WILKES-DISSERTATION-2018.Pdf (4.378Mb)

Chemostat and Modeling Investigations of Algal Photosynthetic Carbon Isotope Fractionation

The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters

Citation Wilkes, Elise. 2018. Chemostat and Modeling Investigations of Algal Photosynthetic Carbon Isotope Fractionation. Doctoral dissertation, Harvard University, Graduate School of Arts & Sciences.

Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:41127152

Terms of Use This article was downloaded from Harvard University’s DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http:// nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of- use#LAA

Chemostat and Modeling Investigations of Algal Photosynthetic Carbon Isotope Fractionation

A dissertation presented by

Elise Wilkes

The Department of Earth and Planetary Sciences

In partial fulfillment of the requirements for the degree of Doctor of Philosophy

in the subject of

Earth and Planetary Sciences

Harvard University Cambridge, Massachusetts

April, 2018

Dissertation Adviser: Ann Pearson Elise Wilkes

Chemostat and Modeling Investigations of Algal Photosynthetic Carbon Isotope Fractionation

Abstract Marine eukaryotic phytoplankton produce organic matter that is depleted in 13C relative to ambient dissolved carbon dioxide. This photosynthetic carbon isotope fractionation (εP) is recorded in marine sediments and used to resolve changes in the global carbon cycle, including variations in atmospheric pCO2.

These applications rely upon a coherent understanding of the environmental and physiological controls on

P. While classical models for εP are based on the balance between diffusion of CO2 and its fixation into biomass by the enzyme RubisCO, the details of phytoplankton carbon dynamics in reality are more complex. Phytoplankton employ a diversity of RubisCO types, and they also use carbon concentrating mechanisms (CCMs) that enhance intracellular CO2 concentrations. It is essential to understand the significance of these physiological features as controls on εp, as they may play important roles in explaining sedimentary archives.

Here I performed CO2 and growth rate (μ) manipulation experiments with modern phytoplankton in chemostat cultures to address outstanding questions regarding the mechanistic underpinning of P. First,

I characterized the stable carbon isotope ratios of coccolith-associated polysaccharides (CAPs) and other cellular constituents (bulk biomass, coccolith calcite, and alkenones) of Emiliania huxleyi. CAPs are involved in regulating calcification and have been recovered from sediments dating back ~180 Ma. It has been proposed that the carbon isotopic contents of CAPs may be used in combination with other proxies to reconstruct ancient atmospheric pCO2 levels. I find that the CAPs are isotopically enriched relative to bulk biomass and vary with μ and CO2. These results are explained by a simple model that predicts cellular carbon allocation to major organic carbon compound classes in E. huxleyi. My findings suggest that CAPs are less sensitive than alkenones as proxies for pCO2, but that combining CAP data together with data for alkenones and calcite may help reconstruct pCO2 with fewer assumptions than current approaches.

I also performed chemostat culture experiments with the dinoflagellate Alexandrium tamarense, which uses an unusual form of the carbon-fixing enzyme RubisCO (Form II). It commonly is assumed that

iii the kinetic isotope effect associated with RubisCO establishes the theoretical maximum value of εP, which is known as εf. I found that P values for A. tamarense varied with the ratio /[CO2(aq)] and approached an

εf value of 27‰. This value is larger than theoretical predictions for Form II RubisCO and is not significantly different from the f values observed for more recently-evolved taxa that employ Form ID

RubisCO, including E. huxleyi. This consistency across taxa may help to explain the broad uniformity of carbon isotope fractionation between organic and inorganic pools observed throughout the Phanerozoic, and it may pave the way for new algal pCO2 proxies based on dinoflagellate biomarkers or fossil dinoflagellate cysts.

My work on A. tamarense also implies that an f value of 25-27‰ may be a universal property of red-lineage eukaryotic phytoplankton. This finding has major implications for reinterpreting the classical models for P, because it indicates that its maximum value (f) is unlikely to reflect the intrinsic isotope fractionation of RubisCO. By extension, this implies that RubisCO activity is not the kinetically slow step of carbon fixation in these phytoplankton, at least when cultivated in nutrient-limited chemostats. Based on this finding and other support from the literature, I propose a generalized model of carbon isotope fraction in eukaryotic phytoplankton that is able to reconcile the apparent uniformity of f (as inferred from in vivo studies) vs. the isotopic heterogeneity of RubisCO (as inferred from in vitro studies). The model introduces a nutrient- and light-dependent step upstream of RubisCO that is proposed to be a kinetic barrier to carbon acquisition and a significant source of isotope fractionation. Together, the results of this thesis imply that the kinetics, intrinsic discrimination, and taxonomy of RubisCO may be largely irrelevant to the expression of p under growth conditions of low nutrients and high photosynthetic activity, e.g., in the ocean gyres or away from coastal upwelling zones. Existing environmental data are consistent with this idea and suggest that alkenone and/or other organic pCO2 proxies should be reevaluated from the perspective of local nutrient dynamics and cellular growth conditions.

Table of Contents Abstract iii Acknowledgements vi List of Tables and Figures viii

Chapter 1 − Introduction 1

Chapter 2 − Carbon isotope ratios of coccolith-associated polysaccharides of Emiliania huxleyi 17 as a function of growth rate and CO2 concentration

Chapter 3 − CO2-dependent carbon isotope fractionation in the dinoflagellate Alexandrium 47 tamarense Chapter 4 − A general model for carbon isotope fractionation in eukaryotic phytoplankton 80

Chapter 5 − Ongoing work and future directions 120

Appendix A – Supporting information for Chapter 2 133

Appendix B – Supporting information for Chapter 3 140

Appendix C – Supporting information for Chapter 4 144

Appendix D – Supporting information for Chapter 5 169

Acknowledgements I am grateful to Ann Pearson for her endless optimism, and for her confidence in me and my ideas.

I will always appreciate the time she has spent with me discussing science, the patience she has shown me throughout my time at Harvard, and her thorough, careful revisions of this dissertation.

I would also like to thank Dave Johnston for his support and mentorship. I am grateful to Andy

Knoll, Jerry Mitrovica, and Miaki Ishii for serving on my committee, past and present, and to my co-authors

Rosalind Rickaby, Harry McClelland, and Renee Lee.

Thank you to all the current and former members of the Pearson lab. Susie Carter deserves significant recognition for her contributions to this research. She has helped me with method development and to overcome many chemostat-related setbacks. I am grateful to Jenan Kharbush, Nagissa Mahmoudi, and Felix Elling for their friendship, advice, and for always being willing to answer my questions about biology lab techniques. Thank you to Ana C. Gonzalez Valdes, Jiaheng Shen, and Carolyn Zeiner for being generous with their time and friendship. I have shared my desk space and many conversations over the years with Lindsay Hays, Greg Henkes and Jordon Hemingway. I know I have benefitted tremendously from their perspectives. I am thankful to Sarah Hurley, Roderick Bovee, and Hilary Close for helping me to learn from their unique experiences. It has been a pleasure working with my student intern, Katie

Mabbott, who helped with lab analyses supporting this work. I would also like to thank Einat Segev and

Roberto Kolter for the opportunity to join the scientific party of the R/V Endeavor for field work.

I would like to acknowledge my former mentors who inspired me to pursue graduate school and scientific research: Laura Wasylenki, Jill Mikucki, Ivan Apahamian, Ariel Anbar, Mukul Sharma, Justin

Foy, and Bridget Alex.

I appreciate the hard work of the Harvard EPS department administration and support staff. I feel fortunate to have shared my time in this department with so many wonderful people, including Harriet Lau,

Alex Turner, Chris Parendo, Sunny Parker, Simon Lock, Tamara Pico, Anna Waldeck, Emma Bertran,

Jocelyn Fuentes, Marena Lin, and Athena Eyster. I am also grateful for my time in HGWISE and the friendships I developed through that organization.

I have an incredible support network outside of Harvard. Thank you to my dear friends in Boston:

Mac Krumpak, Dalia Larios, Emily and Andy Stuntz, Emma Smithayer, Michelle and Alex Davis, and

Jason Goodman. To Becky Rapf, Libby Parker, and Rachel Rosenberg—your support from afar means a lot. Thank you to my sister, Christine Wilkes. I am always so impressed by your wisdom, maturity, and adventurousness. To my grandmother, Joan Wilkes—you are an inspiration, and I appreciate all the experiences I have had because of you. To my parents-in-law, Robert and Marianne Kondziolka – I feel your support and love in all our interactions. And to my parents, Roger and Heather Wilkes – you have given me every opportunity. I love and appreciate you both.

Finally, to my husband, John Kondziolka—I cannot thank you enough for all the many ways you have supported me throughout this degree and over the past ten years. I am the luckiest because I have you by my side.

vii

List of tables and figures

Tables

Table 2.1 Experimental conditions and isotopic data for E. huxleyi 20

Table 2.2 Isotopic offsets between cellular pools 27

Table 2.3 Four carbon allocation scenarios consistent with δbiomass and δCAP measurements 33

Table 3.1 Experimental conditions and bulk isotopic data for A. tamarense 52

Table 3.2 Results of compound-specific isotope analysis of A. tamarense lipids 61

Table 4.1 Compiled εRubisCO and εf values for different RubisCO forms 85

Table 4.2 CCM components combined and generalized as active transport mechanisms 89 in the model

Table 5.1. Steady-state algal characteristics under three CO2 and growth rate combinations 122 Table 5.2. Steady-state carbonate system parameters 124 Table A.1 Carbonate system parameters 133

Table A.2 Steady-state cell densities and residual nutrient concentrations 133

Table A.3 Isotopic fractionation between cellular pools 133

Table A.4 Three carbon allocation scenarios associated with different flip values 135

Table B.1 Summary of growth rate and CO2 combinations 140

Table B.2 Summary of the carbonate chemistry parameters and isotope measurements 141

Table B.3 Cell dimensions 142

Table B.4 Experimental design comparison with Hoins et al. (2016b) 142

Table C.1. Assumed kinetic isotope effects used for model calibration 148

- Table C.2. Measured and calculated kinetic isotope effects for the conversion of CO2 to HCO3 148

Table C.3. Model parameter definitions and units 149

Table C.4. Literature data used to test model 155

Table C.5 Input parameters and model outcomes accompanying Figures 4.4-4.7 166

viii

Figures

Figure 1.1. Paired marine sedimentary carbon isotope records 1

Figure 1.2. Summary of factors affecting the magnitudes of εTOC and εp 3

Figure 1.3. εp responses to dissolved CO2 concentrations in chemostat culture experiments 5

Figure 1.4. Small changes in εp result in large changes in pCO2 as εp approaches εf 6

Figure 1.5. Reconstructions of pCO2, εTOC, and εp throughout the Phanerozoic 7

Figure 2.1 13C values of calcite, CAPs, biomass, and alkenones 25

Figure 2.2 Isotopic fractionations (δ values, ‰) between compound classes 26

Figure 2.3 Predicted cellular carbon allocation to major organic compound classes 35

Figure 3.1 εp as a function of growth rate and CO2 concentration, Alexandrium tamarense 57

Figure 3.2 Comparison with eukaryotic species data from Popp et al. (1998a) 58

Figure 3.3 Isotopic fractionations of lipids from three biomarker classes 62

Figure 4.1. Passive diffusion model vs. the revised model 82

Figure 4.2. Comparison of εRubisCO values measured in vitro with εf values determined in vivo 86

Figure 4.3. Model structure 87

Figure 4.4. Behavior of the generalized model 94

Figure 4.5. Modeled vs. measured P values for the diatom P. tricornutum 97

Figure 4.6 Modeled vs. measured P values for the haptophyte E. huxleyi 98

Figure 4.7. Modeled vs. measured P values for Alexandrium dinoflagellate species 99

Figure 4.8. Modeled vs. measured P values for the diatom P. glacialis 100

Figure 4.9. Modeled vs. measured P values for all taxa and conditions (n = 135) 101

Figure 5.1. Photosynthetic carbon isotope fractionation of Micromonas pusilla 126

Figure 5.2. Compilation of nitrate-limited chemostat culture experiments 127

Figure 5.3. Weighted-mean δ13C values of fatty acids derived from glycolipids 129

Figure 5.4. Average εp values estimated for three particle size classes 129

Figure A.1. SDS PAGE gel verifying the presence of a single polysaccharide 136

Figure A.2. Comparison with data from the chemostat incubations of Bidigare et al. (1997) 137

Figure A.3. Alternative cellular carbon allocation in E. huxleyi 138

Figure B.1. The chemostat system and its associated components 143

Figure C.1. Model structure 144

Figure C.2. Alternative model structure 144

Figure C.3. Experimental datasets used for model testing (E. huxleyi, P. tricornutum) 153

Figure C.4. Experimental datasets used for model testing (Alexandrium spp.) 154

Figure D.1. Multisizer cell count agreement with microscopy count 169

Figure D.2. Temporal measurements for Experiment #1 170

Figure D.3. Temporal measurements for Experiment #2 171

Figure D.4. Temporal measurements for Experiment #3 172

Figure D.5. Line P transect 173

Chapter 1

Introduction

13 The paired stable carbon isotope records of marine carbonates (δ Ccarb) and sedimentary organic

13 matter (δ Corg) are commonly used to interpret environmental and evolutionary changes in Earth history

(Hayes et al., 1999; Kuhn, 2007; Pancost et al., 2013; Krissansen-Totton et al., 2015). The average isotopic

13 13 difference between these two carbon reservoirs is described by the quantity εTOC ≈ δ Ccarb - δ Corg (Figure

1.1.; Hayes et al., 1999]). Small variations in the magnitude of εTOC over geologic time have been attributed

to ecological factors, algal physiological factors, and variations in atmospheric CO2 levels (Freeman and

Hayes, 1992; Kump and Arthur, 1999; Hayes et al., 1999; Pagani, 2014). At a broad scale, εTOC values have

1 been surprisingly constant for much of Earth’s history despite major changes in global carbon cycling

(Krissansen-Totten et al., 2015). In order to accurately decipher εTOC in terms of major geological and

biological events, modern relationships must be used to understand the processes influencing εTOC, as well

as the limits of interpretation.

Figure 1.1. Paired marine sedimentary carbon isotope records. δ13C values of bulk organic matter and carbonate are plotted as a function of time, with the modern on the left. Adapted from Krissansen-Totton et al., 2015.

The marine record of εTOC traditionally has been interpreted by analogy with carbon isotope fractionation models that first were developed in the terrestrial biosciences (Farquhar et al., 1982, 1989).

Carbon isotope models of photosynthesis in land plants were adapted for phytoplankton, with minor modifications, based on laboratory studies involving modern marine taxa (Laws et al., 1995; Popp et al.,

1998a,b; Burkhardt et al., 1999a, b; Laws et al., 2001). These studies isolated the factors controlling isotopic fractionation during photosynthesis by measuring the isotopic compositions of dissolved carbon dioxide

13 13 (δ CCO2(aq)) and whole cell biomass (δ Corg) under varying cellular growth conditions. The difference between these two values corresponds to the photosynthetic carbon isotope fractionation (εp) of phytoplankton. Although εp is impossible to measure directly in ancient sediments, because the value of

13 δ CCO2(aq) is not strictly known, εp is implicitly considered in any interpretation of past εTOC records (Pancost et al., 2013; Pagani, 2014). For modern marine taxa, the value of εp has been shown to approach a maximum magnitude of approximately 25‰ (Popp et al., 1998a). This experimentally determined and widely accepted relationship, εp ≤ 25‰, influences both the paleoenvironmental and the evolutionary conclusions drawn from εTOC (Pagani et al., 2011; Pancost et al., 2013; Pagani, 2014).

