. Nature microbiology, 2(9), 1-10.

Coloma, S. E., Dienstbier, A., Bamford, D. H., Sivonen, K., Roine, E., & Hiltunen, T. (2017).

Newly isolated Nodularia phage influences cyanobacterial community dynamics.

Environmental microbiology, 19(1), 273-286.

Croft, M. T., Lawrence, A. D., Raux-Deery, E., Warren, M. J., & Smith, A. G. (2005). Algae

acquire vitamin B12 through a symbiotic relationship with bacteria. Nature, 438(7064), 90-

93.

Dobin, A., Davis, C. A., Schlesinger, F., Drenkow, J., Zaleski, C., Jha, S., ... & Gingeras, T. R.

(2013). STAR: ultrafast universal RNA-seq aligner. Bioinformatics, 29(1), 15-21.

Gan, N., Xiao, Y., Zhu, L., Wu, Z., Liu, J., Hu, C., & Song, L. (2012). The role of microcystins

in maintaining colonies of bloom-forming Microcystis spp. Environmental Microbiology,

14(3), 730-742.

Gerphagnon, M., Macarthur, D. J., Latour, D., Gachon, C. M., Van Ogtrop, F., Gleason, F. H.,

& Sime‐Ngando, T. (2015). Microbial players involved in the decline of filamentous and

colonial cyanobacterial blooms with a focus on fungal parasitism. Environmental

Microbiology, 17(8), 2573-2587.

Ghazalpour, A., Doss, S., Zhang, B., Wang, S., Plaisier, C., Castellanos, R., ... & Horvath, S.

(2006). Integrating genetic and network analysis to characterize genes related to mouse

109 weight. PLoS Genet, 2(8), e130.

Guidi, L., Chaffron, S., Bittner, L., Eveillard, D., Larhlimi, A., Roux, S., ... & Gorsky, G. (2016).

Plankton networks driving carbon export in the oligotrophic ocean. Nature, 532(7600),

465-470.

Guildford, S. J., & Hecky, R. E. (2000). Total nitrogen, total phosphorus, and nutrient limitation

in lakes and : is there a common relationship?. Limnology and oceanography, 45(6),

1213-1223.

Haas, B. J., Papanicolaou, A., Yassour, M., Grabherr, M., Blood, P. D., Bowden, J., ... & Regev,

A. (2013). De novo transcript sequence reconstruction from RNA-seq using the Trinity

platform for reference generation and analysis. Nature protocols, 8(8), 1494-1512.

Harke, M. J., Steffen, M. M., Gobler, C. J., Otten, T. G., Wilhelm, S. W., Wood, S. A., & Paerl,

H. W. (2016). A review of the global ecology, , and biogeography of the toxic

cyanobacterium, Microcystis spp. Harmful algae, 54, 4-20.

Huisman, J., Codd, G. A., Paerl, H. W., Ibelings, B. W., Verspagen, J. M., & Visser, P. M. (2018).

Cyanobacterial blooms. Nature Reviews Microbiology, 16(8), 471-483.

Huergo, L. F., Chandra, G., & Merrick, M. (2013). PII signal transduction proteins: nitrogen

regulation and beyond. FEMS microbiology reviews, 37(2), 251-283.

Kopylova, E., Noé, L., & Touzet, H. (2012). SortMeRNA: fast and accurate filtering of

ribosomal RNAs in metatranscriptomic data. Bioinformatics, 28(24), 3211-3217.

Kuno, S., Sako, Y., & Yoshida, T. (2014). Diversification of CRISPR within coexisting

genotypes in a natural population of the bloom-forming cyanobacterium Microcystis

110 aeruginosa. Microbiology, 160(5), 903-916.

Langfelder, P., & Horvath, S. (2008). WGCNA: an R package for weighted correlation network

analysis. BMC bioinformatics, 9(1), 1-13.

Langmead, B., & Salzberg, S. L. (2012). Fast gapped-read alignment with Bowtie 2. Nature

methods, 9(4), 357.

Li, B., & Dewey, C. N. (2011). RSEM: accurate transcript quantification from RNA-Seq data

with or without a reference genome. BMC bioinformatics, 12(1), 1-16.

Li, H., Barber, M., Lu, J., & Goel, R. (2020). Microbial community successions and their

dynamic functions during harmful cyanobacterial blooms in a freshwater lake. Water

Research, 185, 116292.

Love, M. I., Huber, W., & Anders, S. (2014). Moderated estimation of fold change and

dispersion for RNA-seq data with DESeq2. Genome biology, 15(12), 1-21.

Lu, J., Zhu, B., Struewing, I., Xu, N., & Duan, S. (2019). Nitrogen-phosphorus-associated

metabolic activities during the development of a cyanobacterial bloom revealed by

metatranscriptomics. Scientific reports, 9(1), 1-11.

Makarova, K. S., Wolf, Y. I., Snir, S., & Koonin, E. V. (2011). Defense islands in bacterial and

archaeal genomes and prediction of novel defense systems. Journal of bacteriology,

193(21), 6039-6056.

Menzel, P., Ng, K. L., & Krogh, A. (2016). Fast and sensitive taxonomic classification for

metagenomics with Kaiju. Nature communications, 7(1), 1-9.

Nalewajko, C., & Murphy, T. P. (2001). Effects of temperature, and availability of nitrogen and

111 phosphorus on the abundance of Anabaena and Microcystis in Lake Biwa, Japan: an

experimental approach. Limnology, 2(1), 45-48.

Oberhaus, L., Briand, J. F., Leboulanger, C., Jacquet, S., & Humbert, J. F. (2007). Comparative

effects of the quality and quantity of light and temperature on the growth of Planktothrix

agardhii and P. rubescens1. Journal of , 43(6), 1191-1199.

Oksanen, J. (2011). Multivariate analysis of ecological communities in R: vegan tutorial. R

package version, 1(7), 1-43.

Otten, T. G., & Paerl, H. W. (2015). Health effects of toxic cyanobacteria in US drinking and

recreational waters: our current understanding and proposed direction. Current

environmental health reports, 2(1), 75-84.

Paerl, H. (2008). Nutrient and other environmental controls of harmful cyanobacterial blooms

along the freshwater-marine continuum. In Cyanobacterial harmful algal blooms: State of

the science and research needs (pp. 217-237). Springer, New York, NY.

Paerl, H. W., Hall, N. S., & Calandrino, E. S. (2011). Controlling harmful cyanobacterial

blooms in a world experiencing anthropogenic and climatic-induced change. Science of

the total environment, 409(10), 1739-1745.

Paerl, H. W., & Otten, T. G. (2013). Harmful cyanobacterial blooms: causes, consequences,

and controls. Microbial ecology, 65(4), 995-1010.

Paerl, H. W., & Otten, T. G. (2016). Duelling ‘CyanoHABs’: unravelling the environmental

drivers controlling dominance and succession among diazotrophic and non-N2-fixing

harmful cyanobacteria. Environmental microbiology, 18(2), 316-324.

112 Paerl, H. (2017). The cyanobacterial nitrogen fixation paradox in natural waters.

F1000Research, 6.

Price, G. D., Badger, M. R., Woodger, F. J., & Long, B. M. (2008). Advances in understanding

the cyanobacterial CO2-concentrating-mechanism (CCM): functional components, Ci

transporters, diversity, genetic regulation and prospects for engineering into plants. Journal

of experimental botany, 59(7), 1441-1461.

Rabalais, N. N., Diaz, R. J., Levin, L. A., Turner, R. E., Gilbert, D., & Zhang, J. (2010).

Dynamics and distribution of natural and human-caused hypoxia. Biogeosciences, 7(2),

585-619.

Rohart, F., Gautier, B., Singh, A., & Lê Cao, K. A. (2017). mixOmics: An R package for ‘

feature selection and multiple data integration. PLoS computational biology, 13(11),

e1005752.

Rohrlack, T., Christiansen, G., & Kurmayer, R. (2013). Putative antiparasite defensive system

involving ribosomal and nonribosomal oligopeptides in cyanobacteria of the genus

Planktothrix. Applied and environmental microbiology, 79(8), 2642-2647.

Scheffer, M., Hosper, S. H., Meijer, M. L., , B., & Jeppesen, E. (1993). Alternative

equilibria in shallow lakes. Trends in ecology & evolution, 8(8), 275-279.

Schindler, D. W., Hecky, R. E., Findlay, D. L., Stainton, M. P., Parker, B. R., Paterson, M. J., ...

& Kasian, S. E. M. (2008). Eutrophication of lakes cannot be controlled by reducing

nitrogen input: results of a 37- whole-ecosystem experiment. Proceedings of the

National Academy of Sciences, 105(32), 11254-11258.

113 Smith, V. H. (1983). Low nitrogen to phosphorus ratios favor dominance by blue-green algae

in lake phytoplankton. Science, 221(4611), 669-671.

Stein, L. (2015). Microbiology: cyanate fuels the . Nature, 524(7563), 43-44.

Tang, X., Krausfeldt, L. E., Shao, K., LeCleir, G. R., Stough, J. M., Gao, G., ... & Wilhelm, S.

W. (2018). Seasonal gene expression and the ecophysiological implications of toxic

Microcystis aeruginosa blooms in Lake Taihu. Environmental science & technology,

52(19), 11049-11059.

Thorp, J. H., & Covich, A. P. (Eds.). (2009). Ecology and classification of North American

freshwater invertebrates. Academic press.

Verspagen, J. M., Van de Waal, D. B., Finke, J. F., Visser, P. M., Van Donk, E., & Huisman, J.

(2014). Rising CO2 levels will intensify phytoplankton blooms in eutrophic and

hypertrophic lakes. PloS one, 9(8), e104325.

Wang, K., Razzano, M., & Mou, X. (2020). Cyanobacterial blooms alter the relative

importance of neutral and selective processes in assembling freshwater bacterioplankton

community. Science of The Total Environment, 706, 135724.

Woodhouse, J. N., Kinsela, A. S., Collins, R. N., Bowling, L. C., Honeyman, G. L., Holliday,

J. K., & Neilan, B. A. (2016). Microbial communities reflect temporal changes in

cyanobacterial composition in a shallow ephemeral freshwater lake. The ISME journal,

10(6), 1337-1351.

Xie, M., Ren, M., Yang, C., Yi, H., Li, Z., Li, T., & Zhao, J. (2016). Metagenomic analysis

reveals symbiotic relationship among bacteria in Microcystis-dominated community.

114 Frontiers in Microbiology, 7, 56.

Yoshida, M., Yoshida, T., Kashima, A., Takashima, Y., Hosoda, N., Nagasaki, K., & Hiroishi,

S. (2008). Ecological dynamics of the toxic bloom-forming cyanobacterium Microcystis

aeruginosa and its cyanophages in freshwater. Applied and environmental microbiology,

74(10), 3269-3273.

Zhang, B., & Horvath, S. (2005). A general framework for weighted gene co-expression

network analysis. Statistical applications in genetics and molecular biology, 4(1).

115

Figure 15 Temporal dynamics of (a) cyanobacterial genera and (b) their associated microbiome organisms during CyanoHABs in Harsha Lake (only those organisms with relative abundance > 0.1% were shown). (c) Correlation analysis of cyanobacterial abundance between measurements by metatranscriptomic sequences (relative sequence abundance, %) and microscopically determined biomass (mg/L). (d) PCA analysis of cyanobacterial metatranscriptomes (The dot size is proportional to the total cyanobacterial biomass).

116 250 9.5 28 14 ) ） 240 15 (mg/L) ℃ 27 12 9.0 ( uS.cm ( (FNU) 230 pH 26 10 oxygen 10 erature 8.5 8 25 Turbidity 220 Temp Conductivity 5 Dissolved 6 24 8.0 210

ANA bloom ANA bloom ANA bloom ANA bloom ANA bloom 10.0 40 500 80 9.5 400 70 30 9.0 1400 /L) 300 ug 60 /L) /L) (mg/L) 8.5 /L) ug ug ug 20 ( ( ( TP TP ( 200 2 3 TOC 8.0 50 TN 1000 NO NO 100 7.5 10 40 800 0 7.0 ANA bloom ANA bloom ANA bloom ANA bloom ANA bloom 5 45 45 4 40 /L) 40 35 3 ug ( /L) 30 ug 35 2 25 TN/TP 30 SRP ( 20 1 Microcystins 15 25 0 ANA bloom ANA bloom ANA bloom

Figure 16 Environmental variables during sampling period in Harsha Lake.

117

Figure 17 Volcano plots depicting differentially expressed cyanobacterial genes in metabolisms of (a) Nitrogen, (b) Phosphorus, (c) Gas vesicle, (d) Extracellular polysaccharide, and (e) Secondary metabolites between different bloom stages (i.e., a1: ANA vs MIC-PLA; a2: MIC-PLA vs Post-bloom; a3: ANA vs Post-bloom). Red and blue dots represent over- and under-represented genes (|fold change| ≥ 1.5 and P < 0.05) in cyanobacterial metatranscriptomes, respectively. Gray dots represent genes that had no differentially expressions between stages. Text colors are used to indicate genes of different cyanobacteria: Orange, Anabaena; Green, Microcystis; Blue, Planktothrix. The locus tag name (e.g., ANA_C13742) was shown if the gene name is not available.