Critically, this maximum value of 25‰ was obtained from studies involving only three species of unicellular, eukaryotic algae grown under uniform, nitrate-limited chemostat culture conditions (Laws et al., 1995; Bidigare et al., 1997; Popp et al., 1998a). More recent theories that have been developed to explain the cellular-level mechanisms controlling εp also are based on a similar subset of modern phytoplankton

(e.g., Burkhardt et al., 1999a,b; Laws et al., 2001; Cassar et al., 2006; Schulz et al., 2007; Hopkinson et al.,

2011; McClelland et al., 2017). These organisms all employ the same form of the carbon-fixing enzyme

RubisCO—the enzyme thought to be the primary control on marine photosynthetic carbon isotope fractionation. However, RubisCO exists in multiple structurally, catalytically, and phylogenetically distinct forms in phytoplankton: Forms IA, IB, ID, and II (Tabita et al., 2007, 2008). Thus, the current framework for explaining mechanistic controls on εp, and hence εTOC, may not sufficiently account for the complexity of algal communities and their associated enzymatic heterogeneity.

Relationship between εTOC and εp

The magnitude of εTOC is thought to reflect several processes in addition to photosynthesis. It integrates the combined effects of temperature-dependent equilibrium fractionations between dissolved inorganic carbon (DIC) and dissolved CO2 (εHCO3-CO2(aq)), equilibrium fractionations between DIC and carbonate minerals (εppt), and fractionations associated with secondary biological processing (εhet, Figure

1.2; Hayes et al., 1999; Killops and Killops, 2005). Values of εTOC can be converted to approximate values

13 13 of εp by estimating the C content of dissolved CO2 from δ Ccarb, reconstructed sea surface temperatures, and carbonate system isotopic equilibrium, and by assuming εhet is small or zero (Hayes et al., 1999).

Figure 1.2. Summary of factors affecting the magnitudes of εTOC and εp. Adapted from Hayes et al., 1999 and Killops and Killops, 2005.

To minimize issues associated with secondary biological processing or terrestrial inputs, molecular biomarkers known to be derived from restricted taxonomic sources often are analyzed in place of sedimentary total organic carbon (Pagani, 2002; Freeman and Pagani, 2005). This precaution improves the likelihood that the measured isotopic signature reflects the fractionation due to photosynthesis rather than some other process, and by definition removes the uncertainty of εhet. For sediments deposited in the

Cenozoic Era, εp is most often estimated from a class of molecular biomarkers called alkenones, which are produced only by haptophyte algae (reviewed by Pagani, 2014). Newer approaches still largely in the

3 exploratory stage involve analyzing organic matter bound to fossilized components of biomineralizing phytoplankton (e.g., the silica frustules of diatoms or the calcite liths of coccolithophores; Heureux and

Rickaby, 2015; McClelland et al., 2015; Mejía et al., 2017; Stoll et al., 2017). For older sediments lacking these constituents, isotopic analyses of n-alkanes, acyclic isoprenoids, hopanes, and total organic carbon have been employed in various combinations for the purposes of constraining and interpreting sedimentary

εp values (Joachimski et al., 2001; Kuhn, 2007; Pancost et al., 2013).

εp and pCO2

Field and laboratory studies demonstrate that concentrations of ambient dissolved CO2 ([CO2(aq)]) influence the extent of isotopic discrimination against 13C during algal photosynthesis (Rau et al., 1989,

1992; Francois et al., 1993; Laws et al., 1995). When [CO2(aq)] is high, algae discriminate more strongly

13 13 against CO2(aq) and produce organic matter that is relatively depleted in C. The partial pressure of atmospheric CO2 (pCO2) is thought to affect phytoplankton εp through its control of [CO2(aq)] in surface waters. Hence, estimated values of εp from the sedimentary record are used to reconstruct changes in pCO2 through time, an application known as paleobarometry (e.g., Popp et al., 1989; Hayes et al., 1999; Pagani et al., 2011; Zhang et al., 2013; Heureux and Rickaby, 2015).

These approaches for estimating paleo-pCO2 are informed by empirical relationships among growth conditions, carbon dioxide concentrations, and the photosynthetic carbon isotope fractionation of modern alkenone-producing algae. Culture experiments and surveys of marine suspended particulate matter suggest that εp is related to dissolved CO2 concentrations according to the following equation:

b εp = 25‰ − (1) [CO2(aq)] where b is a term representing physiological factors, including growth rate and surface area-to-volume relationships (Laws et al., 1995, 1997; Bidigare et al., 1997; Popp et al., 1998a; Laws et al., 2001). The

εp-intercept of equation (1) corresponds to high ambient concentrations of dissolved CO2 and converges on the maximum value of εp = 25‰ for the three species of eukaryotic microalgae studied prior to this thesis

(Figure 1.3). When algal growth rates () are low or ambient CO2 concentrations are high, the rate-limiting step in photosynthesis is presumed to be carbon fixation by RubisCO, and the isotope effect associated with this process is though to set this maximum value of εP, which is denoted εf (for “fixation”). Until recently, it has been assumed that the value of εf equals the fractionation measured in vitro for RubisCO from higher plants (Roeske and O’Leary 1984; Guy et al., 1993; Scott et al., 2004; McNevin et al., 2006), adjusted slightly for the effects of anaplerotic reactions (-carboxylations; Francois et al., 1993).

The fractionation associated with carbon fixation imposes a fundamental limit on the sensitivity of

εp (and thus εTOC) to variations in pCO2. This limit is apparent when plotting εp directly against pCO2

(Equation 1; Figure 1.4). Since εp asymptotically approaches εf for increasing pCO2, pCO2 levels become prohibitively sensitive to changes in εp and cannot be accurately reconstructed for εp values near εf (small changes in εp result in large changes in pCO2; shaded lines, Figure 1.4). Assuming the relationship in equation (1) holds over geologic timescales, including εf = 25‰, the upper threshold level of sensitivity for paleo-reconstructions is pCO2 ≈ 2200 ppm (Kump and Arthur, 1999; Freeman and Pagani, 2005; Pancost et al., 2013). Similarly, if εp varies, this may be evidence that pCO2 falls below the threshold level of sensitivity (Freeman, 2001; Freeman and Pagani, 2005; Pancost et al., 2013).

Figure 1.3. εp responses to dissolved CO2 concentrations in chemostat culture experiments. (a) At a given [CO2(aq)] and growth rate (μ) the εp associated with photosynthesis differs for three species of eukaryotic phytoplankton. For all three species, εf is 25‰, corresponding to the εp-intercept. (b) The differences in (a) can be normalized by accounting for cell geometries. Figure adapted by Pagani, 2002 from data reported in Popp et al., 1998a.

Figure 1.4. Small changes in εp result in large changes in pCO2 as εp approaches its maximum value of 25‰. Figure adapted from Pancost et al., 2013.

One purpose of this thesis is to investigate what we recognize as the converse of this problem: if εp varies and/or has an estimated value < 25‰ during times when pCO2 levels (as suggested by other proxy approaches) are above the 2200 ppm threshold level of sensitivity (e.g., between ~500-370 Ma (Cambrian through Devonian) and ~170-140 Ma (late Jurassic), Figure 1.5), then the controls on εp require an alternative explanation. The existence of any record where εp < εf while pCO2 is presumed to be > 2200 ppm underscores the need for a more thorough examination of other factors influencing εp and εTOC.

Phytoplankton employ diverse versions of the enzyme RubisCO, as well as carbon concentrating mechanisms (CCMs) that enchance intracellular CO2 relative to the ambient environment. It is essential to understand the significance of these physiological features as controls on εp and εf as they may play important roles in explaining these records.

(a)

(b)

Figure 1.5. (a) pCO2 reconstruction adapted from the GEOCARB III model (Berner and Kothavala, 2001). The dashed red line indicates the sensitivity threshold of 2200 ppm. (b) Values of εTOC and associated values of εp vary throughout the Phanerozoic, including during times when CO2 levels exceed the sensitivity threshold of ~2200 ppm (Hayes et al., 1999). The solid green line corresponds to the accepted maximum fractionation associated with carbon fixation, εf = 25‰. Note several inconsistencies among pCO2, εTOC, and εp that emerge from this figure: from ~500-370 Ma and ~170-140 Ma, εTOC varies by 4 and 5‰, respectively, despite pCO2 exceeding 2200 ppm. Moreover, estimated εp values during these periods are smaller than the assumed εp maximum of 25‰. From ~370-245 Ma, when pCO2 was lower than 2200 ppm, εTOC remains relatively uniform. Thus, the magnitudes of εTOC do not capture the proposed 2000 ppm decline in pCO2 during this time, followed by a 2000 ppm increase thirty million years later.

Taxonomic influences on εf

Hayes et al. (1999) argue based on the structure of the εTOC record (Figure 1.5) that εf ≈ 25‰ over the entire Phanerozoic and justify this conclusion by noting that no primary eukaryotic biomass observed in nature or the laboratory has isotopic compositions requiring εp to be larger than 25‰.

However, from a taxonomic perspective, the relative consistency of εTOC is rather surprising.

Different groups of phytoplankton possess distinct forms of RubisCO (IA, IB, ID, II) which are associated

7 with different kinetic characteristics, isotope fractionations (RubisCO values), and substrate specificities

(Tortell, 2000; Tcherkez et al., 2006; McNevin et al., 2007; Tabita et al., 2008; Young et al., 2016). Thus, it is conceivable that marine phytoplankton species could vary in the extent to which they fractionate carbon

(Roeske and O’Leary, 1985; Goericke et al., 1994; Boller et al., 2011; Boller et al., 2015). This idea is supported by high-precision in vitro studies of RubisCO which show that RubisCO is ~26-30‰ for higher plants (form IB RubisCO; Roeske and O’Leary,1984; McNevin et al., 2007), 22-24‰ for cyanobacteria

(forms IA and IB RubisCO; Guy et al., 1993; Scott and Cavanaugh, 2007), 18-23‰ for proteobacteria

(form II RubisCO; Roeske and O’Leary, 1985; Robinson et al., 2003), and only 11.1‰ and 18.5‰ for E. huxleyi and the diatom Skeletonema costatum, respectively (form ID RubisCO; Boller et al., 2011; Boller et al., 2015).

13 Prior to this thesis, the in vivo C fractionation associated with carbon fixation (εf) had only been studied systematically in continuous culture using two species of diatoms (P. tricornutum and P. glacialis), one species of haptophyte algae (E. huxleyi) and a cyanobacterium (Synechococcus spp.; Laws et al., 1995,

1997; Bidigare et al., 1997; Popp et al., 1998a,b). Although the eukaryotic organisms converged on the consensus εf value of 25‰ (Figure 1.3), the cyanobacterium Synechococcus exhibited a lower εf value of approximately 17‰ (Popp et al., 1998a); this differing behavior for cyanobacteria remains a subject for future study and will not be addressed further here.

Influences of cellular physiology on εP

A central goal of this thesis is to reconcile the apparent uniformity in εf values with the heterogeneity in RubisCO values measured in vitro. The consistent εf value for marine phytoplankton is surprising, if it is related to RubisCO; alternatively, it argues that the value of εf reflects some other process.

A growing body of evidence suggests a variety of physiological factors significantly influence the isotopic compositions of algae. Modern phytoplankton have evolved inorganic carbon concentrating mechanisms (CCMs) to support carbon fixation (Giordano et al., 2005; Reinfelder, 2011; Raven and

Beardall, 2015; Matsuda et al., 2017). CCMs may function by actively transporting bicarbonate or CO2 across the cell membrane or enhancing CO2 levels around the site of RubisCO through catalyzed

- interconversions of HCO3 and CO2(aq) to promote diffusion or accumulation of these inorganic carbon species in cellular substructures (Reinfelder, 2011; Hopkinson, 2014). CCMs may affect εp values in a multitude of ways, for example: (1) enabling phytoplankton to import relatively 13C-enriched bicarbonate through cell membranes, or (2) elevating the intracellular concentration of CO2 compared to the cell’s surroundings. The latter presents a well-known conundrum for pCO2 reconstructions: namely, it predicts that the intracellular environment is disconnected from external [CO2(aq)] (Laws et al., 2002).

In addition to physiological responses to CCMs, variations in growth conditions also are known to influence values of εp. Phytoplankton grown in batch cultures have very different εp responses to growth rate and [CO2(aq)] when compared to chemostat cultures. The primary differences between these two culturing approaches is the resource limiting growth. In chemostat cutures, growth is limited by the delivery of nutrients, while in batch cultures, growth commonly is limited by light. Dilute batch cultures of E. huxleyi under nutrient-replete conditions and variable irradiance yielded a shallower slope for equation (1) and εp values that were up to 8‰ smaller than experiments using nitrate-limited chemostat cultures (Riebesell et al., 2000a; Rost et al., 2002). Similar results were obtained using P. tricornutum (Burkhardt et al., 1999a,b;

Riebesell et al., 2000b). For cells grown in nutrient-replete batch culture, εp values are generally smaller under low light levels or diel irradiance than under high or constant levels of irradiance (Rost et al., 2002;

Pagani, 2014). These results have stimulated speculation that different reliance on carbon acquisition mechanisms is triggered depending on which resource is limiting growth (Riebesell et al., 2000a,b; Cassar et al., 2006; Hoins et al., 2016).

The major outcome of this thesis has been to address this problem, and with it, to propose an answer as to why heterogeneity in RubisCO is not a significant control on εp unless cells are grown under light-limited conditions.

Summary of thesis

This dissertation aims to expand the understanding of phytoplankton photosynthetic carbon isotope fractionation (P) through chemostat culture experiments and modeling approaches, and to evaluate the potential for new paleobarometry (pCO2) proxies derived from algal biomass.

In Chapter 2, I characterize the 13C values of coccolith-asssociated polysaccharides (CAPs) and other cellular carbon pools (bulk biomass, coccolith calcite, and alkenones) of Emiliania huxleyi from chemostat culture experiments. CAPs are involved in regulating calcification in coccolithophores and have been recovered from sedimentary deposits dating back ~180 Ma. It has been proposed that the carbon isotopic contents of CAPs, measured in parallel with coccolith calcite 13C values and alkenone 13C values

(when available), may be used as a novel proxy for ancient atmospheric CO2 levels. I find that CAPs are isotopically enriched by 4.5−10.1‰ relative to bulk biomass, displaying smaller isotopic offsets at faster growth rates and lower CO2 concentrations. These results are used in a simple model to predict cellular carbon allocation to major organic carbon compound classes in E. huxleyi that may account for the isotopic patterns observed across experiments. My findings suggest that CAPs are less sensitive paleobarometry proxies than alkenones, but that CAPs may record a complementary signature for constraining CO2 in the geologic past.

For Chapter 3 I perform nitrate-limited chemostat culture experiments with the marine dinoflagellate Alexandrium tamarense. This species uses an unusual form of the carbon-fixing enzyme:

Form II RubisCO. I find that P values display a robust linear relationship to /[CO2(aq)] with a maximum fractionation (f) value of 27‰. This value is larger than theoretical predicitions for Form II RubisCO and is not significantly different from the f values observed for the more recently-evolved taxa that employ

Form ID RubisCO, e.g, E. huxleyi. I also report the 13C values of dinosterol, hexadecenoic acid, and phytol from each experiment, and discusss my results in the context of a recent proposal to use fossil dinocyst 13C values as a paleobarometry proxy. The major implication of my work is that an f value of 25-27‰ is

10 unlikely to reflect RubisCO, not only in this taxon but perhaps more generally across eukaryotic phytoplankton.

In Chapter 4, I build upon my findings in Chapter 3 to propose a generalized model of carbon isotope fraction (P) in eukaryotic phytoplankton. This model was developed to (i) reconcile apparent contradictions between f as inferred from in vivo studies vs. RubisCO as inferred from in vitro studies, and

(ii) to explain the differences in P that are observed using nutrient-limited chemostat vs. light-limited batch culturing techniques. The model introduces a nutrient- and light-dependent process upstream of RubisCO that is proposed to represent the kinetically slow step in phytoplankton inorganic carbon acquisition. This new model is able to reproduce existing chemostat and batch culture datasets across three prominent eukaryotic phytoplankton groups with an average error of 7.6% (nRMSE).

In Chapter 5, I conclude by reporting preliminary chemostat culture results from CO2 and growth rate manipulation experiments with the picoeukaryotic green algal species Micromonas pusilla.

References

Berner R, Kothavala Z (2001) Geocarb III: A revised model of atmospheric CO2 over Phanerozoic time. American Journal of Science 301:182–204. Bidigare R, Fluegge A, Freeman K, Hanson K, Hayes J, Hollander D, Jasper J, King L, Laws E, Milder J et al. (1997) Consistent fractionation of 13C in nature and in the laboratory: Growth-rate effects in some haptophyte algae. Global Biogeochemical Cycles 11:279–292. Boller A, Thomas P, Cavanaugh C, Scott K (2011) Low stable carbon isotope fractionation by coccolithophore RubisCO. Geochimica et Cosmochimica Acta 75:7200–7207. Boller AJ, Thomas PJ, Cavanaugh CM, Scott KM (2015) Isotopic discrimination and kinetic parameters of RubisCO from the marine bloom-forming diatom, Skeletonoma costatum. Geobiology 13:33−43.

Burkhardt S, Riebesell U, Zondervan I (1999a) Effects of growth rate, CO2 concentration, and cell size on the stable carbon isotope fractionation in marine phytoplankton. Geochimica et Cosmochimica Acta 63:3729–3741. Burkhardt S, Riebesell U, Zondervan I (1999b) Stable carbon isotope fractionation by marine phytoplankton in response to daylength, growth rate, and CO2 availability. Marine Ecology Progress Series 184:31–41. Cassar N, Laws EA, Popp BN (2006) Carbon isotopic fractionation by the marine diatom Phaeodactylum tricornutum under nutrient- and light-limited growth conditions. Geochimica et Cosmochimica Acta 70:5323−5335. Farquhar GD, O’Leary MHB, Berry JA (1982) On the relationship between carbon isotope discrimination and the intracellular carbon dioxide concentration in leaves. Australian Journal of Plant Physiology 9: 121–137. Farquhar GD, Ehleringer JR, Hubick KT (1989) Carbon isotope discrimination and photosynthesis. Annual Revies in Plant Physics 40:503–537. Francois R, Altabet MA, Goericke R, McCorkle DC, Brunet C, Poisson A (1993) Changes in the δ13C of surface water particulate organic matter across the subtropical convergence in the SW Indian Ocean. Global Biogeochemical Cycles 7:627–644. Freeman KH (2001) Isotopic biogeochemistry of marine organic carbon. Reviews in Mineralogy and Geochemistry 43:579–605. Freeman KH, Hayes J (1992) Fractionation of carbon isotopes by phytoplankton and estimates of ancient CO2 levels. Global Biogeochemical Cycles 6:185–198.

Freeman KH, Pagani M (2005) Alkenone-based estimates of past CO2 levels: A consideration of their utility based on an analysis of uncertainties In A History of Atmospheric CO2 and Its Effects on Plants, Animals, and Ecosystems, pp. 35–61. Springer.

Giordano M, Beardall J, Raven JA (2005) CO2 concentrating mechanisms in algae: Mechanisms, environmental modulation, and evolution. Annual Review of Plant Biology 56:99–131. Goericke R, Montoya J, Fry B (1994) Physiology of isotopic fractionation in algae and cyanobacteria In Lajtha K, Michener R, editors, Stable Isotopes in Ecology and Environmental Science, pp. 187– 221. Blackwell Scientific Publications, Oxford.

Guy RD, Fogel ML, Berry JA (1993) Photosynthetic fractionation of the stable isotopes of oxygen and carbon. Plant Physiology 101:37–47. Hayes JM (2001) Fractionation of carbon and hydrogen isotopes in biosynthetic processes. Reviews in Mineralogy and Geochemistry 43:225–277. Hayes JM, Strauss H, Kaufman AJ (1999) The abundance of 13C in marine organic matter and isotopic fractionation in the global biogeochemical cycle of carbon during the past 800 Ma. Chemical Geology 161:103–125.

Heureux A., Rickaby REM (2015) Refining our estimate of atmospheric CO2 across the Eocene−Oligocene climatic transition. Earth and Planetary Science Letters 409:329−338. Hoins M, Eberlein T, Groβmann CH, Brandenburg K, Reichart G-J, Rost B, Sluijs A., Van de Waal DB (2016) Combined effects of ocean acidification and light or nitrogen availabilities on 13C fractionation in marine dinoflagellates. PLoS ONE 11:e0154370.

Hopkinson BM (2014) A chloroplast pump model for the CO2 concentrating mechanism in the diatom Phaeodactylum tricornutum. Photosynthesis Research 121:223−233. Joachimski MM, Ostertag-Henning C, Pancost RD, Strauss H, Freeman KH, Littke R, Sinninghe Damsté JS, Racki G (2001) Water column anoxia, enhanced productivity and concomitant changes in δ13C and δ34S across the Frasnian–Famennian boundary (Kowala—Holy Cross Mountains/Poland). Chemical Geology 175:109–131. Keller K, Morel FMM (1999) A model of carbon isotopic fractionation and active carbon uptake in phytoplankton, Marine Ecological Progress Series 182:295−298. Killops S, Killops V (2005) Introduction to Organic Geochemistry. Blackwell Publishing, Malden, Massachusetts. Krissansen-Totton J, Buick R, Catling DC (2015) A statistical analysis of the carbon isotope record from the Archean to Phanerozoic and implications for the rise of oxygen. American Journal of Science 315: 275−316.

Kuhn T (2007) The evolution of the photosynthetic carbon isotope fractionation εp of marine phytoplankton during the Devonian to Permian time interval Ph.D. diss., University Erlangen-Nuremberg. Kump LR, Arthur MA (1999) Interpreting carbon-isotope excursions: Carbonates and organic matter. Chemical Geology 161:181–198.

Laws EA, Bidigare RR, Popp BN (1997) Effect of growth rate and CO2 concentration on carbon isotopic fractionation by the marine diatom Phaeodactylum tricornutum. Limnology and Oceanography 42:1552–1560. Laws EA, Popp BN, Bidigare RR, Kennicutt MC, Macko SA (1995) Dependence of phytoplankton carbon isotopic composition on growth rate and CO2aq: Theoretical considerations and experimental results. Geochimica et Cosmochimica Acta 59:1131–1138. Laws EA, Popp BN, Bidigare RR, Riebesell U, Burkhardt S, Wakeham SG (2001) Controls on the molecular distribution and carbon isotopic composition of alkenones in certain haptophyte algae. Geochemistry, Geophysics, Geosystems 2.

Laws EA, Popp BN, Cassar N., Tanimoto J (2002) 13C discrimination patterns in oceanic phytoplankton: likely influence of CO2 concentrating mechanisms, and implications for palaeoreconstructions. Funct. Plant. Biol. 29, 323−333. Matsuda Y, Hopkinson BM, Nakajima K, Dupont CL, Tsuji Y (2017) Mechanisms of carbon dioxide acquisition and CO2 sensing in marine diatoms: a gateway to carbon metabolism. Philosophical Transactions of the Royal Society of London Series B. 372:20160403. McClelland, HLO, Hermoso M, Bruggeman J, Rickaby REM (2015). Towards a unifying theory for carbon isotopic partitioning in coccolithophores: Implications for paleo-proxies. Goldschmidt Abstracts, 2069. McClelland HLO, Bruggeman J, Hermoso M, Rickaby REM (2017) The origin of carbon isotope vital effects in coccolith calcite. Nature Communications 8:10.1038/ncomms14511. McNevin DB, Badger MR, Kane HJ, Farquhar GD (2006) Measurement of (carbon) kinetic isotope effect by Rayleigh fractionation using membrane inlet mass spectrometry for CO2-consuming reactions. FunctionalPlant Biology 33:1115−1128. McNevin DB, Badger MR, Whitney SM, von Caemmerer S, Tcherkez GGB, Farquhar GD (2007) Differences in carbon isotope discrimination of three variants of D-Ribulose-1,5-bisphosphate Carboxylase/Oxygenase reflect differences in their catalytic mechanisms. The Journal of Biological Chemistry 282: 36068−36076. Mejía LM, Méndez-Vicente A, Abrevaya L, Lawrence KT, Ladlow C, Bolton C, Cacho I, Stoll H. (2017) A diatom record of CO2 decline since the late Miocene. Earth and Planetary Science Letters 479, 18−33.

Pagani M (2002) The alkenone-CO2 proxy and ancient atmospheric carbon dioxide. Philosophical Transactions of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences 360:609–632.

Pagani M. (2014) Biomarker-based inferences of past climate: The alkenone pCO2 Proxy. In Treatise on Geochemistry 2nd Ed. (eds. H. Holland and K. Turekian). pp. 361-378, Elsevier, Amsterdam. Pagani M, Huber M, Liu Z, Bohaty SM, Henderiks J, Sijp W, Krishnan S, DeConto RM (2011) The role of carbon dioxide during the onset of Antarctic glaciation. Science 334:1261–1264. Pancost RD, Freeman KH, Herrmann AD, Patzkowsky ME, Ainsaar L, Martma T (2013) Reconstructing Late Ordovician carbon cycle variations. Geochimica et Cosmochimica Acta 105:433–454. Popp B, Takigiku R, Hayes J, Louda J, Baker E (1989) The post-Paleozoic chronology and mechanism of 13C depletion in primary marine organic matter. American Journal of Science 289:436–454. Popp BN, Laws EA, Bidigare RR, Dore JE, Hanson KL, Wakeham SG (1998a) Effect of phytoplankton cell geometry on carbon isotopic fractionation. Geochimica et Cosmochimica Acta 62:69–77. Popp BN, Kenig F, Wakeham SG, Laws EA, Bidigare RR, (1998b) Does growth rate affect ketone unsaturation and intracellular carbon isotopic variability in Emiliania huxleyi? Paleoceanography 13:35–41.

Rau GH, Takahashi T, Des Marais D (1989) Latitudinal variations in plankton C: Implications for CO2 and productivity in past oceans. Nature 341:165.

Rau GH, Takahashi T, Des Marais DJ, Repeta DJ, Martin JH (1992) The relationship between 13C of organic matter and [CO2(aq)] in ocean surface water: Data from a JGOFS site in the northeast Atlantic Ocean and a model. Geochimica et Cosmochimica Acta 56:1413–1419.

Raven JA, Beardall J (2015) The ins and outs of CO2. Journal of Experimental Botany67, 1-13. Reinfelder JR (2011) Carbon concentrating mechanisms in eukaryotic marine phytoplankton. Marine Science 3.

Riebesell U, Revill AT, Holdsworth DG, Volkman JK (2000a) The effects of varying CO2 concentration on lipid composition and carbon isotope fractionation in Emiliania huxleyi. Geochimica et Cosmochimica Acta 64:4179–4192. Riebesell U, Burkhardt S, Dauelsberg A, Kroon B (2000b) Carbon isotope fractionation by a marine diatom: Dependence on the growth-rate-limiting resource. Marine Ecology Progress Series 193:295–303. Robinson JJ, Scott KM, Swanson ST, O Leary MH, Horken K, Tabita FR, Cavanaugh CM (2003) Kinetic isotope effect and characterization of form II RubisCO from the chemoautotrophic endosymbionts of the hydrothermal vent tubeworm Riftia pachyptila. Limnology and Oceanography 48:48–54. Roeske C, O’Leary MH (1984) Carbon isotope effects on enzyme-catalyzed carboxylation of ribulose bisphosphate. Biochemistry 23:6275–6284. Roeske C, O’Leary MH (1985) Carbon isotope effect on carboxylation of ribulose bisphosphate catalyzed by ribulose bisphosphate carboxylase from Rhodospirillum rubrum. Biochemistry 24:1603–1607. Romanek CS, Grossman EL, Morse JW (1992) Carbon isotopic fractionation in synthetic aragonite and calcite: Effects of temperature and precipitation rate. Geochimica et Cosmochimica Acta 56:419– 430. Rost B, Zondervan I, Riebesell U (2002) Light-dependent carbon isotope fractionation in the coccolithophorid Emiliania huxleyi. Limnology and Oceanography 47:120–128.

Scott KM, Cavanaugh CM (2007) CO2 uptake and fixation by endosymbiotic chemoautotrophs from the bivalve Solemya velum. Applied and Environmental Microbiology 73:1174–1179. Stoll HM, Mendez-Vicente A, Abrevaya L, Anderson R, Rigual-Hernández AS, Gonzalez-Lemos S. (2017) Growth rate and size effect on carbon isotopic fractionation in diatom-bound organic matter in recent Southern Ocean sediments. Earth and Planetary Science Letters 457, 87−99. Tabita FR, Hanson TE, Li H, Satagopan S, Singh J, Chan S (2007) Function, structure, and evolution of the RubisCO-like proteins and their RubisCO homologs. Microbiology and Molecular Biology Reviews 71, 576−599. Tabita FR, Satagopan S, Hanson TE, Kreel NE, Scott SS (2008) Distinct form I, II, III, and IV Rubisco proteins from the three kingdoms of life provide clues about Rubisco evolution and structure/function relationships. Journal of Experimental Botany 59, 1515-1524. Tcherkez GGB, Farquhar GD, Andrews TJ (2006) Despite slow catalysis and confused substrate specificity, all ribulose bisphosphate carboxylases may be nearly perfectly optimized. Proceedings of the National Academy of Sciences USA 103, 7246−7251. Young JN, Heureux AMC, Sharwood RE, Rickaby REM, Morel FMM, Whitney SM (2016) Large variation in the Rubisco kinetics of diatoms reveals diversity among their carbon-concentrating mechanisms. Journal of Experimental Botany 67:3445−3456.

Zhang YG, Pagani M, Liu Z, Bohaty SM, DeConto R (2013) A 40-million-year history of atmospheric CO2. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 371:20130096.

Chapter 2 Carbon isotope ratios of coccolith-associated polysaccharides of Emiliania huxleyi as a function of growth rate and CO2 concentration This chapter has been published: Wilkes E. B., McClelland H. L. O., Rickaby R. E. M., and Pearson A. (2018) Carbon isotope ratios of coccolith-associated polysaccharides of Emiliania huxleyi as a function of growth rate and CO2 concentration. Organic Geochemistry, 119, 1−10.

Abstract

The calcite plates, or coccoliths, of haptophyte algae including Emiliania huxleyi are formed in intracellular vesicles in association with water-soluble acidic polysaccharides. These coccolith-associated polysaccharides (CAPs) are involved in regulating coccolith formation and have been recovered from sediment samples dating back to ~180 Ma. Paired measurements of the carbon isotopic compositions of

CAPs and coccolith calcite have been proposed as a novel paleo-pCO2 barometer, but additional proxy validation and development are still required. Here we present culture results quantifying carbon isotopic offsets between CAPs and other cellular components: bulk organic biomass, alkenones, and calcite. E. huxleyi was grown in nitrate-limited chemostat experiments at growth rates (µ) of 0.20–0.62/d and carbon dioxide concentrations of 10.7–17.6 µmol/kg. We find that CAPs are isotopically enriched by 4.5−10.1‰ relative to bulk organic carbon, exhibiting smaller isotopic offsets at faster growth rates and lower CO2 concentrations. This variability suggests that CAPs record a complementary signature of past growth conditions with different sensitivity than alkenones or coccolith calcite. By measuring the isotopic compositions of all three molecules and minerals of self-consistent origin, the ratio µ/[CO2(aq)] may be reconstructed with fewer assumptions than current approaches.

Introduction

Emiliania huxleyi is a cosmopolitan, bloom-forming marine algal species that uses dissolved inorganic carbon (DIC) for calcification and photosynthesis (Westbroek, 1993). It is the dominant coccolithophore in modern oceans, capable of producing calcifying plates (“coccoliths”) that interlock around the cell (De Vargas et al., 2007, Henriksen and Stipp, 2009). The stable carbon and oxygen isotopic

17 compositions of fossilized coccoliths have been used to reconstruct paleoclimatic and evolutionary events

(Stoll, 2005, Rickaby et al., 2007, Hermoso et al., 2009, Bolton and Stoll, 2013). E. huxleyi and its ancestors within the family Noëlaerhabdaceae also synthesize long-chain unsaturated ketones called alkenones that are preserved in the geologic record and used as paleotemperature and paleobarometry (pCO2) proxies (e.g.,

Volkman et al., 1980, Marchal et al., 2002, Pagani et al., 2005, Zhang et al., 2013, Brassell, 2014). Although

E. huxleyi only became dominant in the fossil record around 70 ka, alkenones first appear in Cretaceous sediments, and fossil coccoliths have been dated to the Late Triassic (Thierstein et al., 1977, Farrimond et al., 1986, Bown et al., 1987).