118

Figure 18 WGCNA networks of (a) Anabaena, (b) Microcystis, (c) Planktothrix, (d) non- cyanobacterial, and (e) cyanophage gene expression matrices. Each network is decomposed into smaller coherent modules (y-axis). Pearson’s correlations between the coherent modules’ eigenvectors and environmental parameters as well as cyanobacterial biomass were calculated. Significant correlations and associated P values are noted. Names of modules that had significant correlations with cyanobacterial biomass are shown in black font.

119 a (Ana1) b (Ana1)

r = 0.68 8 LOOCV-R2: 0.234 p = 4.7e-06 0.6

biomass 0.5 6 biomass 0.4 4 Anabaena 0.3 Anabaena for for 0.2

sig. 2 Measured 0.1 Gene 0 0.0

0.4 0.5 0.6 0.7 0.8 0.9 1.0 4 6 8 10

d (NC8) d (NC8)

r = 0.6 8 LOOCV-R2: 0.151

0.7 p = 5.1e-07 0.6

biomass 6 biomass 0.5 4 Anabaena Anabaena for for 0.4

sig. 2 0.3 Measured Gene

0.2 0

0.6 0.7 0.8 0.9 1.0 2 4 6 8 Predicted Anabaena biomass Module membership

Figure 19 (a) The WGCNA approach directly links (a) Ana1 and (c) NC8 module structures

(topology) to Anabaena biomass; sPLS regression was used to predict Anabaena biomass using the relative abundances of genes in (b) Ana1 and (d) NC8 modules.

120

Figure 20 Temporal change of averaged expression levels (Transcripts Per Million) of genes of the enriched GO terms within the (a) Ana1, and (c) NC8 modules. The value importance in projection (VIP) values of genes within the (b) Ana1 and (d) NC8 modules. The size of dots in (b) and (d) is proportional to the VIP value. Genes of interest are pinpointed.

121 a (Mic4) b (Mic4) r = 0.45 0.8 p = 0.024 20 LOOCV-R2: 0.138 biomass biomass

0.6 15

10 0.4 Microcystis Microcystis for for 5 0.2 Sig.

Measured 0 Gene 0.0 0.3 0.4 0.5 0.6 0.7 0.8 0.9 3 6 9 c (Mic6) d (Mic6) r = 0.8

0.6 p = 3.2e-07 20 LOOCV-R2: 0.071 biomass 0.5 biomass 15 0.4 10 Microcystis 0.3 Microcystis for for

ig. 5 0.2 Measured 0.1

Gene s 0 0.4 0.5 0.6 0.7 0.8 0.9 3 5 7 9 e (Mic7) f (Mic7) r = 0.79 p = 1.8e-10 20

0.8 LOOCV-R2: 0.381 biomass 0.7 biomass 15 0.6 10 0.5 Microcystis Microcystis for for 0.4

ig. 5 s 0.3 Measured

Gene 0 0.2 0.4 0.5 0.6 0.7 0.8 0.9 4 8 12 g (NC2) h (NC2) r = 0.43 p = 6.8e-14 20 LOOCV-R2: 0.578 0.8 biomass biomass 15 0.6

10 Microcystis 0.4 Micorcystis for for 5 ig. 0.2 s

Measured 0 Gene 0.0 0.4 0.5 0.6 0.7 0.8 0.9 1.0 5 10 15 20 i (CP9) j (CP9)

0.8 r = 0.85 p = 2.4e-18 20 LOOCV-R2: 0.443 biomass 0.7

biomass 15 0.6

10 0.5 Microcystis Microcystis for for

0.4 5 ig. s 0.3

Measured 0 Gene 0.2 0.6 0.7 0.8 0.9 1.0 5 10 15 20 Module membership Predicted Microcystis biomass Figure 21 (a) The WGCNA approach directly links (a) Mic4, (c) Mic6, (e) Mic7, (g) NC2, and (i) CP9 module structures (topology) to Microcystis biomass; sPLS regression was used to predict Microcystis biomass using the relative abundances of genes in (b) Mic4, (d) Mic6,

(f) Mic7, (h) NC2, (j) CP9 modules.

122

Figure 22 Temporal change of averaged expression levels (Transcripts Per Million) of genes of the enriched GO terms within the (a) Mic4, Mic6, Mic7 and (c) NC2 modules. (e)

Temporal change of averaged expression levels of high predictive power genes (VIP > 1) within the CP9 module. The value importance in projection (VIP) values of genes within the

(b) Mic4, Mic6, Mic7, (d) NC2, and (f) CP9 modules. The size of dots in (b), (d) and (f) is proportional to the VIP value. Genes of interest are pinpointed.

123 a (Pla12) b (Pla12)

r = 0.58 LOOCV-R2: 0.211 0.6 p = 0.0057 15 0.5 biomass biomass

0.4 10 Planktothrix 0.3 Planktothrix for for 5 0.2 sig. Measured Gene 0.1 0

0.60 0.70 0.80 0.90 5 10 15

c (NC5) d (NC5) r = 0.48 p = 1.9e-16 LOOCV-R2: 0.7 15 0.8 biomass biomass

10 0.6 Planktothrix Planktothrix

sig. 5 0.4 Gene Measured

0.2 0

0.4 0.5 0.6 0.7 0.8 0.9 1.0 5 10 15 20

e (CP8) f (CP8)

r = 0.39 p = 0.037 0.8 15 LOOCV-R2: 0.111 biomass 0.7 biomass

10 0.6 Planktothrix 0.5 Planktothrix for for 5 sig. 0.4 Measured Gene 0.3 0

0.7 0.8 0.9 1.0 4 8 12 16 Module membership Predicted Planktothrix biomass Figure 23 The WGCNA approach directly links (a) Pla12, (c) NC5, and (e) CP8 module structures (topology) to Planktothrix biomass; sPLS regression was used to predict

Planktothrix biomass using the relative abundances of genes in (b) Pla12, (d) NC5, (f) CP8 modules.

124

Figure 24 Temporal change of averaged expression levels (Transcripts Per Million) of genes of the enriched GO terms within the (a) Pla12 and (c) NC5 modules. (e) Temporal change of averaged expression levels of high predictive power genes (VIP > 1) within the CP8 module.

The value importance in projection (VIP) values of genes within the (b) Pla12, (d) NC5, and

(f) CP8 modules. The size of dots in (b), (d) and (f) is proportional to the VIP value. Genes of interest are pinpointed.

125

Figure 25 A schematic diagram showing potential interactions between cyanobacteria and their microbiome. DOM, dissolved organic matter; EPS, extracellular polysaccharides; AA, amino acids; ALP, alkaline phosphatase; CAZymes, carbohydrate-active enzymes; TCA, tricarboxylic acid cycle; GS-GOGAT, glutamine synthetase and glutamine:2-oxoglutarate amidotransferase; nifB, nitrogenase biosynthesis protein; nifK, Mo-nitrogenase

MoFe protein; amtB, ammonium transporter; Gln, glutamine; Glu, glutamate; glnA, glutamine synthase; glt genes, glutamate synthases; gdhA, glutamate dehydrogenase; 2-OG,

2-oxoglutarate; Arg, arginine; Asp, aspartate; CA, ; ccmK, carbon dioxide-concentrating mechanism protein; rbcS, ribulose-bisphosphate carboxylase.

Illustration of the CO2-concentrating mechanism was modified after Huisman et al., 2018.

Gray color genes were not detected in the present study, red color genes were either overexpressed or VIP genes of Anabaena in the ANA stage, and blue color genes were either overexpressed or VIPs of Microcystis/Planktothrix in the MIC-PLA stage.

126 Table 5 The summary of genes within ABR-Ana1 module.

Gene id node centrality VIP score Function ANA_C12902 80 4.217 gas vesicle structural protein ANA_C11502 65 4.195 ammonium transporter ANA_C10080 41 4.070 aldehyde dehydrogenase ANA_C12901 77 4.007 gas vesicle structural protein ANA_C13195 47 3.407 ribosome small subunit-dependent GTPase A ANA_C12191 85 3.285 prepilin-type N-terminal cleavage/methylation domain-containing protein ANA_C10716 28 3.209 RNA-binding protein ANA_C12732 63 3.088 4Fe-4S dicluster domain-containing protein ANA_C12614 54 3.024 hypothetical protein ANA_C13379 53 2.990 NA ANA_C10509 90 2.975 secretion protein HlyD ANA_C13533 39 2.965 nitrogenase molybdenum-iron protein subunit beta ANA_C11717 79 2.871 5-histidylcysteine sulfoxide synthase ANA_C13415 80 2.814 nitrogenase cofactor biosynthesis protein NifB ANA_C13706 59 2.788 restriction endonuclease ANA_C20359 69 2.717 hypothetical protein ANA_C20258 56 2.503 addiction module component ANA_C20412 43 2.484 Asp-tRNA(Asn)/Glu-tRNA(Gln) amidotransferase subunit GatC ANA_C20149 33 2.429 ATP-dependent Clp endopeptidase, proteolytic subunit ClpP ANA_C10060 28 2.228 polyphosphate kinase 1 ANA_C11291 44 2.225 potassium-transporting ATPase subunit A ANA_C20685 49 2.212 peptidase M48 ANA_C12607 7 2.138 serine/threonine protein kinase

127 ANA_C20202 44 1.980 acyl-CoA desaturase ANA_C12356 38 1.703 metallophosphoesterase ANA_C11068 20 1.672 subunit beta ANA_C20625 62 1.656 magnesium protoporphyrin IX methyltransferase ANA_C10462 84 1.568 cobyric acid synthase CobQ ANA_C13208 58 1.499 hypothetical protein ANA_C11990 11 1.498 N-acetyltransferase ANA_C12674 103 1.462 NA ANA_C12600 90 1.444 ribose-phosphate diphosphokinase ANA_C20184 10 1.440 II reaction center protein T ANA_C10661 34 1.436 UDP-N-acetylglucosamine 1-carboxyvinyltransferase ANA_C12569 73 1.373 indole-3-glycerol phosphate synthase TrpC ANA_C13062 12 1.339 non-ribosomal peptide synthetase ANA_C12051 24 1.330 hypothetical protein ANA_C20104 18 1.253 HU family DNA-binding protein ANA_C12745 21 1.232 TetR/AcrR family transcriptional regulator ANA_C11069 24 1.226 phycocyanin subunit alpha ANA_C20155 27 1.181 pyruvate dehydrogenase (acetyl-transferring) E1 component subunit alpha ANA_C11387 21 1.157 DUF4327 domain-containing protein ANA_C12355 46 1.085 50S ribosomal protein L28 ANA_C20379 51 1.084 peptidase ANA_C12048 70 1.082 photosystem II q(b) protein ANA_C10740 42 1.069 hypothetical protein ANA_C10475 12 1.042 magnesium-protoporphyrin IX monomethyl ester (oxidative) cyclase ANA_C12322 5 1.040 DUF1257 domain-containing protein

128 Table 6 The summary of genes within ABR-NC8 module.

Seq id node centrality VIP Function Taxa T_DN485346_c0_g1 5 1.582 L-aspartate oxidase Chlorophyta T_DN5451_c0_g1 16 1.248 Alanine dehydrogenase Gammaproteobacteria T_DN39128_c0_g1 35 1.193 ATP-dependent Clp protease ATP-binding subunit ClpX Fungi T_DN20766_c2_g1 4 1.188 Geranylgeranyl transferase type-2 subunit alpha uncultured Bacteria T_DN1850_c0_g1 26 1.168 P700 chlorophyll a apoprotein A1 Chlorophyta T_DN286_c0_g3 27 1.125 Photosystem II CP47 reaction center protein Chlorophyta T_DN4998_c1_g3 22 1.113 Photosystem II protein D1 Bacillariophyta T_DN2930_c0_g3 7 1.109 Photosystem II protein D1 Chlorophyta T_DN28597_c0_g1 6 1.100 Photosystem II protein D1 Chlorophyta T_DN1392_c0_g1 33 1.085 Ribosomal protein Firmicutes T_DN1_c1_g2 32 1.071 Cytochrome b6 Actinobacteria T_DN90_c0_g1 33 1.068 Caprin-1 Ciliophora T_DN31_c0_g2 35 1.053 Fungi T_DN142_c0_g1 36 1.039 Ribosomal protein Chlorophyta T_DN132_c4_g2 33 1.029 Ribosomal protein Chlorophyta T_DN175_c0_g2 14 1.022 Chaperone protein DnaK uncultured Bacteria T_DN87_c0_g1 33 1.019 Uncharacterized protein Bacillariophyta T_DN11779_c0_g1 19 1.017 50S ribosomal protein L2 Betaproteobacteria T_DN4930_c0_g1 33 1.017 Outer membrane protein Gammaproteobacteria T_DN11347_c0_g1 1 1.008 Ferredoxin uncultured Bacteria T_DN91_c0_g1 32 1.007 ATP-binding subunit Firmicutes T_DN18_c0_g3 36 1.002 7,8-didemethyl-8-hydroxy-5-deazariboflavin synthase Firmicutes