Coccolith precipitation occurs intracellularly in vesicles that maintain a controlled chemical composition (Henriksen and Stipp, 2009). Coccolithogenesis begins with the formation of precursor organic templates that provide a framework of binding sites for crystal nucleation and growth (Young et al., 2003).

The calcite crystals grow to form complex units, with mineral expression regulated by acidic polysaccharides (Marsh, 2003). Completed coccoliths and coccolith-associated polysaccharides (CAPs) trapped within ultimately are expelled to the outside of the cell. E. huxleyi contains one type of CAP, consisting of a polymeric mannose backbone with sidechains of galacturonic acid and ester-bound sulfate groups (De Jong et al., 1976, Fichtinger-Schepman et al., 1981, Kok et al., 1986). Another coccolithophore species, Pleurochrysis carterae, contains three types of CAPs, with both galacturonic and glucuronic moieties (Marsh et al., 1994, Marsh et al., 2002). CAPs interact with the carbonate chemistry of the coccolith vesicle through the carboxyl groups of the uronic acid residues, which can shed protons to preferentially bind calcium cations (Borman et al., 1982).

CAPs are increasingly being used to study the interplay between the ambient environment and the carbonate chemistry of intracellular carbon pools (Henriksen and Stipp, 2009, Lee et al., 2016, Rickaby et al., 2016). Lee et al. (2016) extracted intact CAPs from both modern cultures and fossil coccoliths dating back to ~180 Ma. The uronic acid contents of these extracts correlated with the predicted internal saturation state of the coccolith vesicle in modern cultures and approximately tracked Phanerozoic pCO2 reconstructions obtained from other paleo-proxies (Lee et al., 2016).

13 The stable carbon isotopic composition of CAPs ( CCAP) may provide further complementary

13 13 information.  CCAP values, measured in conjunction with coccolith calcite  C values, have been proposed as a novel paleobarometer for ancient pCO2 (Hermoso, 2014, McClelland et al., 2017). Such

13 13 application assumes that  CCAP values predictably track  Cbiomass values and can be used to calculate the isotope fractionation accompanying photosynthesis (P values; Freeman and Hayes, 1992). Preliminary

13 work shows that  CCAP values are measurable (R.B.Y. Lee, unpublished, McClelland et al., 2015), but no

13 paired measurements of  CCAP values and bulk organic carbon have been reported. In this study we measure the 13C values of CAPs, bulk cellular organic carbon, calcite, and alkenones of E. huxleyi grown in nitrate-limited chemostat cultures. Results are analyzed as a function of growth rate and CO2 availability, with all other culture conditions held constant.

Materials and methods

Chemostat culture methods

A coccolith-bearing strain of E. huxleyi (CCMP3266, isolated from the South Pacific in 1998) was grown in a nitrate-limited chemostat at a constant temperature of 18 °C. Experimental conditions were selected to mimic the cultures of Bidigare et al. (1997) and employed the chemostat system described in

Wilkes et al. (2017). Cool-white fluorescent light was supplied continuously with a saturating photon flux density of ~150 µmol photons/m2s (400–700 nm radiation). The vessel was stirred at 50 rpm. The growth medium consisted of 0.2 µm-filtered and autoclaved Gulf of Maine natural seawater enriched with metals and vitamins according to L1 medium (Guillard and Hargraves, 1993). Initial nitrate and phosphate concentrations were adjusted to approximately 100 µM and 36 µM, respectively. Four different growth rates (µ; 0.20–0.62/d; Table 2.1) were achieved by adjusting the dilution rate.

Table 2.1. Experimental conditions and isotopic data for E. huxleyia,b

Expt [CO ] 2(aq) µc (/d) δ13C (‰) δ13C (‰) δ13C (‰) δ13C (‰) # (µmol/kg) calcite biomass CAP alkenone

1 17.6 ± 1.2 0.20 ± 0.01 –15.44 ± 0.01 –48.2 ± 1.4 –38.2 ± 0.2 –51.5 ± 1.0 2 14.0 ± 0.4 0.40 ± 0.01 –12.90 ± 0.01 –45.7 ± 0.1 –37.3 ± 0.5 –49.8 ± 0.9 3 11.9 ± 0.5 0.48 ± 0.01 –10.44 ± 0.01 –42.9 ± 0.3 –35.4 ± 0.5 –46.4 ± 0.5 4 10.7 ± 0.6 0.62 ± 0.01 –8.26 ± 0.01 –39.5 ± 0.9 –34.9 ± 0.5 –42.4 ± 0.5 aValues reflect the mean of steady-state conditions, ± 1σ. Sampling of cells for isotopic analysis was not begun until at least four doublings were completed at a given dilution rate. b The order in which the experiments were performed is: 2, 3, 1, 4. cValues and associated errors correspond to the dilution rates applied to the chemostat system, equaling  at steady- state.

Cell densities were monitored daily by OD600 and by cell counts using a hemocytometer counting chamber and a light microscope, yielding reasonable correlation between approaches (r2=0.72, all experiments). Residual nitrate and phosphate concentrations were determined spectrophotometrically on

0.22 µm-filtered and refrigerated samples using the resorcinol (Zhang and Fischer, 2006) and mixed molybdate (Strickland and Parsons, 1968) methods, respectively. Daily sample removal never exceeded

4% of the culture volume to minimize perturbations to steady-state conditions. Cell size measurements were not performed in this study. By analogy with other nitrate-limited chemostat studies of E. huxleyi over relevant CO2 ranges, we assume that cell diameter changed minimally between experiments and is not a primary control on our isotopic results (Popp et al.,1998a; Müller et al., 2012).

Carbonate system chemistry

Four CO2 concentrations from 10.7–17.6 µmol/kg (Table 2.1) were maintained by bubbling with

13 mixtures of tank CO2 ( CCO2 = –38.58 ± 0.03‰) and 4:1 N2:O2. pH was monitored continuously using an in-process pH probe (EasyFerm Plus, Hamilton) and ranged from 8.0−8.2 across all four experiments. Total dissolved inorganic carbon (DIC) and alkalinity samples were taken daily during the steady-state phase of each experiment (final 3–6 days). Samples for DIC were collected without headspace, poisoned with 0.2

20 w/w % sodium azide, and stored in darkness at 4 °C. DIC was converted to CO2 by acidification with

H3PO4, purified on a vacuum line, and quantified manometrically, yielding a range from 1740−1880

µmol/kg. Total alkalinity was determined by Gran titration with 0.01 N HCl solution prepared in a 0.7 M

NaCl background (Gran, 1952, Dickson et al., 2007), using certified reference materials supplied by A.G.

Dickson (Scripps Institution of Oceanography) to monitor precision. The carbonate system was calculated from DIC, pH, phosphate, temperature, and salinity using CO2SYS (Lewis and Wallace, 1998, van Heuven et al., 2011, Table A.1) and the dissociation constants of Mehrbach et al. (1973), as refitted by Dickson and

Millero (1987), and Dickson (1990). The combined uncertainties in calculated inorganic carbon speciation were estimated numerically following Bevington and Robinson (2003).

Isotopic analysis

13 Cells were pelleted by centrifugation and stored at –80 °C. Values of  Cbiomass were measured on thawed, acidified by wet HCl addition (1N), and dried (60 °C) samples using an elemental analyzer interfaced to a continuous flow isotope ratio mass spectrometer (EA-IRMS; UC Davis Stable Isotope

Facility). Alkenone 13C values were measured from total lipid extracts (Bligh and Dyer, 1959) using gas chromatography-isotope ratio mass spectrometry (GC-IRMS: Thermo Scientific Delta V Advantage interfaced to a Trace GC Ultra via a GC Isolink). Alkenones were separated on a 60-m DB1-MS capillary column using the following oven ramp program: ramp from 65–110 °C at 40 °C/min and hold 2 minutes, ramp to 270 °C at 40 °C/min, ramp to 320 °C at 2 °C/min and hold 36 minutes. Values of 13C were measured and averaged for C37:3 and C37:2 alkenones (peak sizes, 0.5–5V, m/z 44; peak areas 8–131 Vs) using n-C32, n-C38, and n-C41 alkane external standards.

CAPs were isolated from freeze-dried biomass pellets according to the protocol described in Lee et al. (2016). Briefly, cells were cleaned with 1% v/v Triton X-100, 4.5% v/v NaOCl in 0.05 M NaHCO3.

The coccoliths were centrifuged through a gradient of Ludox TM-50 colloidal silica (Sigma-Aldrich) layered with 20% w/v sucrose. After additional rinses with NH4HCO3, the pellet was decalcified with 0.5

M EDTA (pH 8.0, 12 h) and sonicated. Insoluble residues were removed by centrifugation and the supernatant was diafiltered with an Amicon Ultra-4 centrifugal filter unit to remove all molecules less than

14 kDA (Millipore). The CAP was isolated by anion exchange liquid chromatography using a HiTrap

DEAE FF (GE Healthcare). CAP extracts were subjected to sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and PAGE, followed by staining with Alcian Blue to verify the presence of

CAP and its existence as a single polysaccharide (Figure. A.1). CAP identity was also confirmed through reverse-phase high-performance liquid chromatography (RP-HPLC) with apple pectin (polygalacturonic acid) as a positive control. Extracts were shown to be free of any contaminating proteins by performing a

Bradford assay and by staining PAGE gels with Coomassie Blue. CAP 13C values were analyzed by spooling wire micro-combustion isotope ratio mass spectrometry (SWiM-IRMS, Harvard University) using a pectin standard; full process blanks were assessed using both MilliQ water and the pectin standard.

13  C values of DIC were measured at Woods Hole Oceanographic Institution on CO2 gas samples collected in Pyrex tubes following purification and DIC quantification on a vacuum line. Carbon isotope compositions of coccolith calcite were measured using a Thermo Delta V Advantage isotope ratio mass spectrometer fitted to a Kiel IV carbonate device (University of Oxford). All measured isotope values are reported on the international V-PDB scale. Carbon isotope fractionations between different cellular

13 13 constituents are reported as simple linear differences ( (A-B) =  CA– CB) and as epsilon values

13 13 (following Freeman and Hayes, 1992, Zeebe and Wolf-Gladrow, 2001): (A-B) = ( CA– CB)/(1+

13 3  CB/10 ).

Results

Chemostat Cultures

Each experiment reached steady state for all measured variables 7–12 days after setting the initial conditions (dilution rate and CO2 combination). Multiple samples from the stable steady state were collected over the subsequent 3–6 days, treated as replicates, and averaged. Final cell densities were

22 approximately 3 x106 cells/mL across experiments (Table A.2), as expected based on the low half-saturation constant for nitrate (KM= 0.35 M, Perrin et al. 2016) relative to the initial concentration in the growth medium (100 M). Routine monitoring of residual nitrate concentrations indicated at least a 25-fold reduction from the concentration in the feed media, yielding residual molar N:P ratios of less than 0.2 in all experiments (Table A.2). Dissolved CO2 concentrations (Table 2.1) were within the ranges of both the modern ocean (~8–30 µmol/kg) and prior chemostat culture investigations of E. huxleyi (9.6–274.1

µmol/kg; Bidigare et al., 1997, Popp et al., 1998a,b). Growth rates (µ, Table 2.1) were between 16 and 51% of the maximum growth rates achieved in nutrient-replete batch cultures (max =1.22/d, Hermoso et al.,

2016a).

Carbon isotopic compositions of cellular constituents and DIC

Coccolith calcite was the most 13C-enriched cellular component, with 13C values ranging from -15.4 to –8.3‰ (Table 2.1). CAPs were the most enriched organic component (–39.5 to –38.2‰), followed by bulk biomass (–48.2 to –39.5‰) and alkenones (–51.5 to –42.4‰) (Table 2.1). The strongly

13 negative absolute values of  C in all carbon pools reflect the influence of the tank CO2 (–38.58 ± 0.03‰) used to adjust the seawater carbonate chemistry.

13  CDIC values were measured for five consecutive sampling days at steady state during Experiment

13 #1, with a mean value of  CDIC = –18.3 ± 0.2‰. The isotopic composition of the dissolved CO2 (–27.8‰)

13 was calculated from  CDIC and carbonate speciation (Zeebe and Wolf-Gladrow, 2001), using the fractionation factors of Mook et al. (1974) and Zhang et al. (1995). Unfortunately, due to a technical

13 malfunction, δ CDIC measurements are unavailable for the remaining three experiments. However, the measurements from Experiment #1 confirm that the fractionations expressed between DIC, biomass, and coccolith calcite are consistent with literature ranges. Photosynthetic carbon isotope fractionation (εP) was

21.4 ± 1.5‰, calculated relative to CO2 as the inorganic carbon source, or 31.4‰ calculated relative to total

DIC. This result agrees with Bidigare et al. (1997; Figure A.2a), and falls within theoretical bounds

23 established for eukaryotic phytoplankton (Goericke et al, 1994, Popp et al., 1998a; Wilkes et al., 2017).

Interestingly, coccolith calcite is enriched by 2.9‰ relative to DIC in Experiment #1 (and 1.9‰ relative to thermodynamic predictions for inorganically-precipitated calcite, Romanek et al., 1992). This nearly 3‰

13 enrichment is at the upper end of the range reported from prior batch cultures of E. huxleyi ( Ccalcite-DIC

= –4.2 to 3‰; Rost et al., 2002, Ziveri et al., 2003, Hermoso et al., 2016b, Katz et al., 2017, McClelland et al., 2017), likely reflecting differences in nutrient conditions and carbonate system manipulation between studies (see Discussion).

Isotopic sensitivities to changing growth rate and [CO2(aq)]

The 13C values of bulk cellular biomass, calcite, and alkenones are influenced by the ratio

µ/[CO2(aq)] (Bidigare et al., 1997, Popp et al., 1998a,b, Riebesell et al., 2000, Rost et al., 2002, Tchernov et al., 2014, Hermoso et al., 2016b, Holtz et al., 2017). These three cellular components, as well as CAPs,

2 display strong linear relationships with respect to µ/[CO2(aq)] in this study (r > 0.90, p < 0.05, Figure 2.1a;

Figure A.2b). The 13C values for each cellular pool grow increasingly 13C-enriched with increasing

13 13 µ/[CO2(aq)], although not all pools have the same slope (Figure 2.1a).  Cbiomass and  Calkenone values display the greatest sensitivity to µ/[CO2(aq)], with nearly identical slopes of 190 and 202 (‰ µmol d)/kg,

13 respectively.  Ccalcite values display slightly less sensitivity with a slope of 157, which corresponds to a

13 13 25% slope difference relative to  Cbiomass.  CCAP values respond least sensitively to µ/[CO2(aq)], with a slope of 75 (‰ µmol d)/kg. When the 13C values of alkenones and CAPs are compared to the bulk phases

(calcite and biomass) the slopes of the cross-plots are approximately 1 and 0.5, respectively (Figure 2.1b,c).

The differences in isotopic sensitivities to µ/[CO2(aq)] are apparent by examining  values, the

13 13 isotopic differences between any two cellular carbon pools (Figure 2.2).  Ccalcite-CAP and  CCAP-biomass

13 values both display robust linear relationships: with increasing µ/[CO2(aq)],  Ccalcite-CAP values increase

13 from 22.7 to 26.7‰ (Figure 2.2a) while  CCAP-biomass values decrease from 10.1 to 4.5‰ (Figure 2.2b).

13 13 These opposing trends are also evident in Figure 2.1:  CCAP values approach  Cbiomass values and diverge

13 from  Ccalcite values at the limit of faster growth or lower [CO2(aq)].

Figure 2.1. The 13C values (‰) of coccolith calcite (black triangles), CAPs (grey circles), bulk biomass (white squares), and alkenones (black diamonds) (a) as a function of the ratio of growth rate (µ) to CO2 13 13 13 concentration ([CO2(aq)]), (b) comparing δ Calkenone values to  Cbiomass and  Ccalcite values, with slopes of 13 13 13 approximately 1, and (c) comparing δ CCAP values to  Cbiomass and  Ccalcite, showing slopes of approximately 0.5.

Figure 2.2. Isotopic fractionations (δ values, ‰) between compound classes, calculated as simple linear 13 13 13 differences and plotted versus /[CO2(aq)]: (a) δ Ccalcite-CAP, (b) δ C CAP-biomass, (c) δ Cbiomass-alkenone, (d) 13 δ Ccalcite-biomass. Error bars represent ±1σ propagated error. Note: all vertical axes span 12‰ ranges, but with different values.

13 Consistent with prior work,  Cbiomass-alkenone values are constant within error across all four experiments, with a mean value of 3.5 ± 0.5‰ (Table 2.2), and display no significant linear dependence on

2 µ/[CO2(aq)] (r = 0.20, p > 0.05, Figure 2.2c). Prior E. huxleyi chemostat cultures under analogous conditions

13 (µ = 0.2–0.6/d, 18 C, nitrate-limited) similarly exhibited  Cbiomass-alkenone values (3.5–5.2‰) with no

13 clear dependence on growth rate (Popp et al., 1998b).  Ccalcite-biomass values also do not display a

26 statistically significant response to µ/[CO2(aq)] (p > 0.05, Figure 2.2d; mean 32.3 ± 0.7‰). By analogy,

13 therefore, the difference  Ccalcite-alkenone would also be statistically constant over the range of the experiments; and collectively, the  values indicate that biomass, alkenone, and calcite 13C values all

13 effectively respond similarly to µ/[CO2(aq)] (Figures 1, 2). Only  CCAP values have a different µ/[CO2(aq)] sensitivity.

Table 2.2. Isotopic offsets between cellular pools (‰).a Expt Δδ13C Δδ13C Δδ13C Δδ13C Δδ13C # calcite-CAP CAP-biomass biomass-alkenone calcite-biomass alkenone-CAP

1 22.7 ± 0.2 10.1 ± 1.4 3.3 ± 1.7 32.8 ± 1.4 –13.4 ± 1.0 2 24.4 ± 0.5 8.4 ± 0.5 4.2 ± 0.9 32.8 ± 0.1 –12.6 ± 1.1 3 24.9 ± 0.5 7.5 ± 0.6 3.5 ± 0.6 32.5 ± 0.3 –11.0 ± 0.7 4 26.7 ± 0.5 4.5 ± 1.0 2.9 ± 1.0 31.2 ± 0.9 –7.4 ± 0.7

Mean 24.7 8 3.5 32.3 –11.1 SD 1.6 2 0.5 0.7 2.6 aValues are ± 1σ propagated error.

Maximum isotopic offsets between substrates and carbon pools occur at the limit of infinite

13 [CO2(aq)] or zero growth (Goericke et al., 1994, Laws et al., 1995). For example, P values (≈  CCO2-biomass) have an expected intercept of approximately 25‰, inferred from prior chemostat cultures of eukaryotic algae including E. huxleyi (Popp et al., 1998a). This value has been interpreted as the maximum

13 fractionation accompanying enzymatic carbon fixation (f). Although we were unable to measure  CDIC values for three of our experiments and thus cannot directly solve the εP = εf – µ/[CO2(aq)] equation in this study, the analogous intercept values for other carbon pools may provide insight into the inherent isotope

13 effects of the governing reactions.  Ccalcite-CAP approaches an intercept of approximately 22‰,

13 13 13  CCAP-biomass approaches 11‰,  Cbiomass-alkenone approaches 4‰, and  Ccalcite-biomass approaches 33‰.

Discussion

Intracellular carbon isotope patterns – bulk classes

The isotopic ordering among compound classes matches general expectations for photoautotrophs.

Organic carbon is 13C-depleted relative to inorganic carbon (DIC and coccolith calcite) in our samples due

13 to kinetic discrimination against CO2 during photosynthesis (e.g., Goericke et al., 1994). Within the organic carbon pools, typical intracellular isotopic ordering also prevails: carbohydrates are more enriched in 13C than total cellular biomass, while lipids are 13C-depleted (Wong et al., 1975, Sakata et al., 1997,

13 13 13 Hayes, 2001). This consistent isotopic pattern ( CCAP >  Cbiomass >  Calkenone) can be explained by a biosynthetic reaction network in which simple carbohydrates, used to synthesize CAPs, are the first compounds generated, while downstream kinetic processes discriminate against 13C and produce relatively

13C-depleted products including alkenones (Hayes, 2001).

13 All  values (Figure 2.2) are also consistent with expectations. Our mean  Cbiomass-alkenones value of 3.5 ± 0.5‰ (equivalent to biomass-alkenones = 3.7‰; Table A.3) is within error, although on the lower end, of the 3.8–4.2‰ corrections that have been implemented in paleobarometry studies (Jasper and Hayes,

1990, Jasper et al., 1994, Pagani et al., 2005, Bijl et al., 2010, Seki et al., 2010) and is typical of the expressed fractionation for acetogenic lipids relative to total biomass (e.g., Hayes, 2001).

Paired measurements of calcite and biomass 13C values from the same study are rare. Our mean

13  Ccalcite-biomass value of 32.3 ± 0.7‰ (equivalent to calcite-biomass = 33.8 ± 0.9‰; Table A.3) exceeds the

13 values measured in two nutrient-replete batch culture studies of E. huxleyi ( Ccalcite-biomass =15.3–28.8‰,

Rost et al., 2002, McClelland et al., 2017), but the difference must partially reflect the larger P values obtained in nitrate-limited chemostats (P = 21.4‰, Experiment #1;17.2–24.9‰ Bidigare et al., 1997,

Figure A.2a) relative to those from the batch cultures (P = 6.7–17.1‰, Rost et al., 2002, McClelland et al.,

2017). The larger fractionations in our study may also result from the continuous, relatively high light conditions employed (associated with larger P values; Rost et al., 2002, Holtz et al., 2017) and differences in carbonate system manipulation between studies (McClelland et al., 2017).

No prior investigations have characterized the carbon isotope fractionation between CAPs and other cellular pools; and interestingly, the  values we observe for CAPs appear to be different from expectations for bulk carbohydrates. Bulk carbohydrates can be up to 3–4‰ enriched relative to total biomass for photoautotrophs, but more commonly are thought to be on average 1–2‰ enriched (Abelson and Hoering, 1961, Coffin et al., 1990, Macko et al., 1990, Hayes, 2001). However, analyses of individual monosaccharides indicate substantial isotopic heterogeneity within the bulk carbohydrate pool, with some

13 13 being significantly enriched in C (van Dongen et al., 2002, Teece and Fogel, 2007). Our  CCAP-biomass values (4.5−10.1‰, Table 2.2) fall within ranges observed for individual monosaccharides in marine and freshwater algae (0–9‰ for unicellular algae, and up to 13‰ for the macroalgal species Ulva lactuca; van

Dongen et al., 2002, Teece and Fogel, 2007). Teece and Fogel (2007) measured 13C values for glucose and galactose from a batch culture of E. huxleyi, finding enrichments of 5 and 7‰ relative to biomass,

13 respectively, and leading to predictions of  CCAP-biomass values ~6‰ (Benthien et al., 2007, Boller et al.,

2011). Yet the most prevalent monomers within the CAP structure (mannose, galacturonic acid, and rhamnose; Fichtinger-Schepman, 1981) have not to date been isotopically characterized for this species. In a field specimen of Ulva lactuca, these monomers were 13C-enriched relative to biomass by 10‰

(mannose), 13‰ (galactose), and 7‰ (rhamnose) (Teece and Fogel, 2007). Because our

13  CCAP-biomassvalues of 4.5–10.1‰ are broadly consistent with these existing measurements, it is likely

13 13 that  CCAP values are more C-enriched than the mass-weighted average carbohydrate composition of the cells, perhaps to an even greater extent than predicted by Benthien et al. (2007) and Boller et al. (2011).

Linear isotopic responses to changing µ/[CO2(aq)]

13 The linear increase in  Cbiomass values with increasing µ/[CO2(aq)] (Figure 2.1) is consistent with predictions from simple algal models invoking primarily diffusive entry of CO2 into the cell for photosynthetic fixation (e.g., Freeman and Hayes, 1992, Laws et al., 1995). Bidigare et al. (1997) showed

13 similar linear increases in δ Cbiomass values under analogous chemostat growth conditions (Figure A.2b).

13 While it is not necessary to invoke non-diffusive carbon supply to explain either chemostat study, δ Cbiomass values cannot be used to rule out the use of carbon concentrating mechanisms (CCMs) in E. huxleyi. Laws

13 et al. (2002) hypothesized that the δ Cbiomass values of E. huxleyi should become insensitive to the ratio

-1 µ/CO2 at values exceeding ~0.1 kg/(mol d ) if an inducible carbon concentrating mechanism (CCM) is employed (assuming a cell radius of 2.6 m and a permeability of 10-5 m/s). Neither Bidigare et al. (1997) nor the present study tested µ/CO2 values larger than 0.07 kg/(mol d). Indeed, substantial physiological evidence indicates that E. huxleyi uses a variety of CCMs to actively enhance intracellular CO2 (e.g., Rost et al., 2006; Mackinder et al., 2011; Bach et al., 2013; Isensee et al., 2014). We also cannot rule out the

13 13 possibility that changes in  CDIC contributed to the variation in  Cbiomass that we observe. Distinct CO2 conditions were achieved by bubbling with varying amounts of isotopically-depleted tank CO2 and N2:O2.

13 Under higher [CO2(aq)], δ Cbiomass values may therefore be driven lighter, in part, by a lighter isotopic composition of DIC. Notably, the carbonate system chemistry in Bidigare et al. (1997) was manipulated

13 in a similar manner and yielded linear responses of both δ Cbiomass and P values to µ/[CO2(aq)] (Figure A.2).

13 The parallel increase in  Ccalcite values with increasing µ/[CO2(aq)] (Figure 2.1) suggests that a common physiological mechanism governs the isotopic signature of both calcite and biomass.

13 13 Photosynthesis discriminates against C in the chloroplast, producing depleted  Cbiomass values while simultaneously enriching 13C in the remaining intracellular DIC pool. If this enriched DIC is also used for

13 13 calcification,  Ccalcite and  Cbiomass values might both be expected to increase in response to increasing

13 µ/[CO2(aq)], consistent with our observed constant value for  Ccalcite-biomass (32.3 ± 0.7‰; Table 2.2) and measurements of other lightly-calcifying coccolithophores (Hermoso et al., 2016b, McClelland et al.,

2017). In species with higher ratios of particulate inorganic to organic carbon (high PIC:POC ratios),

13 13  Ccalcite values would instead be expected to decrease with increasing  Cbiomass values since the Rayleigh- type fractionation associated with calcification within the coccolith vesicle would obscure the chloroplast- derived fraction of the signal (Hermoso et al., 2016b, McClelland et al., 2017).

Similar reasoning also explains the linear responses of the organic compound classes (alkenones

13 and CAPs) to µ/[CO2(aq)], but it cannot explain why the slope of the  C vs. µ/[CO2(aq)] signature is not

13 conserved in the CAPs (Figure 2.1). Under high utilization of CO2 (larger values for µ/[CO2(aq)]),  CCAP

13 13 values appear to be approaching  Cbiomass values, while at very low utilization, the value of  CCAP-biomass has a predicted maximum of 11‰ (y-axis intercept, Figure 2.2b). This changing signature specifically

13 indicates that  CCAP values are approximately half as sensitive as the other cellular components to the ratio

µ/[CO2(aq)], and it points towards a more complex control on the isotopic composition of this polymer.

13 Physiological interpretation of  CCAP patterns

13 The fractionation between CAPs and biomass ( CCAP-biomass) spans a 5.6‰ range without any

13 compensatory change in  Calkenone-biomass values. This could occur if isotopically-distinct carbohydrate pools exist within E. huxleyi and the relative amount of carbon flowing to these pools varies systematically with growth conditions. Simple assumptions about cellular composition are used to illustrate this hypothesis, following Hayes (2001).

13  Cbiomass values can be decomposed into the fractional allocation of cellular carbon to the major organic compound classes, as well as the mass-weighted average isotopic composition of each class (Eqn.

1). The subscripts in Equation 1 correspond to proteins, carbohydrates, and lipids, respectively; f is the fractional flux of fixed carbon flowing to each compound class at steady-state. We also assume that the bulk carbohydrate pool, denoted carb, results from the balance of two distinct carbohydrate fractions: acidic polysaccharides (fCAP) and other saccharides (fsacc) (Eqn. 2).

biomass = fprotprot + fcarbcarb + fliplip (1)

carb = fCAPCAP + fsaccsacc (2)

Values of biomass and CAP correspond to the values listed in Table 2.1, and lip values are estimated

13 to be 4‰ C-depleted relative to biomass (lip= biomass – 4‰, Eqn. 3, Table 2.3) by analogy with our

13  Calkenones measurements and other consensus observations (e.g., Schouten et al., 1998, Laws et al., 2001).

Alkenones and total lipids are assumed to be isotopically equivalent because alkenones represent a significant fraction of the lipids synthesized in E. huxleyi. Quantitative analyses of carbon fluxes show that up to 18% of photosynthetic carbon is dedicated to alkenones in cells harvested from batch cultures (Tsuji et al., 2015). We further assume that proteins and biomass would be isotopically equal (prot = biomass), since proteins are expected to represent the majority of total cell carbon (e.g., Hayes, 2001, Tang et al., 2017).

Finally, carb, representing the total combined carbohydrate fraction in the cell, is assumed to be enriched by 4‰ relative to biomass (carb = biomass + 4‰, Eqn. 3, Table 2.3). This assumption is at the upper end of the range reported for bulk carbohydrates in other autotrophs (Wong et al., 1975, van der Meer et al., 2001, van Dongen et al., 2002).

biomass = fprot(biomass) + fcarb(biomass + 4‰) + flip(biomass – 4‰) (3)

Solving Eqn 3 requires an estimate of one fractional flux, so we estimate flip = 0.25 by considering how biosynthesis affects the carbon isotopic composition of n-alkyl lipids. Acetyl-coenzyme A (CoA) is produced from the decarboxylation of pyruvate by the enzyme pyruvate dehydrogenase. This step is accompanied by an isotope effect of ~23‰, generating a 13C-depleted carboxyl group relative to the methyl group in acetyl-CoA (Monson and Hayes, 1982, Hayes, 2001). This translates to a ~12‰ depletion for the overall acetyl-CoA molecule relative to pyruvate. However, the net isotopic composition of the resulting acetogenic lipids relative to biomass is controlled by the branching ratio of pyruvate destined for lipid synthesis vs. that used directly, either for cellular biosynthesis or for carboxylation to oxaloacetate by anaplerotic reactions. In the absence of anaplerotic reactions, an expressed 4‰ depletion in lipids relative to biomass would indicate allocation of ~33% of pyruvate to lipid synthesis, but accounting for additional anaplerotic fluxes (e.g., Tang et al., 2017) decreases this estimate to ~25%.