129 Table 7 The summary of genes within MBR-Mic modules.

Gene id node centrality VIP score Function MAE_55400 57 4.098 carbon dioxide-concentrating mechanism protein CcmK MAE_38000 61 3.450 Na+/H+ antiporter MAE_04020 26 3.121 membrane protein insertase YidC MAE_47870 39 3.062 ribulose bisphosphate carboxylase small subunit MAE_37630 55 2.274 DUF1350 domain-containing protein MAE_37640 62 2.141 hypothetical protein MAE_19990 56 2.091 bicarbonate transport system substrate-binding protein MAE_49220 42 1.995 RNA polymerase sigma factor MAE_29520 39 1.968 hypothetical protein MAE_49490 65 1.940 hypothetical protein MAE_12110 39 1.830 hypothetical protein MAE_37590 18 1.821 gas vesicle structural protein MAE_37620 67 1.624 gas vesicle structural protein MAE_02790 43 1.603 thioredoxin MAE_10220 26 1.566 PSII MAE_26780 38 1.554 lysine--tRNA MAE_63090 75 1.496 transposase MAE_26040 26 1.451 DUF3318 domain-containing protein MAE_56790 42 1.438 MAE_20830 41 1.437 insulinase family protein MAE_23280 65 1.422 hypothetical protein MAE_31170 27 1.412 efflux RND transporter periplasmic adaptor subunit MAE_60760 76 1.411 glycoside hydrolase

130 MAE_08280 61 1.383 MAE_42740 17 1.356 AraC family transcriptional regulator MAE_29510 54 1.341 hypothetical protein MAE_52860 64 1.303 methionyl-tRNA formyltransferase MAE_46050 49 1.292 CBS domain-containing protein MAE_04110 62 1.272 type II secretion system F family protein MAE_57600 27 1.265 hypothetical protein MAE_55260 56 1.263 DUF2949 domain-containing protein MAE_54560 27 1.238 hemolysin secretion protein D MAE_37800 26 1.224 EamA family transporter MAE_46450 37 1.214 metallophosphoesterase MAE_62720 27 1.212 ATP-dependent Clp endopeptidase, proteolytic subunit ClpP MAE_04620 36 1.211 DUF3370 domain-containing protein MAE_62690 65 1.188 aspartate-semialdehyde dehydrogenase MAE_54030 26 1.174 1,4-alpha-glucan branching protein GlgB MAE_40560 42 1.172 adenosylmethionine decarboxylase MAE_27990 26 1.148 porin MAE_24040 41 1.113 hypothetical protein MAE_31350 41 1.096 tail length tape measure protein MAE_13670 63 1.044 /fumarate reductase iron-sulfur subunit MAE_28100 43 1.044 DUF1997 domain-containing protein MAE_23930 39 1.043 DUF2488 domain-containing protein MAE_61870 42 1.039 prolipoprotein diacylglyceryl transferase MAE_40050 37 1.019

131 Table 8 The summary of genes within MBR-NC2 module.

Seq id WGCNA centrality VIP Function Taxonomy T_DN603_c0_g1 21 3.200 serine/threonine phosphatase Alphaproteobacteria T_DN5263_c0_g1 12 3.105 Uncharacterized protein Betaproteobacteria T_DN5856_c0_g1 25 3.019 lipoprotein Alphaproteobacteria T_DN7129_c1_g1 22 2.827 NADH-ubiquinone Betaproteobacteria T_DN5700_c0_g2 32 2.816 D-aminoacyl-tRNA deacylase Betaproteobacteria T_DN17903_c0_g1 10 2.729 Uncharacterized protein uncultured Bacteria T_DN5689_c0_g1 2 2.695 PEP-CTERM sorting domain-containing protein Deltaproteobacteria T_DN4792_c0_g2 1 2.622 Uncharacterized protein Betaproteobacteria T_DN1483_c0_g1 17 2.620 galactosidase Alphaproteobacteria T_DN9305_c0_g1 2 2.610 Photosystem II reaction center protein Z uncultured Bacteria T_DN294732_c0_g1 1 2.572 3-isopropylmalate dehydratase small subunit Betaproteobacteria T_DN672_c0_g2 3 2.505 50S ribosomal protein Alphaproteobacteria T_DN19893_c0_g1 1 2.471 50S ribosomal protein uncultured Bacteria T_DN19765_c0_g1 3 2.462 hypothetical protein uncultured Bacteria T_DN187_c0_g1 5 2.460 hypothetical protein uncultured Bacteria T_DN187_c0_g3 5 2.460 6-pyruvoyl tetrahydrobiopterin synthase Alphaproteobacteria T_DN507_c0_g1 7 2.430 Alkaline phosphatase-like protein Deltaproteobacteria T_DN2365_c0_g1 2 2.196 Uncharacterized protein Alphaproteobacteria T_DN12951_c0_g1 4 2.191 histone deacetylase Betaproteobacteria T_DN10591_c0_g2 1 2.176 Uncharacterized protein Betaproteobacteria T_DN29219_c1_g1 2 2.059 Photosystem I assembly protein Microcystis T_DN11807_c0_g1 3 2.038 Uncharacterized protein uncultured Bacteria T_DN4132_c0_g1 10 2.028 Photosystem I assembly protein Cuspidothrix_issatschenkoi

132 T_DN4849_c0_g1 12 1.974 Chemotaxis protein uncultured Bacteria T_DN8728_c0_g2 1 1.972 Uncharacterized protein Actinobacteria T_DN6604_c0_g1 4 1.955 Photosystem I assembly protein Microthlaspi_erraticum T_DN18481_c0_g1 1 1.948 glucosyltransferase Bacteroidetes T_DN5131_c0_g2 15 1.945 Uncharacterized protein Betaproteobacteria T_DN311_c0_g1 1 1.829 hypothetical protein GUITHDRAFT Betaproteobacteria T_DN50891_c0_g1 3 1.767 hypothetical protein AN484 22385 Deltaproteobacteria T_DN16735_c0_g1 3 1.700 Uncharacterized protein Gammaproteobacteria T_DN13846_c0_g1 1 1.657 Uncharacterized protein Bacillariophyta T_DN847_c1_g2 41 1.633 Photosystem I assembly protein Cuspidothrix_issatschenkoi T_DN291375_c0_g1 2 1.453 type II toxin-antitoxin system VapC family toxin Actinobacteria T_DN3621_c0_g1 1 1.390 histone deacetylase Bacteroidetes T_DN186_c0_g5 2 1.351 group II intron reverse transcriptase/maturase Deltaproteobacteria T_DN6797_c0_g1 1 1.346 Uncharacterized protein T_DN3371_c0_g1 1 1.332 Uncharacterized protein Bacteroidetes T_DN15962_c0_g1 1 1.319 Uncharacterized protein Deltaproteobacteria T_DN579_c0_g2 2 1.172 Uncharacterized protein Fusobacteriia T_DN43041_c0_g1 1 1.113 ParA family chromosome partitioning ATPase Verrucomicrobia T_DN20512_c0_g1 4 1.110 exo-_-1,4-glucanase / cellodextrinase T_DN16393_c0_g1 1 1.076 Uncharacterized protein Gammaproteobacteria T_DN1108_c0_g1 11 1.073 Uncharacterized protein Verrucomicrobia T_DN537_c0_g1 1 1.045 hypothetical protein EZS28 006851 Chlorophyta T_DN34442_c0_g1 4 1.028 Uncharacterized protein Actinobacteria

133 Table 9 The summary of genes within MBR-CP9 module.

Gene id node centrality VIP score Strain name AGR48654.1 5 3.436 Ma MaMV-DC AGR48703.1 9 3.396 Ma MaMV-DC BAF36214.1 5 3.165 Ma-LMM01 ARB07039.1 26 2.867 Ma MACPNOA1 AGR48585.1 18 2.867 Ma MaMV-DC

134 Table 10 The summary of genes within PBR-Pla12 module.

Gene id node centrality VIP Function A19Y_3467 14 3.738 ribulose-1,5-bisphosphate carboxylase/oxygenase small subunit A19Y_1120 21 3.179 carbon dioxide-concentrating mechanism protein CcmK A19Y_1536 19 2.936 cyanophycin synthetase A19Y_2824 9 2.796 nitrate reductase A19Y_1148 42 2.033 gas vesicle structural protein A19Y_1150 9 1.866 gas vesicle protein GvpN A19Y_0963 11 1.852 photosystem II protein PsbQ A19Y_4533 45 1.590 acyl-phosphate glycerol 3-phosphate acyltransferase A19Y_3363 16 1.540 glycine--tRNA ligase subunit beta A19Y_1793 36 1.461 alpha/beta hydrolase A19Y_1285 12 1.453 [acyl-carrier-protein] S-malonyltransferase A19Y_1438 10 1.356 ABC transporter permease A19Y_4010 29 1.329 hypothetical protein A19Y_3031 15 1.292 JAB domain-containing protein A19Y_4639 29 1.282 TIGR03032 family protein A19Y_0993 41 1.273 protein-methionine-sulfoxide reductase catalytic subunit MsrP A19Y_0106 68 1.272 MFS transporter A19Y_0718 18 1.262 MoxR family ATPase A19Y_0549 51 1.239 membrane protein A19Y_4190 20 1.237 peptide chain release factor 2 A19Y_2946 61 1.221 phosphoenolpyruvate mutase A19Y_0515 42 1.171 cytochrome C biogenesis protein CcdA A19Y_1112 17 1.171 BMC domain-containing protein

135 A19Y_3699 33 1.171 ribonuclease P A19Y_0662 57 1.164 serine/threonine-protein phosphatase A19Y_4300 45 1.146 polyphosphate:AMP phosphotransferase A19Y_2589 46 1.145 geranylgeranyl reductase A19Y_3850 31 1.125 MvdD family ATP-grasp ribosomal peptide maturase A19Y_7008 54 1.123 AAA family ATPase A19Y_2060 30 1.122 hypothetical protein A19Y_3739 43 1.116 GTPase A19Y_3448 18 1.112 type I glyceraldehyde-3-phosphate dehydrogenase A19Y_1025 27 1.110 thiol:disulfide interchange protein A19Y_0668 36 1.090 hypothetical protein A19Y_0916 45 1.080 phytoene/squalene synthase family protein A19Y_1040 24 1.080 hypothetical protein A19Y_2438 20 1.078 hypothetical protein A19Y_0635 13 1.075 excinuclease ABC subunit UvrC A19Y_3848 12 1.062 acyl carrier protein A19Y_2720 13 1.049 hypothetical protein A19Y_0213 37 1.048 peroxiredoxin A19Y_4499 26 1.046 glycogen debranching protein A19Y_4080 28 1.040 2-isopropylmalate synthase A19Y_3770 12 1.039 ABC transporter ATP-binding protein A19Y_4598 43 1.035 macrolide ABC transporter ATP-binding protein A19Y_3299 47 1.034 alpha/beta hydrolase A19Y_2021 46 1.027 N-acetyltransferase A19Y_2365 48 1.005 hypothetical protein

136 Table 11 The summary of genes within PBR-NC5 module.

Seq id node centrality VIP Function Taxonomy T_DN5041_c0_g1 3 3.199 glycogen or starch phosphorylase Alphaproteobacteria T_DN10739_c0_g2 2 3.023 galactosidase Actinobacteria T_DN6648_c0_g1 24 2.783 Uncharacterrized protein Betaproteobacteria T_DN2814_c0_g1 8 2.779 glycogen or starch phosphorylase Betaproteobacteria T_DN8925_c0_g2 3 2.685 Protein TAR1 uncultured Bacteria T_DN1883_c0_g1 12 2.684 Tryptophan halogenase Acidobacteria T_DN9929_c0_g1 25 2.661 Phosphoribulokinase Alphaproteobacteria T_DN1943_c0_g2 24 2.592 Peptidase-like protein Alphaproteobacteria T_DN539_c1_g1 10 2.570 Uncharacterized protein Betaproteobacteria T_DN5519_c1_g1 7 2.553 Replicase large subunit Alphaproteobacteria T_DN923_c0_g1 26 2.544 Uncharacterized protein Actinobacteria T_DN13425_c0_g1 5 2.493 Isocitrate dehydrogenase Alphaproteobacteria T_DN3967_c0_g1 8 2.411 Uncharacterrized protein Alphaproteobacteria T_DN891_c0_g1 4 2.406 hypothetical protein Alphaproteobacteria T_DN23736_c0_g2 14 2.392 PEP-CTERM sorting domain-containing protein Alphaproteobacteria T_DN5298_c0_g3 18 2.377 Tryptophan halogenase Fusobacteriia T_DN303_c0_g3 18 2.365 Protein TAR1 Alphaproteobacteria T_DN234_c0_g1 15 2.349 galactosidase Bacillariophyta T_DN35971_c0_g1 5 2.274 Phosphoribulokinase Chlorophyta T_DN252_c0_g2 13 2.261 Uncharacterized protein Alphaproteobacteria T_DN7158_c0_g1 4 2.180 Histidinol-phosphate aminotransferase Alphaproteobacteria T_DN4689_c0_g1 27 2.160 Uncharacterrized protein Deltaproteobacteria T_DN928_c0_g1 13 2.123 Glycosyl transferase group Alphaproteobacteria