Table 2.3. Four carbon allocation scenarios consistent with δbiomass and δCAP measurements. % carbon Relative  Expt # δbiomass flip δlip fprot δprot fcarb δcarb fCAP δCAP fsacc δsacc allocation to CAPs

Prota 0 1 –48.2 0.25 –52.2 0.50 –48.2 0.25 –44.2 0.14 –38.2 0.86 –45.2 3.5 Carba 4 2 –45.7 0.25 –49.7 0.50 –45.7 0.25 –41.7 0.37 –37.3 0.63 –44.3 9.3 Lipa –4 3 –42.9 0.25 –46.9 0.50 –42.9 0.25 –38.9 0.50 –35.4 0.50 –42.4 12.5 a 7 4 –39.5 0.25 –43.5 0.50 –39.5 0.25 –35.5 0.91 –34.9 0.09 –41.9 22.8 CAP-sacc Prot 1 1 –48.2 0.25 –52.2 0.63 –47.2 0.13 –45.2 0.13 –38.2 0.87 –46.2 1.6 33 3 2 –45.7 0.25 –49.7 0.63 –44.7 0.13 –42.7 0.33 –37.3 0.67 –45.3 4.1 Carb Lip –4 3 –42.9 0.25 –46.9 0.63 –41.9 0.13 –39.9 0.44 –35.4 0.56 –43.4 5.5

CAP-sacc 8 4 –39.5 0.25 –43.5 0.63 –38.5 0.13 –36.5 0.80 –34.9 0.20 –42.9 10.0 Prot 0 1 –48.2 0.25 –52.2 0.42 –48.2 0.33 –45.2 0.13 –38.2 0.87 –46.2 4.3 Carb 3 2 –45.7 0.25 –49.7 0.42 –45.7 0.33 –42.7 0.33 –37.3 0.67 –45.3 10.9 Lip –4 3 –42.9 0.25 –46.9 0.42 –42.9 0.33 –39.9 0.44 –35.4 0.56 –43.4 14.5

CAP-sacc 8 4 –39.5 0.25 –43.5 0.42 –39.5 0.33 –36.5 0.80 –34.9 0.20 –42.9 26.4 Prot 1 1 –48.2 0.25 –52.2 0.50 –47.2 0.25 –46.2 0.11 –38.2 0.89 –47.2 2.8 Carb 2 2 –45.7 0.25 –49.7 0.50 –44.7 0.25 –43.7 0.29 –37.3 0.71 –46.3 7.3 Lip –4 3 –42.9 0.25 –46.9 0.50 –41.9 0.25 –40.9 0.39 –35.4 0.61 –44.4 9.8

CAP-sacc 9 4 –39.5 0.25 –43.5 0.50 –38.5 0.25 –37.5 0.71 –34.9 0.29 –43.9 17.8 aScenario described in the body of the text.

Proteins and carbohydrates must together constitute the remaining biomass (75%) in Eqn. 3, allowing us to solve for the remaining unknowns. We find that bulk cellular carbon allocation remains invariant across our experiments: flip = 0.25, fprot = 0.5, fcarb = 0.25 (Table 2.3, Figure 2.3). This finding is plausible but prescribed by the assumptions of the model: mass balance calculations with constant isotopic offsets between pools will result in constant allocation to fcarb and fprot regardless of choice of flip. For example, other reasonable choices for flip (ranging from 0.15−0.33) are explored in Table A.4, producing different fractional carbon allocations to fprot (0.7−0.33) and fcarb (0.15−0.33).

The fractional allocation to CAPs (fCAP) versus other saccharides (fsacc) is calculated by assuming a net fractionation factor between the two pools (CAP-sacc) of 7‰ (Equation 4, Table 2.3). This value

13 corresponds to the maximum  CCAP-biomass value of 11‰ inferred from the y-intercept of Figure 2.2b, corrected by 4‰ to account for our assumption that carb is 4‰ enriched relative to biomass (Equation 3).

carb = biomass + 4‰ = fCAPCAP + (1–fCAP)(CAP–7‰) (4)

The fCAP results (Table 2.3) indicate that the total cellular carbon allocation to CAPs increases from

3.5 to 22.8% from Experiment #1 to #4, in agreement with 14C-labeling experiments showing that ~15–

20% of fixed carbon is used for the synthesis of acidic polysaccharides in E. huxleyi (Kayano and Shiraiwa

2009, Tsuji et a., 2015, Taylor et al., 2017).

Figure. 2.3. Predicted cellular carbon allocation to major organic compound classes in E. huxleyi. Arrow widths correspond to calculated fractional fluxes (f). Relative shading intensities indicate the measured or predicted isotopic offsets between the organic carbon pools, including the bulk biomass. (a) Cells from

Experiment #1 (left; lowest /[CO2(aq)]) preferentially allocate carbon within the carbohydrate pool to saccharides other than CAPs. Cells from Experiment #4 (right; highest /[CO2(aq)]) allocate more carbon to CAP synthesis. (b) Percentages of fixed carbon allocated to CAPs for each experiment, calculated for four plausible combinations of prot and carb model inputs. Regardless of choice, carbon allocation to CAPs is expected to increase from Experiment #1 to Experiment #4 (with increasing /[CO2(aq)]). Abbreviations: PDH = pyruvate dehydrogenase, OAA = oxaloacetate.

To test the sensitivity of our model results, other plausible choices of prot and carb inputs are evaluated in Table 2.3, Figure 2.3b, and Figure A.3; all cases suggest that E. huxleyi produces more CAPs relative to other carbohydrates as µ/[CO2(aq)] increases (Table 2.3, Figure 2.3). It remains unknown whether this response in our cultures is a consequence of the faster growth rate or the relatively lower CO2 concentrations under these conditions. Several studies imply that E. huxleyi regulates carbon flow among

CAPs, neutral polysaccharides, and low molecular weight (LMW) metabolites in response to changing growth conditions (Kayano and Shiraiwa, 2009, Borchard and Engel, 2012, Chow et al., 2015). Most of these studies have been conducted with batch cultures or phosphorous-limited chemostat cultures, so as yet no corroborating evidence for nitrate-limited conditions exists. However, our estimates are consistent with

14C-labeling experiments with the green alga Dunaliella tertiolecta in nitrate-limited chemostat cultures

(Halsey et al., 2011), which showed that at faster growth rates, proportionally more 14C is allocated to polysaccharides, and at slower growth rates, more 14C is allocated to LMW metabolites. For E. huxleyi, this may reflect a strategy by the cell to support calcification when the intracellular carbon pool is relatively depleted.

Alternatively, our data might imply that CAPs are synthesized from proportionally more recycled cellular carbon with increasing growth rates. If proteins are recycled for the synthesis of CAPs, and prot ≈

13 biomass, then CAP and biomass  C values would be expected to converge with enhanced recycling.

However, such explanation requires that the recycled carbon be incorporated at a biosynthetic stage downstream of simple sugars and prior to the polymerization of CAPs. If the recycled carbon first passed through simple sugars, then no difference in slopes would result among organic compound classes since simple sugars are precursors to proteins and lipids, as well as CAPs. One additional possibility is that the monosaccharide composition may have changed systematically with the changing growth conditions. Only one polysaccharide was detected per extract by PAGE, SDS-PAGE, and HPLC, but we did not explicitly verify that the monosaccharide composition of the CAP remained constant across experiments. Future

36 studies could clarify this contribution by comparing the isotopic composition of CAPs with that of individual monomers from the same extract.

Conclusions: Implications for paleoceanography

Our findings indicate that alkenones have greater inherent sensitivity for paleobarometry

13 applications than CAPs due to the steeper slope relating  Calkenone values to µ/[CO2(aq)] (Figure 2.1). The relatively constant isotopic depletion in alkenones compared to biomass across four growth conditions supports the conclusions of Popp et al. (1998b) that growth rate does not significantly or systematically influence the isotopic offset between bulk biomass and alkenones. This result upholds the fundamental

13 assumption employed in alkenone-based pCO2 reconstructions that sedimentary  Calkenone measurements

13 can be used to estimate original  Cbiomass values.

In contrast, CAPs are not direct analogues or replacements for alkenone biomarkers in the

13 13 reconstruction of  Cbiomass values, since  Cbiomass-CAP varies by 5.6‰ in our experiments. This is a large variation relative to the limited range of investigated growth and [CO2(aq)] conditions. However, assuming that our chemostat results are applicable to in situ processes in both ancient and modern environments, then alkenones, calcite, and CAPs all recovered from the same sedimentary deposit may together be more useful than alkenones and calcite alone: isotopic measurements of all three cellular constituents would constrain

3- µ/[CO2(aq)] with fewer assumptions than current approaches using the [PO4 ]-derived “b” parameter (Rau et al., 1992, Bidigare et al., 1997). If the alkenones, calcite, and CAPs are related by  values that map onto Figure 2.2a and 2.2b, this information reconstructs µ/[CO2(aq)] directly. One advantage of this approach

13 is that it does not require seawater  CDIC values, enabling reconstructions from sediments lacking coeval planktonic foraminifera. Another advantage is the potential to recover additional useful information about the coccolithophore community through parallel efforts. CAPs are extracted from fossilized coccoliths, so cell size and taxonomy information should be recoverable with coccolith calcite and CAP isotope values

(Henderiks, 2008, Bolton et al., 2012, Bolton and Stoll, 2013, O’Dea et al., 2014, McClelland et al., 2016).

The uronic acid contents of CAPs provide complementary information about the internal saturation state of the coccolith vesicle and, by extension, atmospheric CO2 levels (Lee et al., 2016, Rickaby et al., 2016). If

13  CDIC can be estimated using independent measurements (e.g., mixed-layer foraminifera), then

13 13 13 measurements of either  CCAP−alkenone or  Ccalcite−CAP could be used to reconstruct  Cbiomass. Together these strategies point to a way forward for paleobarometry: by analyzing a suite of geologically-preserved constituents, including CAPs, it may be possible to reduce some of the uncertainties inherent to any single-proxy approach. Thus, with additional proxy validation efforts, including cultures and field studies encompassing other species and CO2 conditions, the carbon isotopic composition of CAPs may provide an important biochemical window into ancient environments.

Acknowledgements

This work was supported by a National Science Foundation Graduate Research Fellowship [grant number

DGE1144152, to EBW]; the Gordon and Betty Moore Foundation (to AP); and NASA-NAI CAN6 (to AP;

PI Roger Summons, MIT). We thank Susie Carter, Katie Mabbott, and Alan Gagnon for assistance with laboratory analyses, Einat Segev for providing starter cultures and culturing advice, John Volkman and Bart van Dongen for editorial handling, and two anonymous reviewers for their thoughtful comments.

References

Abselson, P.H., Hoering, T.C., 1961. Carbon isotope fractionation in formation of amino acids by photosynthetic organisms. Proceedings of the National Academy of Sciences of the United States of America 47, 623–632. Bach, L.T., MacKinder, L.C.M., Schulz, K.G., Wheeler, G., Schroeder, D.C., Brownlee, C., Riebesell, U., 2013. Dissecting the impact of CO2 and pH on the mechanisms of photosynthesis and calcification in the coccolithophore Emiliania huxleyi. New Phytologist 199, 121-134. Benthien, A., Zondervan, I., Engel, A., Hefter, J., Terbrüggen, A., Riebesell, U., 2007. Carbon isotopic fractionation during a mesocosm bloom experiment dominated by Emiliania huxleyi: Effects of CO2 concentration and primary production. Geochimica et Cosmochimica Acta 71, 1528–1541. Bevington P.R., Robinson, D.K., 2003. Data reduction and error analysis for the physical sciences, second ed. McGraw-Hill, New York. Bidigare, R.R., Fluegge, A., Freeman, K.H., Hanson, K.L., Hayes, J.M., Hollander, D., Jasper, J.P., King, L.L., Laws, E.A., Milder, J., Miller, F.J., Pancost, R., Popp, B.N., Steinberg, P.A., Wakeham, S.G., 1997. Consistent fractionation of 13C in nature and in the laboratory: Growth-rate effects in some haptophyte algae. Global Biogeochemical Cycles 11, 279–292. Bijl, P.K., Houben, A.J.P., Schouten, S., Bohaty, S.M., Sluijs, A., Reichart, G.J., Sinninghe Damsté, D.S., Brinkhuis, H., 2010. Transient middle Eocene atmospheric CO2 and temperature variations. Nature 330, 819–821. Bligh, E.G., Dyer, 1959. A rapid method of total lipid extraction and purification. Canadian Journal of Biochemistry and Physiology 37, 911–919. Boller, A.J., Thomas, P.J., Cavanaugh, C.M., Scott, K.M., 2011. Low stable carbon isotope fractionation by coccolithophore RubisCO. Geochimica et Cosmochimica Acta 75, 7200–7207. Bolton, C.T., Stoll, H.M., Mendez-Vicente, A., 2012. Vital effects in coccolith calcite: Cenozoic climate pCO2 drove the diversity of carbon acquisition strategies in coccolithophores? Paleoceanography 27, PA4204. Bolton, C.T., Stoll, H.M., 2013. Late Miocene threshold response of marine algae to carbon dioxide limitation. Nature 500, 558–562. Borchard, C., Engel, A., 2012. Organic matter exudation by Emiliania huxleyi under simulated future ocean conditions. Biogeosciences 9, 3405–3423.

Borman, A.H., de Jong, E.W., Huizinga, M., Kok, D.J., Westbroek P., Bosch L., 1982. The role in CaCO3 crystallization of an acid Ca2+-binding polysaccharide associated with coccoliths of Emiliania huxleyi. European Journal of Biochemistry 129, 179–183. Bown, P.R., 1987. Taxonomy, biostratigraphy, and evolution of late Triassic-early Jurassic calcareous nannofossils. Special Papers in Palaeontology 38, 1–18. Brassell, S.C., 2014. Climatic influences on the Paleogene evolution of alkenones. Paleoceanography 29, 255–272.

Chow, J.S., Lee, C., Engel, A., 2015. The influence of acidic polysaccharides, growth rate, and free coccoliths on the coagulation efficiency of Emiliania huxleyi. Marine Chemistry 175, 5–17. Coffin, R.B., Velinsky, D.J., Devereux, R. Price, W.A., Cifuentes, L.A., 1990. Stable carbon isotope analysis of nucleic acids to trace sources of dissolved substrates used by estuarine bacteria. Applied Environmental Microbiology 56, 2012–2020. De Jong, E.W., Bosch, L., Westbroek, P., 1976. Isolation and characterization of a Ca2+-binding polysaccharide associated with coccoliths of Emiliania huxleyi (Lohmann) Kamptner. European Journal of Biochemistry 70, 611–612. De Vargas, C., Aubry, M.-P., Probert, I., Young, J., 2007. Origin and evolution of coccolithophores: From coastal hunters to oceanic farmers. In: Falkowski, K, Knoll, A. (Eds.), Evolution of Primary Producers in the Sea. Elsevier, Boston, pp. 133–163.

Dickson, A.G., 1990. Standard potential of the reaction: AgCl(s) + 1/2H2(g)=Ag(s) + HCl(aq), and the - standard aciditiy constant of the ion HSO4 in synthetic seawater from 273.15 to 318.15 K. The Journal of Chemical Thermodynamics 22,113-127. Dickson, A.G., Millero, F.J., 1987. A comparison of the equilibrium constants for the dissociation of carbonic acid in seawater media. Deep Sea Research Part I 34, 1733–1743.

Dickson, A.G., Sabine, C.L., Christian, J.R., 2007. Guide to best practices for ocean CO2 measurements. PICES Special Publication 3, 1–191. Farrimond P., Eglinton, G., Brassell, S.C., 1986. Alkenones in Cretaceous black shales, Blake-Bahama Basin, western North Atlantic. Organic Geochemistry 10, 897–903. Fichtinger-Schepman, A.M., Kamerling, J.P., Versluis, C., Vliegenthart, F.G., 1981. Structural studies of the methylated, acidic polysaccharide associated with coccoliths of Emiliania huxleyi (Lohmann) Kamptner, 1981. Carbohydrate Research 93, 105–123. Freeman, K.H., Hayes, J.M., 1992. Fractionation of carbon isotopes by phytoplankton and estimates of ancient CO2 levels. Global Biogeochemical Cycles 6, 185–198. Goericke, R., Montoya, J., Fry, B., 1994. Physiology of isotopic fractionation in algae and cyanobacteria. In: Lajtha, K., Michener, R. (Eds.), Stable Isotopes in Ecology and Environmental Science. Blackwell Scientific Publications, Oxford, pp. 187–221. Gran, G., 1952. Determination of the equivalence point in potentiometric titrations, Part II. Analyst 77, 661–671. Guillard, R.R.L., Hargraves, P.E., 1993. Stichochrysis immobilis is a diatom, not a chrysophyte. Phycologia 32, 234–236. Halsey, K.H., Milligan, A.J., Behrenfeld, M.J., 2011. Linking time-dependent carbon-fixation efficiencies in Dunaliella tertiolecta (chlorophyceae) to underlying metabolic pathways. Journal of Phycology 47, 66–76. Hayes, J.M., 2001. Fractionation of carbon and hydrogen isotopes in biosynthetic processes. Reviews in Mineralogy and Geochemistry 43, 225–277.

Henderiks, J., 2008. Coccolithophore size rules – Reconstructing ancient cell geometry and cellular calcite quota from fossil coccoliths. Marine Micropaleontology 67, 143–154. Henriksen, K., Stipp, S.L.S., 2009. Controlling biomineralization: The effect of solution composition on coccolith polysaccharide functionality. Crystal Growth and Design 9, 2088–2097. Hermoso, M., Rickaby, R.E.M, Minoletti, F. Diester-Haas, L., 2009. The Middle Miocene paleoceanographic events as seen by the chemistry of calcareous nannofossils in SW Pacific Ocean. Eos Transactions AGU 90, PP41A-1498. Hermoso, M., 2014. Coccolith-derived isotopic proxies in palaeoceanography: where geologists need biologists. Cryptogamie, Algologie 35, 323–351. Hermoso, M., Minoletti, F., Aloisi, G., Bonifacie, M., McClelland, H.L.O., Labourdette, N., REnforth, P., Chaduteau, C., Rickaby, R.E.M., 2016a. An explanation for the 18O excess in Noelaerhabdaceae coccolith calcite. Geochimica et Cosmochimica Acta 189, 132–142. Hermoso, M., Chan, I.Z.X., McClelland, H.L.O., Heureux, A.M.C., Rickaby, R.E.M., 2016b. Vanishing coccolith vital effects with alleviated carbon limitation. Biogeoscience 13, 301–312. Holtz, L.-M., Wolf-Gladrow, D., Thoms, S., 2017. Stable carbon isotope signals in particulate organic and inorganic carbon of coccolithophores – A numerical model study for Emiliani huxleyi. Journal of Theoretical Biology 420, 117–127. Isensee, K., Erez, J., Stoll, H.M., 2014. Detection of a variable intracellular acid-labile carbon pool in Thalassiosira weissflogii (Heterokontophyta) and Emiliania huxleyi (Haptophyta) in response to changes in the seawater carbon system. Physiologia Plantarum 150, 321–338. Jasper, J.P., Hayes, J.M., 1990. A carbon isotope record of CO2 levels during the late Quaternary. Nature 347, 462– 464. Jasper, J.P, Hayes, J.M., Mix, A.C., Prahl, F.G., 1994. Photosynthetic fractionation of 13C and concentrations of CO2 in the central equatorial Pacific during the last 225,000 years. Paleoceanography 9, 781–898. Katz, A., Bonifacie, M., Hermoso, M., Cartigny, P., Calmels, D., 2017. Laboratory-grown coccoliths

exhibit no vital effect in clumped isotope (47) composition on a range of geologically relevant temperatures. Geochimica et Cosmochimica Acta 208, 335–353. Kayano, K., Shiraiwa, Y., 2009. Physiological regulation of coccolith polysaccharide production by phosphate availability in the coccolithophorid Emiliania huxleyi. Plant Cell Physiology 50, 1522– 1531.

Kok, D.J., Blomen, L.J.M.J., Westbroek, P., Bijvoet, O.L.M., 1986. Polysaccharide from coccoliths (CaCO3 biomineral): influence on crystallization of calcium oxalate monohydrate. European Journal of Biochemistry 158, 167–172. Laws, E.A., Popp, B.N., Bidigare, R.R., Kennicutt, M.C., Macko, S.A., 1995. Dependence of phytoplankton carbon isotopic composition on growth rate and CO2aq: theoretical considerations and experimental results. Geochimica et Cosmochimica Acta 59, 1131–1138.

Laws, E.A., Popp, B.N., Bidigare, R.R., Riebesell, U., Burkhardt, S., Wakeham, S.G., 2001. Controls on the molecular distribution and caron isotopic composition of alkenones in certain haptophyte algae. Geochemistry Geophysics Geosystems 2, 2000GC000057. Laws, E.A., Popp, B.N., Cassar, N., Tanimoto, J., 2002. 13C discrimination patterns in oceanic phytoplankton: likely influence of CO2 concentrating mechanisms, and implications for paleoreconstructions. Functional Plant Biology 29, 323−333. Lee, R.B.Y., Mavridou, D.A.I., Papadakos, G., McClelland, H.L.O., Rickaby, R.E.M., 2016. The uronic acid content of coccolith-associated polysaccharides provides insight into coccolithogenesis and past climate. Nature Communications 7, DOI: 10.1038/ncomms13144.

Lewis, E., Wallace, D., 1998. Program developed for CO2 system calculations: Oak Ridge TN Oak Ridge National Laboratory Environmental Sciences Division, v. 4735. Mackinder, L., Wheeler, G., Schroeder, D., von Dassow, P., Riebesell, U., Brownlee, C., 2011. Expression of biomineralization-related ion transport genes in Emiliania huxleyi. Environmental Microbiology 13, 3250–3265. Macko, S.A., Helleur, R., Hartley, G., Jackman, P., 1990. Diagenesis of organic matter – a study using stable isotopes of individual carbohydrates. Organic Geochemistry 16, 1129–1137. Marchal, O., Cacho, I., Stocker, T.F., Grimalt, J.O., Calvo, E., Martrat, B., Shackleton, N., Vautravers, M., Cortijo, E., van Kreveld, S., Andersson, C., Koc, N., Chapman, M., Sbaffi, L., Duplessy, J.C., Sarnthein, M., Turon, J.L., Duprat, J., and Jansen, E., 2002. Apparent long-term cooling of the sea surface in the northeast Atlantic and Mediterranean during the Holocene. Quaternary Science Reviews 21, 455–483. Marsh, M.E., 1994. Polyanion-mediated mineralization–assembly and reorganization of acidic polysaccharides in the Golgi system of a coccolithophorid alga during mineral deposition, Protoplasma 177, 108–122. Marsh, M.E., Ridall, A.L., Azadi, P., Duke, P.J., 2002. Galacturonomannan and golgi-dreived membrane linked to growth and shaping of biogenic calcite. Journal of Structural Biology 139, 39–45. Marsh, M.E., 2003. Regulation of CaCO(3) formation in coccolithophores. Comparative Biochemistry and Physiology Part B: Biochemistry and Molecular Biology 136, 743–754. McClelland, H.L.O., Hermoso, M., Bruggeman, J., Rickaby, R.E.M., 2015. Towards a unifying theory for carbon isotopic partitioning in coccolithophores: Implications for paleo-proxies. Goldschmidt Abstracts, 2069. McClelland, H.L.O, Barbarin, N., Beaufort, L., Hermoso, M., Ferretti, P., Greaves, M., Rickaby, RE.M., 2016. Calcification response of a key phytoplankton family to millennial-scale environmental change. Scientific Reports 6, 34263. McClelland, H.L.O., Bruggeman, J., Hermoso, M., Rickaby, R.E.M., 2017. The origin of carbon isotope vital effects in coccolith calcite. Nature Communications 8, DOI: 10.1038/ncomms14511. Mehrbach, C., Culberson, C.H., Hawley, J.E., Pytkowicz, R.M., 1973. Measurement of the apparent dissociation constants of carbonic acid in seawater at atmospheric pressure. Limnology and Oceanography 18, 897–907.

Monson, K.D., Hayes, J.M., 1982. Carbon isotopic fractionation in the biosynthesis of bacterial fatty acids—Ozonolysis of unsaturated fatty acids as a means of determining the intramolecular distribution of carbon isotopes. Geochimica et Cosmocimica Acta 46, 139–149. Mook, W.G., Bommerson, J.C., Staverman, W.H., 1974. Carbon isotope fractionation between dissolved bicarbonate and gaseous carbon dioxide. Earth and Planetary Science Letters 22, 169–176.

Müller, M.N., Beaufort, L., Bernard, O., Pedrotti, M.L., Talec, A., Sciandra, A., 2012. Influence of CO2 and nitrogen limitation on the coccolith volume of Emiliania huxleyi (Haptophyta). Biogeosciences 9, 4155–4167. O’Dea, S.A., Gibbs, S.J., Bown, P.R., Young, J.R., Poulton, A.J., Newsam, C., Wilson, P.A., 2014. Coccolithophore calcification response to past ocean acidification and climate change. Nature Communications 5, DOI:10.1038/ncomms6363. Pagani, M., Zachos, J., Freeman, K.H., Bohaty, S., Tipple, B., 2005. Marked decline in atmospheric carbon dioxide concentrations during the Paleogene. Science 309, 600–603. Perrin, L. Probert, I., Langer, G., Aloisi, G., 2016. Growth of the coccolithophore Emiliania huxleyi in light- and nutrient-limited batch reactors: relevance for the BIOSOPE deep ecological niche of coccolithophores. Biogeosciences 13, 5983–6001. Popp, B.N., Laws, E.A., Bidigare, R.R., Dore, J.E., Hanson, K.L., Wakeham, S.G., 1998a. Effect of phytoplankton cell geometry on carbon isotopic fractionation. Geochimica et Cosmochimica Acta 62, 69–77. Popp, B.N., Kenig, F., Wakeham, S.G., Laws, E.A., Bidigare, R.R., 1998b. Does growth rate affect ketone unsaturation and intracellular carbon isotopic variability in Emiliania huxleyi? Paleoceanography 13, 35–41.

Rau, G.H., Takahashi, T., Des Marais, D.J., Repeta, D.J., Martin, J.H., 1992. The relationship between 13C of organic matter and [CO2(aq)] in ocean surface water: Data from a JGOFS site in the northeast Atlantic Ocean and a model. Geochimica et Cosmochimica Acta 56, 1413–1419. Rickaby, R.E.M., Bard, E., Sonzogni, C., Rostek, F., Beaufort, L., Barker, S., Rees, G., Schrag, D.P., 2007. Coccolith chemistry reveals secular variations in the global ocean carbon cycle? Earth and Planetary Science Letters 253, 83–95. Rickaby, R.E.M., Hermoso, M., Lee, R.B.Y., Rae, B.D., Heureux, A.M.C., Balestreri, C., Chakravarti, L., Schroeder, D.C., Brownlee, C., 2016. Environmental carbonate chemistry selects for phenotype of recently isolated strains of Emiliania huxleyi. Deep-Sea Research II 127, 28–40.

Riebesell, U., Revill, A.T., Holdsworth, D.G., Volkman, J.K., 2000. The effects of varying CO2 concentration on lipid composition and carbon isotope fractionation in Emiliania huxleyi. Geochimica et Cosmochimica Acta 64, 4179–4192. Romanek, C.S., Grossman, E.L., Morse, J.W., 1992. Carbon isotopic fractionation in synthetic aragonite and calcite: Effects of temperature and precipitation rate. Geochimica et Cosmochimica Acta 56, 419–430. Rost, B., Zondervan, I., Riebesell, U., 2002. Light-dependent carbon isotope fractionation in the coccolithophorid Emiliania huxleyi. Limnology and Oceanography 47, 120–128.

Rost, B., Riebesell, U., Sültemeyer, D., 2006. Carbon acquisition of marine phytoplankton: Effect of photoperiod length. Limnology and Oceanography 51, 12–20. Sakata, S., Hayes, J.M., McTaggart, A.R., Evans, R.A., Leckrone, K.J., Togasaki, R.K., 1997. Carbon isotopic fractionation associated with lipid biosynthesis by a cyanobacterium: Relevance for interpretation of biomarker records. Geochimica et Cosmochimica Acta 61, 5379–5389. Schouten, S., Klein Breteler, W.C.M., Blokker, P., Schogt, N., Rupstra, I.C., Grice, K., Baas, M., Sinninghe Damsté, J.S., 1998. Biosynthetic effects on the stable carbon isotopic compositions of algal lipids: Implications for deciphering the carbon isotopic biomarker record. Geochimica et Cosmochimica Acta 62, 1397–1406. Seki, O., Foster, G.L., Schmidt, D.N., Mackensen, A., Kawamura, K., Pancost, R.D., 2010. Alkenone and boron based Pliocene pCO2 records. Earth and Planetary Science Letters 292, 201–211. Stoll, H.M., 2005. Limited range of interspecific vital effects in coccolith stable isotopic records during the Paleocene-Eocene thermal maximum. Paleoceanography 20, DOI: 10.1029/2004PA001046. Strickland, J.D.H., Parsons, T.R., 1968. Determination of reactive phosphorus. In: A practical handbook of seawater analysis. Fisheries Research Board of Canada, Bulletin 167, pp. 49–56. Tang, T., Mohr, W., Sattin, S.R., Rogers, D.R., Girguis, P.R., Pearson, A., 2017. Geochemically distinct carbon isotope distributions in Allochromatium vinosum DSM 180T grown photoautotrophically and photoheterotrophically. Geobiology 15, 324–339. Taylor, A.R., Brownlee, C., Wheeler G., 2017. Coccolithophore cell biology: chalking up progress. Annual Review of Marine Science 9, 283–310. Tchernov, D., Gruber, D.F., Irwin, A., 2014. Isotopic fractionation of carbon in the coccolithophorid Emiliania huxleyi. Marine Ecology Progress Series 508, 53–66. Teece, M.A., Fogel, M.L., 2007. Stable carbon isotope biogeochemistry of monosaccharides in aquatic organisms and terrestrial plants. Organic Geochemistry 38, 458–473. Thierstein, H.R., Geitzenauer, K.R., Molfino, B., Shackleton, N.J., 1977. Global synchroneity of late Quaternary coccolith datum levels validation by oxygen isotopes. Geology 5, 400–404. Tsuji, Y., Yamazaki, M., Suzuki, I., Shiraiwa, Y., 2015. Quantitative analysis of carbon flow into photosynthetic products functioning as carbon storage in the marine coccolithophore, Emiliania huxleyi. Marine Biotechnology 17, 428–440. Van der Meer, M.T.J., Schouten, S., van Dongen, B.E., Rijpstra, I.C., Fuchs, G., Sinninghe Damsté, J.S., de Leeuw, J.W., Ward, D.M., 2001. Biosynthetic controls on the 13C contents of organic components in the photoautotrophic bacterium Chloroflexus aurantiacus. The Journal of Biological Chemistry 276, 10971–10976. Van Dongen, B.E., Schouten, S., Sinninghe Damsté, J.S., 2002. Carbon isotope variability in monosaccharides and lipids of aquatic algae and terrestrial plants. Marine Ecology Progress Series 232, 83–92. Van Heuven, S., Pierrot, D., Rae, J.W.B., Lewis, E., Wallace, D.W.R., 2011. MATLAB program developed for CO2 system calculations. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory.

Volkman, J.K., Eglinton, G., Corner, E.D.S., Sargent, J.R., 1980. Novel unsaturated straight-chain C37-C39 methyl and ethyl ketones in marine sediments and a coccolithophore Emiliania huxleyi. Physics and Chemistry of the Earth 12, 219–227. Westbroek, P., Brown C.W., van Bleijswijk, J., Brownlee, C., Brummer, G.J., Conte, M., Egge, J., Fernandéz, E., Jordan, R., Knappertsbusch, M., Stefels, J., Veldhuis, M., van der Wal, P., Young, J., 1993. A model system approach to biological climate forcing—the example of Emiliania huxleyi. Global Planetary Change 8, 27–46.

Wilkes, E.B., Carter, S.J., Pearson, A., 2017. CO2-dependent carbon isotope fractionation in the dinoflagellate Alexandrium tamarense. Geochimica et Cosmochimica Acta 212, 48–61. Wong, W., Sackett, W.M., Benedict, C.R., 1975. Isotope fractionation in photosynthetic bacteria during carbon dioxide assimilation. Plant Physiology 55, 475–479. Young, J.R., Henricksen, K., 2003. Biomineralization vesicles: the calcite of coccoliths. Reviews in Mineralogy and Geochemistry 54, 189–214.