137 T_DN18_c1_g1 26 2.105 Uncharacterized protein Actinobacteria T_DN2412_c0_g1 4 2.052 Uncharacterrized protein Bacteroidetes T_DN195_c0_g1 13 2.033 acetylmuramoyl-L-alanine amidase Alphaproteobacteria T_DN195_c0_g1 13 2.033 Genome assembly Cvelia Alphaproteobacteria T_DN5833_c1_g3 4 2.026 cyanobactin precursor peptide Betaproteobacteria T_DN204063_c0_g1 2 1.996 Putative UDP-glucose 4-epimerase Deltaproteobacteria T_DN1_c1_g1 17 1.605 ABC transporter substrate-binding protein uncultured Bacteria T_DN224_c0_g2 10 1.524 Uncharacterrized protein Acidobacteria T_DN5239_c0_g1 3 1.486 Chemotaxis protein Alphaproteobacteria T_DN4917_c0_g1 2 1.460 Uncharacterrized protein Betaproteobacteria T_DN31_c4_g1 6 1.429 Transposase Deltaproteobacteria T_DN783_c1_g1 14 1.418 hypothetical protein Gammaproteobacteria T_DN5761_c0_g1 21 1.387 60 kDa chaperonin Actinobacteria T_DN45051_c1_g1 3 1.378 Uncharacterrized protein Bacteroidetes T_DN887_c0_g1 18 1.321 hypothetical protein Chloroflexi T_DN1279_c0_g2 1 1.301 chromosome partition protein Firmicutes T_DN351_c0_g7 2 1.269 60 kDa chaperonin Flavobacteriia T_DN28_c0_g1 1 1.266 cyanobactin precursor peptide Fusobacteriia T_DN5297_c1_g2 13 1.176 ABC transporter substrate-binding protein Verrucomicrobia T_DN5377_c0_g1 3 1.160 chromosome partition protein Bacillariophyta T_DN8755_c0_g1 18 1.085 Fur family transcriptional regulator Chlorophyta T_DN1960_c1_g1 4 1.021 Phosphoribulokinase Alphaproteobacteria T_DN9637_c0_g1 19 1.016 hypothetical protein Betaproteobacteria

138 Table 12 The summary of genes within PBR-CP8 module.

Gene id node centrality VIP score Strain name ADZ31541.1 12 3.918 Planktothrix PaV-LD BAF36176.1 21 2.648 Planktothrix PaV-LD AXK90455.1 25 1.648 Cylindrospermopsis phage AXK90464.1 25 1.513 Cylindrospermopsis phage AGR48535.1 8 1.271 Ma MaMV-DC AGR48600.1 8 1.209 Ma MaMV-DC AGR48629.1 25 1.123 Ma MaMV-DC AGR48730.1 25 1.123 Ma MaMV-DC

139 CHAPTER IV

COORDINATED DIEL GENE EXPRESSION OF CYANOBACTERIA AND THEIR

MICROBIOME

PREFACE

This chapter was submitted to the Harmful Algae journal under the title “Coordinated diel gene expression of cyanobacteria and their microbiome”. The author list is as follows: Kai

Wang and Xiaozhen Mou. Kai Wang did all of the experiments, data analysis, and wrote this manuscript. Xiaozhen Mou helped me with experiment design, data analysis, and writing of this manuscript.

ABSTRACT

Circadian rhythms in cyanobacteria have been well recognized. However, whether this programmed activity of cyanobacteria could elicit coordinated diel gene expressions in their co-occurring microorganisms (microbiome) and how responses of the microbiome, in turn, impact cyanobacterial metabolism are unknown. To address these questions, a microcosm experiment and metatranscriptome sequencing of cyanobacteria and their microbiome were performed over two day-night cycles. During the experiment, 1205 Microcystis genes showed diel patterns in the whole water (WWs) microcosm. Only 42.7% (515 genes) of these diel

140 Microcystis genes retained the diel feature but had significantly higher expressions in the

Microcystis microcosm, where the microbiome communities were removed from the whole water. A total of 4779 microbiome genes showed diel expression pattern in the WW microcosm, they were affiliated with diverse phyla, including prokaryotic bacteria, eukaryotic algae, , and Fungi. A similar taxonomic composition of the microbiome was found in the microbiome microcosm, where the Microcystis were removed from the whole water; however, no diel microbiome gene was identified there. Correlation analyses showed that expressions of diel genes of Microcystis and the microbiome were significantly coordinated in the WWs, including those for C-processing, nutrient transformation (i.e., N and P), and micronutrient

metabolism (i.e., iron and vitamin B12). Overall, our results suggest that the diel fluxes of OC

and vitamin B12 in Microcystis largely impact the diel expression of microbiome genes.

Meanwhile, the microbiome communities may support the growth of Microcystis by supplying them with recycled nutrients but compete with Microcystis for iron.

INTRODUCTION

Cyanobacteria are a group of oxygenic photosynthetic prokaryotes possessing a number of adaptive strategies that give them competitive advantages over other primary producers in aquatic environments (Huisman et al., 2018). One such critical adaptation is their temporal partitioning of cellular metabolisms according to daily fluctuation of light (Diamond et al.,

2015), i.e., performing light-dependent and energy-consuming biosynthesis (anabolism) mainly during the daytime and generating energy by breaking down synthesized organic

141 molecules (catabolism) mainly at night (Welkie et al., 2019).

Bacterioplankton are universal and long-term partners of cyanobacteria in aquatic environments (Croft et al., 2005; Seymour et al., 2017). Some co-occurring bacteria are attached to cyanobacterial cells (Ploug et al., 2011; Hmelo et al., 2012), whereas others are free-living (Brauer et al., 2015). Despite their living forms, both bacterioplankton proportions closely interact with cyanobacteria and have been recognized as important members of the cyanobacterial microbiome that impact the function and structures of cyanobacterial communities (Alvarenga et al., 2017). Microbiome organisms may regenerate essential inorganic nutrients to promote the growth of cyanobacteria (Cole, 1982). Meanwhile, microbiome organisms can obtain labile organic substrates released from cyanobacteria to fulfill their heterotrophic living (Brauer et al., 2015). While circadian expression of genes is well recognized for cyanobacteria (Diamond et al., 2015; Davenport et al., 2019), the role of the cyanobacterial microbiome in this process remains largely unknown.

Diel gene expression has rarely been studied in bacterioplankton, even though this phenomenon is believed to be universal to all three domains of life (Bell-Pedersen et al., 2005).

Relevant studies have only been available recently and mostly restricted to marine environments (Ottesen et al., 2014; Frischkorn et al., 2018; Harke et al., 2018; Kolody et al.,

2019). These studies have found that marine bacterioplankton exhibit diel expression of genes that are involved in a variety of functions, most of which are involved in nitrogen and phosphorus recycling (Ottesen et al., 2014; Frischkorn et al., 2018; Harke et al., 2018; Kolody

et al., 2019), iron utilization (Kolody et al., 2019), and vitamin B12 biosynthesis (Frischkorn et

142 al., 2018). However, the driving forces underlie the day-night variation of bacterioplankton gene expressions and its role in circadian gene expression in co-occurring cyanobacteria are unknown.

This study aimed to further our understanding of mechanisms governing the circadian expression of cyanobacterial and their microbiome. We hypothesized that cyanobacteria can elicit coordinated diel expression in their microbiome and microbiome organisms may, in turn, impact the diel metabolism of cyanobacteria, due to tight interactions between cyanobacteria and microbiome organisms. Microcosm experiments were set up using surface water collected from Lake Erie, a Laurentian Great Lake that is suffering from annual CyanoHABs (Stumpf et al., 2012). Diel dynamics of the metatranscriptomes of whole water (cyanobacteria + microbiome), cyanobacteria treatment, and non-cyanobacteria (microbiome) treatment samples were examined every 12h over a 2-day period. Comparisons of gene expression patterns between treatments allowed the identification of daily expressed genes among treatment groups.

MATERIALS AND METHODS

Microcosm experiment set up

Surface water samples (~0.5 m) were collected around noon from the western basin of

Lake Erie (41.686°, -83.378°) during an ongoing Microcystis bloom on July 26, 2019. Water samples were immediately placed on ice and transported back to the lab within four hours.

Once arrived, water samples were transferred to a glass aquarium (Length: 50 cm, Width: 25

143 cm, Height: 30 cm) and incubated at room temperature under natural light for 2 hours without agitation, which resulted in a layer of concentrated Microcystis colonies (i.e., scum) at the top surface of the water (confirmed by microscopic analysis). Surface Microcystis scum (~5 cm) were collected and then filtered sequentially through 25.0-μm nylon net filters and 0.2-μm pore size membrane filters to separate Microcystis colonies and the free-living microbial communities. Filtrates passed through 0.2-μm pore size membrane filters were stored in acid- washed bottles and referred to as the filter-sterilized lake water for subsequent incubation experiments. Microcystis colonies collected on the 25.0-μm nylon net filter were further washed with 100 mL of the filter-sterilized lake water five times to remove free-living and loosely attached microbes; washed Microcystis cells were transferred to a glass aquarium filled with 10 L filter-sterilized lake water and designated as the Microcystis (MCY) treatment. The microbiome microorganisms collected on the 0.2-μm pore size membrane filters combined with the washed microbes were transferred to another glass aquarium, mixed with 10 L filter- sterilized lake water, and designated as the microbiome (MIB) treatment. The whole water

(WW) control was set up in a third glass aquarium using 10 L unprocessed raw water. All three aquaria were cleaned with 5% hydrochloric acid and Milli-Q water before use.

All three aquaria (except for the MIB treatment) were incubated in the Kent State

University greenhouse uncovered under the natural light-dark cycle at ~27 oC. Microcosms were pre-incubated for 72 hours after the setup and before sampling to allow microbial communities to acclimate. The MIB treatment was incubated in constant darkness but same temperature conditions during the pre-incubation period to reduce the viability of

144 (Hood et al., 2016). No obvious water loss was observed during the entire experiment. Water in each aquarium was stirred gently every ~six hours with a sterilized glass stir stick. Water samples (500 mL) were collected after preincubation every 12h for a total of two days. Before sample collection, water in each aquarium was mixed gently using a sterilized glass stir stick.

Water temperature (Temp), pH, dissolved oxygen (DO), and electronic conductivity (EC) were measured at sample collection using a multiparameter meter (YSI, ProODO 626281, OH, USA) and a portable pH/conductivity meter (Thermo Orion, StarA325, MD, USA). After collection, water samples were immediately filtered through 0.22-μm pore size membrane filters. Two replicate filters were made from each sampling time. Each filter was stored in a 1.5-mL microtube that contained a 600 μL RLT plus solution with RNase inhibitor (QIAGEN,

Chatsworth, CA, USA) and stored at -80 °C immediately. The filtrates (passed 0.22-μm membrane) were collected in 50-ml sterile conical centrifuge tubes and stored at -20°C until

- - + analysis of nitrate (NO3 ), nitrite (NO2 ), ammonium (NH4 ), and soluble reactive phosphorus

(SRP).

Nutrient analyses

Nutrients were measured following procedures described in standard EPA methods.

- - + Briefly, concentrations of NO3 , NO2 , NH4 , and SRP were determined based on the automated hydrazine reduction method (Ohio EPA Method 4500), phenate method (Ohio EPA method

4500B), and colorimetry method (USEPA method 365.1), respectively.

RNA extraction, library preparation, and sequencing

RNA was extracted following the protocol described in Lu et al., (2019). Briefly, the

145 filters were disrupted and lysed using a Mini-Beadbeater-16 (BioSpec Products, Inc.,

Bartlesville, OK) twice for 30 sec and then centrifuged at 10,000 g for 3 min. The supernatant was then transferred to a new sterile tube, and RNA was extracted and purified using the

AllPrep DNA/RNA kit (QIAGEN). The genomic DNA was removed using the TURBO DNA- free kit (Life Technologies, Foster City, CA). For each sample, approximately 200 ng of purified RNA was treated with Ribo-Zero (Illumina, San Diego, CA) to remove ribosomal

RNA (rRNA) and then purified using the RNeasy kit (QIAGEN). Samples were confirmed to be free of bacterial DNA by PCR using primers 27F and 1522R. Complementary DNA (cDNA) synthesis was performed using the SuperScript double-strand cDNA synthesis kit (Life

Technologies). The cDNA was purified using an Illumina NexteraXT DNA library prep kit according to the manufacturer’s instruction (Illumina). Sequencing was performed on a HiSeq platform for 150bp paired-end reads (Illumina) in Genewiz. Raw reads were deposited in the national center for biotechnology information (NCBI) short read archive database under

BioProject number SRR13450943-SRR13450964.

Sequence analysis

Low-quality reads (Phred score < 20, reads < 100 bp, and adaptor contaminants) were removed using Trimmomatic version 0.39 (Bolger et al., 2014). Ribosomal RNA (rRNA) reads were removed using SortMeRNA version 2.1 (Kopylovz and Touzet, 2012) and the remaining non-rRNA reads were used in downstream analyses.

To examine the transcriptomic patterns of Microcystis, non-rRNA reads were first mapped to Microcystis aeruginosa NIES 843 genome (accession number: NC_010296.1) using

146 STAR software version 2.7.3 (Dobin et al., 2013). The remaining unmapped reads from each sample were assembled by the de novo assembly using Trinity software (Grabherr et al., 2011).

Individual assemblies of all samples were then pooled together and clustered at 98% identity using CD-Hit (Frischkorn et al., 2018) to combine highly similar contigs across the samples.

The merged contigs were then filtered to remove sequences shorter than 200 nucleotides, rRNA reads, and translated into corresponding amino acid sequences using Prodigal software (Hyatt et al., 2011).

Taxonomic affiliation of contigs was assigned based on the Kaiju software version 1.7.2 using a comprehensive protein database (including Archaea, Bacteria, Eukaryotes, and Viruses)

(Menzel et al., 2016). A resampling procedure was performed before running the Kaiju software based on the sample that has the lowest number of sequences to ensure the results are comparable between samples. Functional annotations of translated amino acid sequences were obtained by DIAMOND BLASTP (Buchfink et al., 2015) against the UniRef90 database

(Suzek et al., 2007). Translated amino acid sequences were also annotated by the Kyoto

Encyclopedia of Genes and Genomes (KEGG) with the online Automatic Annotation Server using the single-directional best-hit method targeted to prokaryotes and with the metagenomic option selected.

Read mapping to contigs was carried out with RSEM (Li and Dewey, 2011) and the default settings with the exception of using the paired-end option and the bowtie2 (Langmead and Salzberg, 2012) option. The orthologous groups (OGs) were generated by performing a reciprocal comparison with DIAMOND BLASTP followed by MCL (Markov cluster algorithm)

147 set to an inflation parameter of 1.4 following a protocol described previously (Frischkorn et al.,

2018; Kolody et al., 2019). Read counts were summed across OGs for microbiome contigs.

For each OG, the taxonomic affiliation of the most abundant contig was used as the taxonomic affiliation for that OG.

Statistical analysis

To keep downstream analyses conservative, only those Microcystis genes or microbiome OGs that had the total number of reads >100 across all time points were used in downstream analyses. Statistical comparisons of Microcystis genes and microbiome OGs between a treatment and the control samples (i.e., Microcystis genes of the WW vs the MCY samples; microbiome OGs of the WW vs the MIB samples) were performed using the DESeq2 package (Love et al., 2014). False discovery rate (FDR) values were calculated to correct corresponding P values using the Benjamini-Hochberg algorithm (Benjamini et al., 1995).

Genes with FDR < 0.05 and an absolute fold change ≥ 1.5 were considered as differentially expressed.

Counts data were normalized using “varianceStabilizingTransformation” command in the DESeq2 package (Love et al., 2014). Significant periodicity in normalized Microcystis genes and microbiome OGs was determined using rhythmicity analysis incorporating non- parametric methods (RAIN) (Thaben et al, 2014) in R. The P values were adjusted by calculating FDR values. Microcystis gene transcripts and microbiome OGs with FDR < 0.1 were considered to have significant periodicity (Frischkorn et al., 2018). To examine co- expression between Microcystis and the microbiome, pairwise Pearson’s correlations between

148 Microcystis gene and microbiome OGs were calculated. The corresponding P values were corrected for multiple testing by calculating FDR values (Benjamini and Hochberg, 1995).

RESULTS

Variations of environmental conditions

After pre-incubation and throughout the experiment, none of the measured environmental variables had significant temporal variations within any given treatment (Fig.

26). However, four out of the eight measured environmental variables showed significant differences among treatments (Fig. 26). Specifically, pH (7.03-7.83) and electronic conductivity (EC, 125-157 μS.cm) had the lowest values in the MCYs and the highest values in the WWs (pH: 7.82-8.07; EC: 155-165 μS.cm) (one-way ANOVA, P < 0.05). Dissolved oxygen (DO) values were at the same level in the MCYs (7.05-8.9 mg/L) and WWs (7.05-9.30 mg/L) but were significantly higher in the MIBs (7.05-11.7 mg/L) (one-way ANOVA, P < 0.05).

+ NH4 values were at the same level between the MCYs (26.1-39.7 mg/L) and WWs (23.4-39.7 mg/L) but were significantly lower in the MIBs (15.8-19.7 mg/L) (one-way ANOVA, P < 0.05).

- The other four variables, i.e., temperature (Temp, 26.7-27.2 ℃), nitrate (NO3 , 13.2-37.0 mg/L),

- nitrite (NO2 , 1.3-9.4 mg/L), and soluble reactive phosphorus (SRP, 11.6-24.3 mg/L) showed no significant difference among treatments (one-way ANOVA, P > 0.05; Fig. 26).

Diel expression of Microcystis genes

Differential gene expression analysis yielded a total of 506 differentially expressed genes (DEGs) of Microcystis between MCY and WW samples (Fig. 27a). Among them, 390

149 were over-expressed and 116 were under-expressed in the MCYs relative to the WWs (Fig.

27a). RAIN analysis further identified a total of 1885 diel Microcystis genes (FDR < 0.1) in the MCYs (1195) and WWs (1205); among these diel Microcystis genes, 298 were also DEGs and were further designated as diel DEGs (dDEGs) (Fig. 27a).

Microcystis dDEGs contained 17 genes that are related to photosynthesis (PS) and

, including PS II, PS I, cytochrome b6f (Cb6f) complex, and Calvin-Benson-

Bassham cycle (CBBC) (Figs. 27b-e). Most (15 out of 17) of these Microcystis dDEGs maintained the same diel patterns in the MCYs as those in the WWs, but their expression levels

were higher in the MCYs than the WWs (log2FC=1.6-6.7). The remaining two genes (pgk,

MAE_30020 of CBBC) had opposite diel patterns in the MCYs and WWs; they peaked at night in the MCYs but peaked during the day in the WWs (Fig. 27e). In addition, the expression

levels of these CBBC genes were significantly higher in the MCYs than the WWs (log2FC=1.6-

2.9; Fig. 27e).

Four of Microcystis dDEGs were affiliated with transformations of macronutrients

+ nitrogen (N) and phosphorus (P), including NH4 and nitrate transport, and phosphate transport and regulation (Fig. 28a). All four of these Microcystis dDEGs in the MCYs had the same diel pattern as those in the WWs, with nitrate transport and phosphate regulator genes peaked during

+ the day and NH4 and phosphate transport genes peaked at night (Fig. 28a). The expression levels of nitrate transport and phosphate regulator genes were significantly higher in the MCYs

+ (log2FC=1.5-4.6) than in the WWs (Fig. 28a). In contrast, the expression levels of NH4 and

phosphate transport genes were significantly lower (log2FC = 2.1-2.7) in the MCYs than in the

150 WWs (Fig. 28a).

Seven Microcystis dDEGs belonged to micronutrient metabolism, including 3 for iron

(Fig. 28b), and 4 for cobalamin (vitamin B12) (Fig. 28c). For iron metabolism, oxygenase and ferredoxin genes had the same diel pattern in the MCYs and WWs, both of which peaked during the day. The flavodoxin gene had an opposite diel pattern in the MCYs (peaked at night) from the WWs (peaked during the day; Fig. 28b). The expression levels of all 3 iron metabolism genes were significantly higher in the MCY than in the WW (Fig. 28b, log2FC=1.8-3.2). The diel pattern and expression level of vitamin biosynthesis genes varied between the MCYs and

WWs. The precorrin6 gene exhibited the same diel pattern (peaked at night) in the MCYs and

WWs, but its expression levels were significantly lower in the MCYs than in the WWs

(log2FC=1.9-2.8) (Fig. 28c). The precorrin4 gene only showed a diel pattern in the MCYs

(peaked during the day) but not in the WWs; it expressed significantly higher in the MCYs

(log2FC=2.3-2.8). With significantly lower expression levels (log2FC=1.6-3.4), precorrin8 and

NADPH dependent FMN reductase genes lost their diel expression patterns in MCY and only showed diel pattern in the WWs (peaked at night; Fig. 28c).

Seven Microcystis dDEGs were for stress response (i.e., universal stress protein, chaperone, and peroxidase genes; Fig. 28d), and microcystin synthesis (i.e., mcyADEG) (Fig. 28e). The universal stress protein (peaked during the day) and chaperone

(peaked at night) genes only had diel pattern in the WWs, whereas glutathione peroxidase genes exhibited the same diel pattern (peaked at night) in the MCYs as those in the WWs (Fig. 28d).

All of these stress response dDEGs were expressed at significantly lower levels (log2FC=2.4-

151 4.9) in the MCYs than in the WWs (Fig. 28d). Microcystin synthesis genes (peaked during the day) also exhibited the same diel pattern in the MCYs and WWs (Fig. 28e), their expression

levels were significantly higher (log2FC=1.5-2.3) in the former samples (Fig. 28e).

Diel expression of microbiome genes

A total of 144,259 microbiome genes were obtained and 6949 of them were DEGs between the MIBs and WWs. Among these microbiome DEGs, 2978 were over-expressed and

3971 were under-expressed in the MIBs relative to the WWs. RAIN analysis identified 4779 microbiome genes that had significant diel expression patterns in the WWs but none in the

MIBs (Fig. 29a). Among the 4779 diel microbiome genes, 1222 were also DEGs and they were further designated as microbiome dDEGs (Fig. 29a). None dDEGs were identified in MIBs.

Four microbiome dDEGs in the WWs were affiliated with organic carbon (OC) degradation (i.e., carbon-active enzymes [CAZymes], respiration, peptidase, and alkaline phosphatase,) (Figs. 29b-c). CAZymes dDEGs were affiliated with Bacteroidetes, while respiration dDEGs were assigned to Planctomycetes, Ciliophora, and Zoopagomycota.

Peptidase and alkaline phosphatase dDEGs were affiliated with and

Ascomycota, respectively. CAZymes and respiration dDEGs peaked during the day in the WWs and their expression levels were significantly higher in the WWs than their expressions in the

MIBs (Fig. 29b). In contrast, peptidase and alkaline phosphatase dDEGs peaked at night in the

WWs and their expressions had the same pattern as CAZymes and respiration dDEGs between the WWs and MIBs (Fig. 29c).

Nine microbiome dDEGs in the WWs were related to nutrient transformation (N and P)

152 (Figs. 30a-c). These nutrient transformation dDEGs are associated with different taxa.

+ Glutamate dehydrogenase (GDH), NH4 transport, P-carrier protein, and P transport dDEGs were assigned to (Fungi), Discosea (), Ciliophora, and

(Chromista), respectively. The N regulator protein and deaminase dDEGs were affiliated with

Planctomycetes (Bacteria), while transaminase, accessory protein, and P starvation-

+ inducible protein dDEGs were affiliated with Proteobacteria (Bacteria). The GDH, NH4 transport, P transport, and P-carrier protein dDEGs peaked at night in the WWs, whereas the rest nutrient transformation microbiome dDEGs peaked during the day in the WWs (Figs. 30a- c). The expression levels of GDH, transaminase, P transport, and P carrier protein dDEGs were

significantly higher (log2FC=3.1-7.9) in the WWs than the corresponding genes in the MIBs,

whereas P-starvation-inducible protein dDEG expressed significantly lower (log2FC=1.5-1.6;

Figs. 30a-c) in the WWs than the corresponding gene in the MIBs. The expression levels of

+ NH4 , N regulation, deaminase, and urease accessory dDEGs were different from other nutrient transformation dDEGs and were only significantly differentially expressed during the day or

+ night. NH4 transport dDEG expressed at a similar level at night but significantly lower

(log2FC=1.9) during the day in the WWs than the corresponding gene in the MIBs (Fig. 30a).

N regulation, deaminase, and urease accessory dDEGs expressed similarly during the day but

significantly higher (log2FC=1.5-1.7) at night in the WWs than the corresponding genes in the

MIBs (Figs. 30a-b).

Five microbiome dDEGs in the WWs were affiliated with micronutrient metabolism, including 4 for iron metabolism and 1 for cobalamin-dependent methionine biosynthesis (Fig.

153 30d). No vitamin synthesis dDEGs or genes were identified in the microbiome. Cytochrome biogenesis and flavodoxin microbiome dDEGs were affiliated with Proteobacteria and

Candidatus Omnitrophica, respectively. Iron-regulated membrane protein, non-heme iron enzyme, and methionine synthase dDEGs were affiliated with Bacteroidetes. Cytochrome biogenesis, iron-regulated membrane protein, and methionine biosynthesis dDEGs peaked during the day in the WWs, while the flavodoxin and non-heme iron enzymes dDEGs peaked at night in the WWs (Fig. 30d). The expression levels of these micronutrient metabolism

dDEGs were all significantly higher (log2FC=1.6-6.9) in the WWs than the corresponding genes in the MIBs (Fig. 30d).

Another four microbiome dDEGs in the WWs were associated with stress response (i.e., chaperone, catalase, glutathione peroxidase, and superoxide dismutase) (Fig. 30e). The chaperone dDEG was assigned to Evosea, while the other dDEGs were affiliated with

Ciliophora. Chaperone and catalase dDEGs peaked at night in the WWs, while glutathione peroxidase and superoxide dismutase dDEGs peaked during the day in the WWs (Fig. 30e).

Chaperone, glutathione peroxidase, and superoxide dismutase dDEGs expressed similarly at

night but significantly higher (log2FC=1.5-2.0) during the day in the WWs than the corresponding genes in the MIBs, whereas catalase dDEG only expressed significantly higher

(log2FC=1.9) at night in the WWs than the corresponding gene in the MIBs (Fig. 30e).

The taxonomic structure of the overall microbial community showed no significant diel pattern in any of the treatments (Fig. 31). However, the relative abundances of 10 major (the relative abundance of microbiome sequences > 0.1%) taxa varied between treatments (Fig. 31).

154 The relative abundances of most bacterioplankton (i.e., Actinobacteria, Bacteroidetes,

Planctomycetes) and viruses were significantly higher in the MIBs than in the WWs (one-way

ANOVA; P < 0.05; Fig. 31), except for Proteobacteria which exhibited an opposite pattern

(one-way ANOVA; P < 0.05; Fig. 31). In contrast, most eukaryotes were significantly lower in the MIBs than in the WWs (one-way ANOVA; P < 0.05; Fig. 31).

Coordinated diel gene expression between Microcystis and the microbiome

In the WWs, the expression of essential genes to microbial growth was significantly coordinated between Microcystis and the microbiome (Fig. 32). Specifically, most of the

Microcystis dDEGs related to PS and C-fixation were positively (0.41 < r < 0.84, P < 0.05) correlated with microbiome CZAymes and respiration dDEGs (Fig. 32) and were negatively (-

0.92 < r < -0.61, P < 0.05) correlated with ALP and peptidase dDEGs of the microbiome (Fig.

32).

+ For macronutrient transformation, Microcystis dDEG related to NH4 transport exhibited significant positive correlations with GDH and deaminase dDEGs in the microbiome, while Microcystis dDEG involved in P transport had a significant positive correlation with ALP dDEG in the microbiome (Fig. 32).

In addition, 3 cobalamin biosynthesis dDEGs in Microcystis exhibited significantly positive correlations (0.51 < r < 0.72, P < 0.05) with cobalamin-dependent methionine synthase dDEG in the microbiome. For iron metabolism, one Microcystis dDEG related to heme oxygenase gene had significant positive correlations (0.63 < r < 0.75, P < 0.05) with iron- requiring dDEGs (i.e., cytochrome biogenesis and iron regulation membrane protein) in the

155 microbiome (Fig. 32).

DISCUSSION

Our results provided one of the first empirical data to support the hypothesis that the diel metabolic activities in Microcystis could elicit coordinated diel expression in its microbiome (Gasol et al., 1998). Our study found that less than 50% of Microcystis genes that showed diel expression pattern in the WWs (with microbiome) retained their diel feature in the

MCYs (without microbiome), indicating that the presence of microbiome communities had significant impacts on the diel feature of Microcystis genes. However, the expression levels of most diel Microcystis genes (i.e., PS-, iron-related, and MC synthesis genes) in WW samples

were lower than (log2FC=1.5-6.7) those in the MCYs. The exceptions were genes related to

+ NH4 transport, and vitamin and peroxidase synthesis which expressed significantly higher in the WWs. In contrast, the microbiome genes only showed diel expression patterns when

Microcystis exist (WW samples). This suggests that observed diel expression of microbiome genes in natural communities (Gasol et al., 1998) is likely microbial responses to circadian activities of Microcystis. Our study also found that diel oscillation of microbiome genes not only from heterotrophic bacteria, but also from eukaryotes, including algae, ciliates, and Fungi

(Fig. 31). This broad influence of cyanobacterial circadian activities on microbiome organisms may underly the diel fluctuation of microbial community structure and nutrient flux in the aquatic environments (Ghiglione et al., 2007; Lu et al., 2014; Ottesen et al., 2014).

In the WWs, the expressions of Microcystis C-fixation genes (Fig. 28) were positively

156 correlated with microbiome genes that are related to OC-degradation (i.e., CAZymes and respiration, Figs. 30b-c), suggesting a synchronization between Microcystis and the microbiome communities on C metabolism (Fig. 33) (Frischkorn et al., 2018). Microbiome communities clearly benefit from the acquisition of OC released by (i.e.,

Microcystis) (Geng et al., 2010) and in return, they can help lower the alleviated O2 stress through microbiome organisms’ respiration (Paerl et al., 1989; Lee et al., 2017). Meanwhile, our results found N and P transport genes of Microcystis exhibited significant correlations

(r=0.69-0.82, P < 0.05; Fig. 32) with organic N and P remineralization gene expression of microbiome communities in the WWs. This indicates that Microcystis might be using inorganic

N and P derived from microbiome remineralization (Fig. 33), consistent with results of previous microcosm co-culture experiments by direct measurements (Lidbury et al., 2015) and gene transcript analysis (Christie-Oleza et al., 2017).

Besides carbon and nutrient metabolism, coordinated diel expression was also found for cobalamin-related activities between cyanobacteria and their microbiome (Figs. 28c and

30d). Despite the critical role of the cobalamin in the growth of all organisms, the de novo biosynthesis of cobalamin has only been characterized in selected prokaryotic organisms

(Warren et al., 2002; Sañudo-Wilhelmy et al., 2014). Based on genome sequence analysis,

Microcystis aeruginosa is capable of de novo cobalamin biosynthesis. Three out of four diel

Microcystis genes that were related to cobalamin biosynthesis highly expressed (log2FC=1.6-

3.4) in the WWs and had significant positive correlations with the cobalamin-dependent methionine synthase gene of the microbiome, indicating that Microcystis might be the source

157 of cobalamin for its microbiome communities (Fig. 33). This further reinforces that the diel expression of the methionine synthase gene in the microbiome was likely caused by the diel fluxes of cobalamin biosynthesis in Microcystis.

Iron is another important co-factor to all of the organisms, and it plays important role in electron transport (Weber et al., 2006). Heme oxygenase has been reported to be involved in liberating iron from organic complexes (Saito et al., 2011) and this gene in Microcystis exhibited higher expression in the MCYs than WWs (Fig. 28c). Meanwhile, several iron- requiring genes in both Microcystis and the microbiome showed significant positive correlations with the heme oxygenase gene expression (Fig. 32). The significant positive correlations between heme oxygenase gene and other iron-requiring genes suggest that iron liberated by heme oxygenase during the day could be repurposed for other processes or organisms during the daytime (Frischkorn et al., 2018). In addition, diel flavodoxin gene expression was detected in both Microcystis (Fig. 28b) and the microbiome (Fig. 30d) and its expression was higher at night in both Microcystis (Fig. 28b) and its microbiome (Fig. 30d) than during the day. Flavodoxin protein does not require iron and related genes are usually expressed to substitute for other iron-requiring proteins when the iron is limited (Erdner and

Anderson, 1999). Therefore, the high expression of the flavodoxin gene at night indicates an iron limitation at night for both Microcystis and the microbiome communities. Furthermore, iron-related genes in Microcystis were expressed at significantly higher levels in the MCYs than in the WWs (Fig. 28b) and microbiome iron-related dDEGs showed an opposite expression pattern (Fig. 30d). This suggests an iron competition between Microcystis and its

158 microbiome communities in the WWs, which helps to explain a lower expression of PS-related genes (iron serves as a cofactor) in the WWs than MCYs during the day (Figs. 28b-d) (Kology et al., 2019). Overall, the diel pattern of iron-related genes might underscore a high iron demand by both Microcystis and microbiome organisms during the daytime, which likely causes competition for this limited source.

Expressions of microcystin (MC) synthesis genes were also detected to follow a significant diel fluctuation in the present study and the expression levels were lower in the

WWs than in the MCYs (Fig. 28e). This indicates that the synthesis of MCs was more active when without microbiome communities. This disagrees with several previous studies, which have proposed that cyanotoxins can function as deterrents against grazing (Lemaire et al., 2012;

Jiang et al., 2016). The contrasting findings once again emphasize the complexity of the physiological and ecological role of MCs. MCs have been found to protect C-fixation-related enzymes from oxidative stress (Zilliges et al., 2011). This was in line with our finding that MC synthesis genes exhibited the same diel pattern (peaked during the day) as PS and C-fixation- related genes.

There were a few experimental design limitations. The focus of this research was the impacts of circadian expression of Microcystis genes on their microbiome and its reciprocal effect. Therefore, analysis of interactions between cyanobacteria and microbiome were focused on the diel differentially expressed genes (dDEGs) between samples, although the expression of some non-diel and non-DEG genes might also be coordinated between Microcystis and the microbiome communities (Morris et al., 2008; Sher et al., 2011; Amin et al., 2015). In addition,

159 our microcosm experiments were performed in aquaria which are limited to the size and bottle effects (Zohary et al., 2005). Furthermore, our results found that some of the reads in

Microcystis treatment (MCY) samples were assigned to heterotrophic bacteria, which might limit the comparison of Microcystis gene expression between the MCY and WW samples.

Lastly, the microbiome treatment only had one replicate during the daytime, which might reduce the statistical power to detect periodically expressed genes in the microbiome communities. Nevertheless, our approach allowed us to provide one of the first empirical dataset to examine the tightly coupled expression between autotrophs (i.e., Microcystis) and their microbiome.

CONCLUSIONS

In conclusion, our study found that the diel expression in the microbiome communities are likely elicited by the circadian fluxes in Microcystis. The coordinated diel metabolic activities between Microcystis and the microbiome communities could be clearly observed in

C-processing, nutrient (i.e., N and P) recycling, and vitamin B12 supply. Our study also suggests an iron competition between Microcystis and its microbiome communities during the daytime, which in turn might limit the expression of PS-related gene expression by Microcystis.

160 REFERENCES

Alvarenga, D. O., Fiore, M. F., & Varani, A. M. (2017). A metagenomic approach to

cyanobacterial genomics. Frontiers in microbiology, 8, 809.

Amin, S. A., Hmelo, L. R., Van Tol, H. M., Durham, B. P., Carlson, L. T., Heal, K. R., ... &

Armbrust, E. V. (2015). Interaction and signalling between a cosmopolitan phytoplankton

and associated bacteria. Nature, 522(7554), 98-101.

Aylward, F. O., Boeuf, D., Mende, D. R., Wood-Charlson, E. M., Vislova, A., Eppley, J. M., ...

& DeLong, E. F. (2017). Diel cycling and long-term persistence of viruses in the ocean’s

euphotic zone. Proceedings of the National Academy of Sciences, 114(43), 11446-11451.

Bell-Pedersen, D., Cassone, V. M., Earnest, D. J., Golden, S. S., Hardin, P. E., Thomas, T. L.,

& Zoran, M. J. (2005). Circadian rhythms from multiple oscillators: lessons from diverse

organisms. Nature Reviews Genetics, 6(7), 544-556.

Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and

powerful approach to multiple testing. Journal of the Royal statistical society: series B

(Methodological), 57(1), 289-300.

Bolger, A. M., Lohse, M., & Usadel, B. (2014). Trimmomatic: a flexible trimmer for Illumina

sequence data. Bioinformatics, 30(15), 2114-2120.

Brauer, V. S., Stomp, M., Bouvier, T., Fouilland, E., Leboulanger, C., Confurius-Guns, V., ... &

Huisman, J. (2015). Competition and facilitation between the marine nitrogen-fixing

cyanobacterium Cyanothece and its associated bacterial community. Frontiers in

microbiology, 5, 795.

161 Buchfink, B., Xie, C., & Huson, D. H. (2015). Fast and sensitive protein alignment using

DIAMOND. Nature methods, 12(1), 59-60.

Cole, J. J. (1982). Interactions between bacteria and algae in aquatic ecosystems. Annual

Review of Ecology and Systematics, 13, 291-314.

Christie-Oleza, J. A., Sousoni, D., Lloyd, M., Armengaud, J., & Scanlan, D. J. (2017). Nutrient

recycling facilitates long-term stability of marine microbial phototroph-heterotroph

interactions. Nature microbiology, 2(9), 1-10.

Croft, M. T., Lawrence, A. D., Raux-Deery, E., Warren, M. J., & Smith, A. G. (2005). Algae

acquire vitamin B12 through a symbiotic relationship with bacteria. Nature, 438(7064), 90-

93.

Davenport, E. J., Neudeck, M. J., Matson, P. G., Bullerjahn, G. S., Davis, T. W., Wilhelm, S.

W., ... & McKay, R. M. L. (2019). Metatranscriptomic analyses of diel metabolic functions

during a Microcystis bloom in Western Lake Erie (United States). Frontiers in

microbiology, 10, 2081.

Diamond, S., Jun, D., Rubin, B. E., & Golden, S. S. (2015). The circadian oscillator in

Synechococcus elongatus controls metabolite partitioning during diurnal growth.

Proceedings of the National Academy of Sciences, 112(15), E1916-E1925.

Dobin, A., Davis, C. A., Schlesinger, F., Drenkow, J., Zaleski, C., Jha, S., ... & Gingeras, T. R.

(2013). STAR: ultrafast universal RNA-seq aligner. Bioinformatics, 29(1), 15-21.

Erdner, D. L., & Anderson, D. M. (1999). Ferredoxin and flavodoxin as biochemical indicators

of iron limitation during open-ocean iron enrichment. Limnology and oceanography, 44(7),

162 1609-1615.

Frischkorn, K. R., Haley, S. T., & Dyhrman, S. T. (2018). Coordinated gene expression between

Trichodesmium and its microbiome over day-night cycles in the North Pacific Subtropical

Gyre. The ISME journal, 12(4), 997-1007.

Gasol, J. M., Doval, M. D., Pinhassi, J., Calderón-Paz, J. I., Guixa-Boixareu, N., Vaqué, D., &

Pedrós-Alió, C. (1998). Diel variations in bacterial heterotrophic activity and growth in

the northwestern Mediterranean Sea. Marine Ecology Progress Series, 164, 107-124.

Geng, H., & Belas, R. (2010). Molecular mechanisms underlying Roseobacter-phytoplankton

symbioses. Current opinion in biotechnology, 21(3), 332-338.

Ghiglione, J. F., Mevel, G., Pujo-Pay, M., Mousseau, L., Lebaron, P., & Goutx, M. (2007). Diel

and seasonal variations in abundance, activity, and community structure of particle-

attached and free-living bacteria in NW Mediterranean Sea. Microbial ecology, 54(2), 217-

231.

Grabherr, M. G., Haas, B. J., Yassour, M., Levin, J. Z., Thompson, D. A., Amit, I., ... & Regev,

A. (2011). Trinity: reconstructing a full-length transcriptome without a genome from

RNA-Seq data. Nature biotechnology, 29(7), 644.

Harke, M. J., Frischkorn, K. R., Haley, S. T., Aylward, F. O., Zehr, J. P., & Dyhrman, S. T.

(2019). Periodic and coordinated gene expression between a and its diatom host.

The ISME journal, 13(1), 118-131.

Helliwell, K. E., Lawrence, A. D., Holzer, A., Kudahl, U. J., Sasso, S., Kräutler, B., ... & Smith,

A. G. (2016). Cyanobacteria and eukaryotic algae use different chemical variants of

163 vitamin B12. Current Biology, 26(8), 999-1008.

Heal, K. R., Qin, W., Ribalet, F., Bertagnolli, A. D., Coyote-Maestas, W., Hmelo, L. R., ... &

Ingalls, A. E. (2017). Two distinct pools of B12 analogs reveal community

interdependencies in the ocean. Proceedings of the National Academy of Sciences, 114(2),

364-369.

Hmelo, L. R., Van Mooy, B. A. S., & Mincer, T. J. (2012). Characterization of bacterial

epibionts on the cyanobacterium Trichodesmium. Aquatic microbial ecology, 67(1), 1-14.

Hood, R. D., Higgins, S. A., Flamholz, A., Nichols, R. J., & Savage, D. F. (2016). The stringent

response regulates adaptation to darkness in the cyanobacterium .

Proceedings of the National Academy of Sciences, 113(33), E4867-E4876.

Huisman, J., Codd, G. A., Paerl, H. W., Ibelings, B. W., Verspagen, J. M., & Visser, P. M. (2018).

Cyanobacterial blooms. Nature Reviews Microbiology, 16(8), 471-483.

Hyatt, D., Chen, G. L., LoCascio, P. F., Land, M. L., Larimer, F. W., & Hauser, L. J. (2010).

Prodigal: prokaryotic gene recognition and translation initiation site identification. BMC

bioinformatics, 11(1), 1-11.

Kolody, B. C., McCrow, J. P., Allen, L. Z., Aylward, F. O., Fontanez, K. M., Moustafa, A., ...

& Allen, A. E. (2019). Diel transcriptional response of a California Current plankton

microbiome to light, low iron, and enduring viral infection. The ISME journal, 13(11),

2817-2833.

Kondo, T., Mori, T., Lebedeva, N. V., Aoki, S., Ishiura, M., & Golden, S. S. (1997). Circadian

rhythms in rapidly dividing cyanobacteria. Science, 275(5297), 224-227.

164 Kopylova, E., Noé, L., & Touzet, H. (2012). SortMeRNA: fast and accurate filtering of

ribosomal RNAs in metatranscriptomic data. Bioinformatics, 28(24), 3211-3217.

Langmead, B., & Salzberg, S. L. (2012). Fast gapped-read alignment with Bowtie 2. Nature

methods, 9(4), 357.

Li, B., & Dewey, C. N. (2011). RSEM: accurate transcript quantification from RNA-Seq data

with or without a reference genome. BMC bioinformatics, 12(1), 1-16.

Li, Q., Lin, F., Yang, C., Wang, J., Lin, Y., Shen, M., ... & Zhao, J. (2018). A large-scale

comparative metagenomic study reveals the functional interactions in six bloom-forming

Microcystis-epibiont communities. Frontiers in microbiology, 9, 746.

Lidbury, I. D., Murrell, J. C., & Chen, Y. (2015). Trimethylamine and trimethylamine N-oxide

are supplementary energy sources for a marine heterotrophic bacterium: implications for

marine carbon and nitrogen cycling. The ISME journal, 9(3), 760-769.

Lee, M. D., Walworth, N. G., McParland, E. L., Fu, F. X., Mincer, T. J., Levine, N. M., ... &

Webb, E. A. (2017). The Trichodesmium consortium: conserved heterotrophic co-

occurrence and genomic signatures of potential interactions. The ISME journal, 11(8),

1813-1824.

Lemaire, V., Brusciotti, S., van Gremberghe, I., Vyverman, W., Vanoverbeke, J., & De Meester,

L. (2012). Genotype×genotype interactions between the toxic cyanobacterium Microcystis

and its grazer, the waterflea Daphnia. Evolutionary applications, 5(2), 168-182.

Lu, X., Zou, L., Clevinger, C., Liu, Q., Hollibaugh, J. T., & Mou, X. (2014). Temporal

dynamics and depth variations of dissolved free amino acids and polyamines in coastal

165 determined by high-performance liquid chromatography. Marine Chemistry, 163,

36-44.

Love, M. I., Huber, W., & Anders, S. (2014). Moderated estimation of fold change and

dispersion for RNA-seq data with DESeq2. Genome biology, 15(12), 1-21.

Menzel, P., Ng, K. L., & Krogh, A. (2016). Fast and sensitive taxonomic classification for

metagenomics with Kaiju. Nature communications, 7(1), 1-9.

Morris, J. J., Kirkegaard, R., Szul, M. J., Johnson, Z. I., & Zinser, E. R. (2008). Facilitation of

robust growth of colonies and dilute liquid cultures by “helper”

heterotrophic bacteria. Applied and environmental microbiology, 74(14), 4530-4534.

Ottesen, E. A., Young, C. R., Gifford, S. M., Eppley, J. M., Marin, R., Schuster, S. C., ... &

DeLong, E. F. (2014). Multispecies diel transcriptional oscillations in open ocean

heterotrophic bacterial assemblages. Science, 345(6193), 207-212.

Paerl, H. W., Priscu, J. C., & Brawner, D. L. (1989). Immunochemical localization of

nitrogenase in marine Trichodesmium aggregates: Relationship to N2 fixation potential.

Applied and environmental microbiology, 55(11), 2965-2975.

Ploug, H., Adam, B., Musat, N., Kalvelage, T., Lavik, G., Wolf-Gladrow, D., & Kuypers, M.

M. (2011). Carbon, nitrogen and O2 fluxes associated with the cyanobacterium Nodularia

spumigena in the Baltic Sea. The ISME Journal, 5(9), 1549-1558.

Saito, M. A., Bertrand, E. M., Dutkiewicz, S., Bulygin, V. V., Moran, D. M., Monteiro, F. M., ...

& Waterbury, J. B. (2011). Iron conservation by reduction of metalloenzyme inventories

in the marine diazotroph . Proceedings of the National Academy of

166 Sciences, 108(6), 2184-2189.

Sañudo-Wilhelmy, S. A., Gómez-Consarnau, L., Suffridge, C., & Webb, E. A. (2014). The role

of B vitamins in marine biogeochemistry. Annual review of marine science, 6, 339-367.

Schirrmeister, B. E., Gugger, M., & Donoghue, P. C. (2015). Cyanobacteria and the Great

Oxidation Event: evidence from genes and . Palaeontology, 58(5), 769-785.

Seymour, J. R., Amin, S. A., Raina, J. B., & Stocker, R. (2017). Zooming in on the phycosphere:

the ecological interface for phytoplankton-bacteria relationships. Nature microbiology,

2(7), 1-12.

Sher, D., Thompson, J. W., Kashtan, N., Croal, L., & Chisholm, S. W. (2011). Response of

Prochlorococcus ecotypes to co-culture with diverse marine bacteria. The ISME journal,

5(7), 1125-1132.

Stumpf, R. P., Wynne, T. T., Baker, D. B., & Fahnenstiel, G. L. (2012). Interannual variability

of cyanobacterial blooms in Lake Erie. PloS one, 7(8), e42444.

Suzek, B. E., Huang, H., McGarvey, P., Mazumder, R., & Wu, C. H. (2007). UniRef:

comprehensive and non-redundant UniProt reference clusters. Bioinformatics, 23(10),

1282-1288.

Warren, M. J., Raux, E., Schubert, H. L., & Escalante-Semerena, J. C. (2002). The biosynthesis

of adenosylcobalamin (vitamin B12). Natural product reports, 19(4), 390-412.

Weber, K. A., Achenbach, L. A., & Coates, J. D. (2006). Microorganisms pumping iron:

anaerobic microbial iron oxidation and reduction. Nature Reviews Microbiology, 4(10),

752-764.

167 Welkie, D. G., Rubin, B. E., Diamond, S., Hood, R. D., Savage, D. F., & Golden, S. S. (2019).

A hard day’s night: cyanobacteria in diel cycles. Trends in microbiology, 27(3), 231-242.

Winter, C., Herndl, G. J., & Weinbauer, M. G. (2004). Diel cycles in viral infection of

bacterioplankton in the North Sea. Aquatic Microbial Ecology, 35(3), 207-216.

Yoshida, T., Nishimura, Y., Watai, H., Haruki, N., Morimoto, D., Kaneko, H., ... & Ogata, H.

(2018). Locality and diel cycling of viral production revealed by a 24 h time course cross-

omics analysis in a coastal region of Japan. The ISME journal, 12(5), 1287-1295.

Zilliges, Y., Kehr, J. C., Meissner, S., Ishida, K., Mikkat, S., Hagemann, M., ... & Dittmann, E.

(2011). The cyanobacterial hepatotoxin microcystin binds to proteins and increases the

fitness of Microcystis under oxidative stress conditions. PloS one, 6(3), e17615.

Zohary, T., Herut, B., Krom, M. D., Mantoura, R. F. C., Pitta, P., Psarra, S., ... & Woodward, E.

M. S. (2005). P-limited bacteria but N and P co-limited phytoplankton in the Eastern

Mediterranean-a microcosm experiment. Deep Sea Research Part II: Topical Studies in

Oceanography, 52(22-23), 3011-3023.

168 WW MCY MIB 8.5 50 8 40

7.5 (µg/L) 30 pH + 4 7 20 NH 6.5 10

190 40 170 30 g/L) µ 150 ( - 3 20 130 NO EC (uS.cm) EC 110 10 14 8 12 6 g/L) µ

10 ( (mg/L) - 2 4

DO 8 NO 6 2

27.4 25 27.2 ) 20

℃ 27 g/L) µ

26.8 ( 15 26.6

Temp ( Temp 10 26.4 SRP 26.2 5 Pre 0 12 24 36 Pre 0 12 24 36

Figure 26 Measured physicochemical variables of Lake Erie and microcosm samples. Pre: preincubation; EC: electronic conductivity; DO: dissolved oxygen; Temp: temperature;

+ - - NH4 : ammonium; NO3 : nitrate; NO2 : nitrite; SRP: soluble reactive phosphorus. Green bars indicate preincubation samples, and gray bars indicate dark condition of the microcosm samples.

169

Figure 27 The expression pattern of Microcystis genes between the Microcystis (MCY) and whole water (WW) samples. (a) Differentially expressed Microcystis genes (Log2 Fold change > 2, FDR < 0.05) between the MCY and WW samples. Averaged diel differentially expressed genes (dDEGs) related to (b) PSII; (c) PSI; (d) cb6f; and (e) CBBC. Gray bars and blue bars indicate dark condition samples in the MCY and WW, respectively.

170

Figure 28 Averaged diel differentially expressed genes (dDEGs) related to (a) nitrogen and phosphorus metabolism; (b) iron metabolism; (c) vitamin biosynthesis; (d) stress response

and H2O2 depletion; and (e) microcystins synthesis. Gray bars and blue bars indicate dark conditions samples in the MCY and WW, respectively.

171

Figure 29 The expression pattern of microbiome genes between the microbiome (MIB) and whole water (WW) samples. (a) Differentially expressed microbiome genes (Log2 Fold change > 2, FDR < 0.05) between the MIBs and WWs. Averaged diel differentially expressed genes (dDEGs) of the microbiome related to (b) carbohydrate-active enzymes and microbiome respiration; (c) peptidases and alkaline phosphatase. Gray bars and blue bars indicate dark condition samples in the MIB and WW, respectively.

172

Figure 30 Averaged diel differentially expressed genes (dDEGs) of the microbiome related to

(a) Nitrogen metabolism; (b) Organic nitrogen metabolism; (c) Phosphorus metabolism; (d)

Iron metabolism and methionine synthesis; and (e) Stress response and H2O2 depletion. Gray bars and blue bars indicate dark condition samples in MIB and WW, respectively.

173

Figure 31 Microbiome communities identified in the MIBs and WWs (only those taxa with relative abundance > 0.1% were shown).

174

Figure 32 Correlations between Microcystis genes (X-axis) and microbiome genes (Y-axis).

Only correlations with adjusted P < 0.05 were shown.

175

Figure 33 A schematic diagram showing potential interactions between Microcystis and its microbiome communities. AA, amino acids; ALP, alkaline phosphatase; CAZymes, carbohydrate-active enzymes; TCA, tricarboxylic acid cycle; NR, nitrate reductase; NiR, nitrite reductase; Gln, glutamine; Glu, glutamate; glnA, glutamine synthase; glt genes, glutamate synthases; GDH, glutamate dehydrogenase; α-KG, alpha-Ketoglutarate; MCs, microcystins; Arg, arginine; Leu, Leucine; glgABC, glycogen synthesis; glgP, glycogen degradation; glgX, Glycogen debranching enzyme; pgi, glucose-6-phosphate ; pgk, phosphoglycerate kinase; pgm, phosphoglucomutase; pdh, pyruvate dehydrogenase; pyk, pyruvate kinase. Genes peaked in the daytime and night were in red and blue color, respectively.

176 CHAPTER V

SUMMARY

A global proliferation of CyanoHABs is occurring with increasing frequency, intensity, and duration in response to anthropogenic eutrophication and climate changes (Huisman et al.,

2018). Besides environmental factors, cyanobacterial traits (i.e., N2 fixation, gas vesicle, CCM, and cyanotoxins) can provide them with competitive advantages over other phytoplankton, further promoting the dominance and development of CyanoHABs worldwide (Huisman et al.,

2018). Cyanobacteria have also been suggested interacting closely with co-occurring microbial communities (Seymour et al., 2017). However, little is known about the level of significance of cyanobacteria-bacteria interactions in bacterial community assembly and CyanoHABs development and succession. The overall goal of this dissertation is to further our understanding of the interactions between cyanobacteria and the co-occurring microorganisms.

Specifically, we studied the impacts of cyanobacteria on bacterial community assembly

(Chapter II), the impacts of co-occurring microorganisms on CyanoHABs development

(Chapter III), and the responses of their respective physiological states (Chapter IV). The following sections provide a summary of each subject and the significance.

Bacterial community assembly over different CyanoHABs stages

Microorganisms play important roles in environmental processes (Worden et al., 2015) and they realize their ecological functions largely in the form of a community (Lepp et al.,

177 2004). Thus, elucidating mechanisms that shape microbial community composition is essential in understanding their ecological functions (Cira et al., 2018). It is commonly agreed that selective and neutral processes simultaneously control microbial community structure

(Cira et al., 2018). CyanoHABs represent a series of complex biological disturbances to water ecosystems and can bring substantial impacts on microbial community structure (Paerl, 1988).

The strength of these biological disturbances varies with different bloom succession states and may, in turn, alter natural microbial community structure (Berry et al., 2017). However, the relative importance of biological disturbances and neutral processes in regulating the co- occurring bacterial communities over different CyanoHABs stages remains unknown.

To examine the mechanisms of bacterial community assembly over different

CyanoHABs development stages, samples were collected from Barberton Reservoir, where serves as a primary drinking water source for the city of Barberton Ohio. Barberton Reservoir was selected as a sampling site because it is routinely monitored for water quality by Ohio EPA and has a prior history of CyanoHABs (Ohio EPA). 16S rRNA amplicon sequencing was performed to examine the composition of bacterial communities. Bacterial cell counts were measured using flow cytometry to correct the relative abundance of bacterial communities recovered from 16S amplicon sequencing data. Cyanobacterial cells were also counted using a compound microscope and converted to biovolumes to reflect the stages of CyanoHABs development.

Our results showed that neutral processes explained over 67% of bacterial community variations and its fit was weaker during the cyanobacterial blooms (R2 = 0.322) than the pre-

178 (R2 = 0.549) and after-bloom stages (R2 = 0.535). In addition, environmental factors could explain shifts of non-neutral bacterial communities (63.9%) better than neutral bacterial communities (34.5%). Furthermore, strong positive correlations were found between bacterial communities and cyanobacteria than with other phytoplankton. Overall, our results emphasize the importance of neutral processes in determining bacterial community structure over different

CyanoHABs stages, and its relative importance was weakened by the increased biological disturbances during the bloom period. Our results also provide suggestions for continued effort focusing more on functional variations as changes of community taxonomic structures are mainly neutral processes determined.

Co-occurring microbes affect CyanoHABs development

CyanoHABs have been found to transmit from N2 fixer-dominated bloom to non-N2 fixer-dominated bloom in many freshwater environments along with N concentrations decreasing. This “opposite scenario” has been attributed to temperature as the N-fixation enzyme (i.e., nitrogenase) is temperature-sensitive (Paerl et al., 2016). Nutrient recycling by heterotrophic bacteria has been shown to supply a large proportion of nutrients to cyanobacteria when the external supply is low (Christie-Oleza et al., 2017). Although the impacts of individual environmental and biotic factors on cyanobacterial growth are becoming clear

(Huisman et al., 2018), their interplay in driving cyanobacterial species transitions during

CyanoHABs remains largely unknown. Elucidating mechanisms governing CyanoHAB development and species succession is critical to understand the ecology of CyanoHABs, which knowledge is essential to guide wise management strategies that can help to prevent

179 and/or mitigate CyanoHABs pollution.

To address this knowledge gap, metatranscriptomes of cyanobacteria and the co- occurring microorganisms were examined weekly over a four-month period (June to September)

in Harsha Lake, where often experiences CyanoHABs transition (N2 fixer to non-N2 fixer) in the past decades. Cyanobacterial biomasses were also measured to reflect the stages of

CyanoHABs development and compared with the functional profile recovered from metatranscriptomic data.

Our results showed that the remineralization process of heterotrophic bacteria released inorganic nutrients which support cyanobacteria growth when the nutrient is limited. In

addition, the varied responses to CO2/bicarbonate availability and high temperature provide

Microcystis and Planktothrix competitive advantage over Anabaena, which play important roles in species transition. Gas vesicle production plays important role in bloom development for all of the three cyanobacterial genera (Anabaena, Microcystis, and Planktothrix) in Harsha lake it’s less important in species transition.

Coordinated physiological responses between cyanobacteria and the co-occurring microbes

While circadian expression of genes is well recognized for cyanobacteria (Diamond et al., 2015), the significance of this process in the cyanobacterial microbiome remains largely unknown. Diel gene expression has rarely been studied in bacterioplankton, even though this phenomenon is believed to be universal to all three domains of life (Bell-Pedersen et al., 2005).

Relevant studies have only been available recently and mostly restricted to marine

180 environments (Ottesen et al., 2014; Frischkorn et al., 2018; Harke et al., 2018; Kolody et al.,

2019). These studies have found that marine bacterioplankton exhibit diel expression of genes that are involved in a variety of functions, including N and P recycling (Ottesen et al., 2014;

Frischkorn et al., 2018; Harke et al., 2018; Kolody et al., 2019), iron utilization (Kolody et al.,

2019), and vitamin B12 biosynthesis (Frischkorn et al., 2018). However, the driving forces underlie the day-night variation of bacterioplankton gene expressions and its role in circadian gene expression in cyanobacteria are unknown.

To address this knowledge gap, microcosm experiments were set up using surface bloom water collected from Lake Erie, a Laurentian Great Lake that is suffering from annual

Microcystis bloom (Stumpf et al., 2012). Diel dynamics of the metatranscriptomes of whole water (Microcystis + microbiome), Microcystis treatment, and non-cyanobacteria (microbiome communities) treatment samples were examined every 12 hours over a 2-day period.

Comparisons of gene expression between treatments and control allowed the identification of coordinated periodically expressed genes.

Our results showed that the period gene expression in the microbiome communities is likely elicited by the cyanobacteria host. The presence or absence of microbiome communities does not affect the circadian gene expression in Microcystis. The coordinated periodically expression genes are mainly involved in carbon and nutrients (N and P) cycling and vitamin

B12 supply. Besides cooperation, an iron competition was also observed between Microcystis and the co-occurring microorganisms, which might limit the photosynthesis ability of

Microcystis.

181 Overall, our research on the reciprocal interactions between cyanobacteria and the co- occurring microorganisms provides one of the first pieces of evidence that the periodic gene expression in the co-occurring microorganisms is likely caused by the cyanobacterial host. In addition, our research emphasizes the importance of the co-occurring microorganisms in regulating CyanoHABs, which brings new perspectives of preventing and/or mitigating

CyanoHABs pollution. Moreover, our results provide evidence that cyanobacterial bloom as a biological disturbance can bring substantial impacts on microbial community structure. A continued effort is necessary to study the interactions between cyanobacteria and their associated microorganisms, particularly with an emphasis on interactions within the region immediately surrounding individual cyanobacterial cells.

182 REFERENCES

Bell-Pedersen, D., Cassone, V. M., Earnest, D. J., Golden, S. S., Hardin, P. E., Thomas, T. L.,

& Zoran, M. J. (2005). Circadian rhythms from multiple oscillators: lessons from diverse

organisms. Nature Reviews Genetics, 6(7), 544-556.

Berry, M. A., Davis, T. W., Cory, R. M., Duhaime, M. B., Johengen, T. H., Kling, G. W., ... &

Denef, V. J. (2017). Cyanobacterial harmful algal blooms are a biological disturbance to

Western Lake Erie bacterial communities. Environmental microbiology, 19(3), 1149-1162.

Christie-Oleza, J. A., Sousoni, D., Lloyd, M., Armengaud, J., & Scanlan, D. J. (2017). Nutrient

recycling facilitates long-term stability of marine microbial phototroph–heterotroph

interactions. Nature microbiology, 2(9), 1-10.

Cira, N. J., Pearce, M. T., & Quake, S. R. (2018). Neutral and selective dynamics in a synthetic

microbial community. Proceedings of the National Academy of Sciences, 115(42), E9842-

E9848.

Diamond, S., Jun, D., Rubin, B. E., & Golden, S. S. (2015). The circadian oscillator in

Synechococcus elongatus controls metabolite partitioning during diurnal growth.

Proceedings of the National Academy of Sciences, 112(15), E1916-E1925.

Frischkorn, K. R., Haley, S. T., & Dyhrman, S. T. (2018). Coordinated gene expression between

Trichodesmium and its microbiome over day-night cycles in the North Pacific Subtropical

Gyre. The ISME journal, 12(4), 997-1007.

Harke, M. J., Frischkorn, K. R., Haley, S. T., Aylward, F. O., Zehr, J. P., & Dyhrman, S. T.

(2019). Periodic and coordinated gene expression between a diazotroph and its diatom host.

183 The ISME journal, 13(1), 118-131.

Huisman, J., Codd, G. A., Paerl, H. W., Ibelings, B. W., Verspagen, J. M., & Visser, P. M. (2018).

Cyanobacterial blooms. Nature Reviews Microbiology, 16(8), 471-483.

Seymour, J. R., Amin, S. A., Raina, J. B., & Stocker, R. (2017). Zooming in on the phycosphere:

the ecological interface for phytoplankton-bacteria relationships. Nature microbiology,

2(7), 1-12.

Kolody, B. C., McCrow, J. P., Allen, L. Z., Aylward, F. O., Fontanez, K. M., Moustafa, A., ...

& Allen, A. E. (2019). Diel transcriptional response of a California Current plankton

microbiome to light, low iron, and enduring viral infection. The ISME journal, 13(11),

2817-2833.

Lepp, P. W., Brinig, M. M., Ouverney, C. C., Palm, K., Armitage, G. C., & Relman, D. A.

(2004). Methanogenic Archaea and human periodontal disease. Proceedings of the national

academy of sciences, 101(16), 6176-6181.

Ottesen, E. A., Young, C. R., Gifford, S. M., Eppley, J. M., Marin, R., Schuster, S. C., ... &

DeLong, E. F. (2014). Multispecies diel transcriptional oscillations in open ocean

heterotrophic bacterial assemblages. Science, 345(6193), 207-212.

Paerl, H. W. (1988). Nuisance phytoplankton blooms in coastal, estuarine, and inland waters 1.

Limnology and Oceanography, 33(4part2), 823-843.

Paerl, H. W., & Otten, T. G. (2016). Duelling ‘CyanoHABs’: unravelling the environmental

drivers controlling dominance and succession among diazotrophic and non‐N2‐fixing

harmful cyanobacteria. Environmental microbiology, 18(2), 316-324.

184 Stumpf, R. P., Wynne, T. T., Baker, D. B., & Fahnenstiel, G. L. (2012). Interannual variability

of cyanobacterial blooms in Lake Erie. PloS one, 7(8), e42444.

Worden, A. Z., Follows, M. J., Giovannoni, S. J., Wilken, S., Zimmerman, A. E., & Keeling, P.

J. (2015). Rethinking the marine carbon cycle: factoring in the multifarious lifestyles of

microbes. Science, 347(6223).

185