Zeebe, R.E.,Wolf-Gladrow, D., 2001. CO2 in seawater: Equilibrium, kinetics, isotopes. Elsevier, Amsterdam. Zhang, J., Quay, P.D., Wilbur, D.O., 1995. Carbon isotope fractionation during gas-water exchange and dissolution of CO2. Geochimica et Cosmochimica Acta 59, 107–114. Zhang, J.-Z., Fischer, C.J., 2006. A simplified resorcinol method for direct spectrophotometric determination of nitrate in seawater. Marine Chemistry 99, 220–226. Zhang, Y.G., Pagani, M., Liu, Z., Bohaty, S.M., Deconto, R., 2013. A 40-million-year history of atmospheric CO2. Philosophical Transactions of the Royal Society A 371, 20130096. Ziveri, P., Stoll, H., Probert, I., Klaas, C., Geisen, M., Ganssen, G.M., Young, J., 2003. Stable isotope ‘vital effects’ in coccolith calcite. Earth and Planetary Science Letters 210, 137–149.

Page intentionally left blank

Chapter 3

CO2-dependent carbon isotope fractionation in the dinoflagellate Alexandrium tamarense

This chapter has been published: Wilkes E. B., Carter S. J., and Pearson A. (2017) CO2-dependent carbon isotope fractionation in the dinoflagellate Alexandrium tamarense. Geochimica et Cosmochimica Acta, 212, 48−61.

Abstract

The carbon isotopic composition of marine sedimentary organic matter is used to resolve long-term histories of pCO2 based on studies indicating a CO2-dependence of photosynthetic carbon isotope

13 fractionation (εP). It recently was proposed that the δ C values of dinoflagellates, as recorded in fossil dinocysts, might be used as a proxy for pCO2. However, significant questions remain regarding carbon isotope fractionation in dinoflagellates and how such fractionation may impact sedimentary records throughout the Phanerozoic. Here we investigate εP as a function of CO2 concentration and growth rate in the dinoflagellate Alexandrium tamarense. Experiments were conducted in nitrate-limited chemostat

-1 cultures. Values of εP were measured on cells having growth rates (μ) of 0.14 to 0.35 d and aqueous carbon

-1 2 dioxide concentrations of 10.2 to 63 μmol kg and were found to correlate linearly with μ/[CO2(aq)] (r =

0.94) in accord with prior, analogous chemostat investigations with eukaryotic phytoplankton. A maximum fractionation (εf) value of 27‰ was characterized from the intercept of the experiments, representing the first value of εf determined for an algal species employing Form II RubisCO—a structurally and catalytically distinct form of the carbon-fixing enzyme. This value is larger than theoretical predictions for

Form II RubisCO and not significantly different from the ~ 25‰ εf values observed for taxa employing

Form ID RubisCO. We also measured the carbon isotope contents of dinosterol, hexadecanoic acid, and phytol from each experiment, finding that each class of biomarker exhibits different isotopic behavior. The

13 apparent CO2-dependence of εP values in our experiments strengthens the proposal to use dinocyst δ C values as a pCO2 proxy. Moreover, the similarity between the εf value for A. tamarense and the consensus value of ~ 25‰ indicates that the CO2-sensitivity of carbon isotope fractionation saturates at similar CO2 levels across all three ecologically prominent clades of eukaryotic phytoplankton. This continuity of εf

47 across taxa may help to explain why there is no coherent signature of phytoplankton evolutionary succession in Phanerozoic carbon isotope records.

Introduction

Marine eukaryotic phytoplankton produce organic matter that is depleted in 13C relative to ambient dissolved carbon dioxide (CO2(aq)). This photosynthetic carbon isotope fractionation, known as εP, is approximately equal to the isotopic difference between extracellular CO2(aq) and total biomass (εP ≈

13 13 δ CCO2 - δ Cbiomass; Freeman and Hayes, 1992). The study of factors governing the generation and preservation of εP signals has a long history, driven by the potential for illuminating paleo-environmental conditions (Degens et al., 1968; Arthur et al., 1985; Hayes et al., 1989; Francois et al., 1993; Laws et al.,

1995). Indeed, variations in sedimentary carbon isotope records (δ13C values), particularly in compound- specific or compound-class applications, are used to reconstruct past atmospheric CO2 levels (e.g., Jasper and Hayes, 1990; Pagani et al., 2005; Pagani, 2014; Heureux and Rickaby, 2015) and to study the production and burial of organic matter through time (e.g., Hayes et al., 1999; Katz et al., 2005; Krissansen-

Totten et al., 2015).

Rigorously-controlled continuous culture (chemostat) studies of modern phytoplankton taxa strengthen the interpretation of sedimentary organic matter archives by investigating the carbon isotopic responses of modern phytoplankton taxa to CO2 concentrations ([CO2(aq)]) and growth rates (μ). Chemostat experiments with eukaryotes indicate that εP values decrease linearly as μ increases or [CO2(aq)] decreases

(i.e., as a function of μ/[CO2(aq)]; Laws et al., 1995; Laws et al., 1997; Bidigare et al., 1997; Popp et al.,

-1 1998a). The linear relationship holds at CO2 concentrations exceeding approximately 10 μmol kg and is in accord with a theoretical model of carbon isotope fractionation that first was developed to describe carbon isotope fractionation in C3 higher plants (Farquhar et al., 1982, 1989; Francois et al., 1993; Goericke et al.,

1994). When applied to marine phytoplankton, the model postulates that cells acquire and lose CO2 by diffusive transport with a small associated fractionation (εt ≈ 0.7‰; Popp et al., 1989; Hayes, 1993) and that the dominant driver of carbon isotope fractionation is the intrinsic kinetic discrimination accompanying

48 fixation of CO2 by the enzyme RubisCO (ribulose-1,5-bisphosphate carboxylase/oxygenase; Goericke et al., 1994).

Within this framework, the maximum attainable value of εP is presumed to correspond to the flux- weighted average fractionation by RubisCO and β-carboxylase enzymes (Francois et al., 1993; Cassar and

Laws, 2007). Known as the net enzymatic fractionation (εf), this maximum isotopic fractionation is expressed in vivo when virtually all of the CO2 that diffuses into the cell diffuses back out again due to either high CO2 availability or low cellular carbon demand. In practice, values of εf for different phytoplankton species are inferred from the εP-intercept of chemostat studies as the ratio μ/[CO2(aq)] approaches zero. Such values have only been studied systematically in nutrient-limited continuous cultures using two species of diatoms (Phaeodactylum tricornutum, Laws et al., 1995, 1997 and Porosira glacialis,

Popp et al., 1998a), one species of haptophyte algae (Emiliania huxleyi, Bidigare et al., 1997), and a cyanobacterium (Synechococcus sp., Popp et al., 1998a). The three eukaryotic organisms all converged on an εf value of approximately 25‰, while Synechococcus, a prokaryote, displayed a smaller εf value of 17‰.

The difference may reflect active uptake of bicarbonate by Synechococcus (enriched by ~8‰ relative to

CO2 at experimental temperatures), or a taxon-specific difference in RubisCO fractionation.

Marine phytoplankton communities have undergone significant changes through Earth’s history, with different phytoplankton groups rising to ecological prominence as environmental pressures changed over time (reviewed by Falkowski et al., 2004; Knoll et al., 2007; Kodner et al., 2008; Rickaby, 2015). The influence of this succession on the isotopic composition of sedimentary organic matter remains unclear.

Different phytoplankton groups employ catalytically and phylogenetically distinct forms of the enzyme

RubisCO: Form I and Form II (Tabita et al., 2007, 2008). These forms evolved under different atmospheric concentrations of the two competitive gaseous substrates for RubisCO, CO2 and O2 (Badger et al., 1998;

Young et al., 2012). Accordingly, RubisCOs display significant taxonomic variability in kinetic properties

(Badger et al., 1998; Tcherkez et al., 2006; Young et al., 2016) and in vitro carbon isotope fractionations

(Roeske and O’Leary, 1984, 1985; Guy et al., 1993; Robinson et al., 2003; Tcherkez et al., 2006; Boller et al., 2011, 2015). These innovations have the potential to influence cellular δ13C values. However, the species studied in well-controlled chemostat experiments (e.g., Laws et al., 1995) represent only relatively

49 modern haptophyte and diatom species that use RubisCO Form ID; i.e., despite the widespread supposition that an εf value of ~25‰ can represent the temporal succession of marine eukaryotic phytoplankton, it is based only on recently-evolved taxa.

Dinoflagellates are unicellular eukaryotic phytoplankton that are ecologically important throughout the Meso- and Cenozoic, occupying a variety of environmental niches. Approximately 15% of modern dinoflagellate species can form dormant, fossilizable resting cysts as part of their life cycle (Head, 1996).

These resting cysts accumulate on the ocean floor, allowing the dinoflagellates to hatch and re-enter the motile stage under favorable conditions. The fossil cyst record provides robust evidence for dinoflagellates dating to the Triassic (Sluijs et al., 2005; Delwiche, 2007), with cyst species diversity reaching a maximum in the Cretaceous (Katz et al., 2007; Hackett et al., 2004). Dinosterane molecular biomarkers have been detected in rocks as old as the Neoproterozoic, however, suggesting the origins of the group may be very ancient (Robinson et al., 1984; Summons and Walter, 1990; Moldowan and Talyzina, 1998; Fensome et al., 1999). Basal (peridinin-containing) dinoflagellates with plastids derived by secondary endosymbiosis are unique among eukaryotes for possessing Form II RubisCO (Morse et al., 1995; Whitney et al., 1995;

Rowan et al. 1996). This form of the enzyme was acquired by lateral gene transfer from a proteobacterium

(Morse et al., 1995; Whitney and Andrews, 1998). Dinoflagellates with plastids derived by tertiary endosymbiosis or from kleptoplasty have Form I RubisCOs (Schnepf and Elbrachter, 1999; Beardall and

Raven, 2016). While several studies have measured carbon isotope fractionation as a function of CO2 availability in batch cultures of various dinoflagellates (Burkhardt et al., 1999a; Rost et al., 2006; Hoins et al., 2015) and in a modified continuous culture system (Hoins et al., 2016a), significant questions remain about carbon isotope fractionation in dinoflagellates and how such fractionation may impact sedimentary records throughout the Phanerozoic.

It remains unclear how the presence of Form II RubisCO influences the overall expression of carbon isotope fractionation in dinoflagellates due to its intrinsic fractionating and catalytic properties.

Form II RubisCO’s low carbon dioxide/oxygen selectivity (Whitney and Andrews, 1998) and assumed low carbon dioxide affinity (Badger et al., 1998) are consistent with the idea that dinoflagellates require carbon concentrating mechanisms (CCMs) to sustain observed rates of photosynthesis (Reinfelder, 2011).

- Physiological evidence supports this view, indicating that some dinoflagellates can take up HCO3 in addition to CO2 and use carbonic anhydrase to interconvert these species (Dason et al., 2004; Rost et al.,

2006; Hoins et al., 2016b). CCMs are considered potentially confounding factors in the use of marine sedimentary carbon isotope records as proxies for paleo CO2 concentrations because CCMs may influence

13 13 the δ C value of intracellular CO2 (Laws et al., 2002). It remains currently unknown whether or how δ C values of fossil dinocysts and dinoflagellate biomarker compounds are affected by the activity of CCMs.

In this study, we investigated carbon isotope fractionation in the dinoflagellate Alexandrium tamarense. Experiments were performed under nitrate-limited conditions in a chemostat culture system mimicking that of Laws et al. (1995) to establish a value of εf for this organism – to our knowledge the fifth

13 εf value clearly characterized in the literature. We also measured the δ C values of hexadecanoic acid (C16:0 fatty acid), phytol, and dinosterol (4α, 23, 24-trimethyl-5α-cholest-22E-en-3β-ol) to explore how lipids from major lipid classes vary as a function of μ/[CO2(aq)].

Materials and Methods

Chemostat culture system

Alexandrium tamarense (CCMP 1771) was cultivated in a nitrate-limited chemostat system (Figure

B.1; BIOSTAT® B, Sartorius Stedim). Parameters were selected to recreate the conditions reported in Laws et al. (1995) and Bidigare et al. (1997). Continuous light was provided by cool white fluorescent daylight tubes, with a flux of 150  30 μmol photons m-2 s-1 (400-700 nm radiation; measured inside the filled growth chamber with a QSL-2100 quantum scalar sensor, Biospherical Instruments Inc.). Temperatures were maintained at 18.00 ± 0.01°C by circulating water from a temperature-controlled chiller through the glass jacket of the growth reservoir, monitored continuously using an in-process thermometer and data logger.

The vessel was stirred at 50 rpm.

The culture was continuously supplied with 0.2-μm sterile-filtered and autoclaved Gulf of Maine seawater, which was enriched with 100 μmol l-1 nitrate, 36 μmol l-1 phosphate, and metals and vitamins according to L1 medium (Guillard and Hargraves, 1993). Spent medium and cells were removed through an overflow, positioned to maintain the culture volume at 4 l. Growth rates (; d-1) were varied by adjusting

51 the pumping rate. Steady-state conditions for each experiment were confirmed by monitoring cell density, pH, residual phosphate and nitrate concentrations, and cell morphology. Cell counts were performed every

24-48 hours using a Sedgwick Rafter counting chamber and 1 ml culture suspensions fixed with Lugol’s solution (2% final concentration); cell concentrations were approximately 104 cells ml-1 across all experiments. Dissolved phosphate analyses were performed using the mixed molybdate colorimetric method (adapted from Strickland and Parsons, 1968), and residual nitrate concentrations were monitored using the resorcinol spectrophotometric method (Zhang and Fischer, 2006) on 0.2-μm-filtered samples collected from the growth chamber. Cells were inspected routinely for changes in morphology and motility using a light microscope. Size measurements for calculating surface area and biovolume were performed with a Zeiss Cell Observer microscope and Fiji software.

-1 Six experiments were completed with  values ranging from 0.14 to 0.35 d and dissolved CO2 concentrations ranging from 10.2 ± 0.5 to 63 ± 3 μmol kg-1 (Table B.1; Table 3.1). Carbonate system parameters and isotope measurements for each experiment are summarized in Table B.2. Final biomass samples for isotopic analysis were taken after cultures had completed at least four doublings under steady- state conditions. All daily sample removal volumes were < 5% of the culture volume to avoid significant perturbations of the steady state.

Table 3.1. Experimental conditions and bulk isotopic data for A. tamarense. Values reflect the means of steady-state conditions, 1σ.

a 13 13 b Expt. # [CO2(aq)] Μ δ CCO2 δ Cbiomass εp n (μmol kg-1) (d-1) (‰) (‰) (‰)

1 63 ± 3 0.24 ± 0.02 −34.2 ± 0.4 −59.3 ± 0.2 26.7 ± 0.5 6 2 23.2 ± 0.7 0.14 ± 0.01 −33.4 ± 0.5 −54.5 ± 0.6 22.3 ± 0.8 3 3 22.9 ± 1.2 0.24 ± 0.01 −28.9 ± 0.3 −49.2 ± 0.6 21.3 ± 0.7 4 4 25.0 ± 1.3 0.35 ± 0.01 −28.4 ± 0.3 −46.6 ± 0.4 19.4 ± 0.6 2 5 12.8 ± 0.4 0.24 ± 0.01 −25.43 ± 0.14 −41.9 ± 0.5 17.2 ± 0.5 5 6 10.2 ± 0.5 0.24 ± 0.02 −22.4 ± 0.6 −36.6 ± 1.4 14.7 ±1.6 3 aThe order in which the experiments were performed is: 3, 4, 2, 5, 6, 1 bReplicates of biomass

Carbonate System Chemistry

CO2 concentrations were maintained by bubbling with gas mixtures achieved by mixing pure CO2

13 (δ C = -38.58  0.03‰) with 4:1 N2:O2 in sequential dilution using mass flow controllers (Brooks

Instrument, USA) and rotameters.

The growth medium was monitored continuously with an in-process pH probe (EasyFerm Plus,

Hamilton). Samples for total dissolved inorganic carbon (DIC) and total alkalinity (TA) were collected every 24-48 hours during both the last days of approach to, and throughout the duration of, each steady state condition. DIC samples were collected with no headspace, poisoned with saturated mercuric chloride (0.02% v/v HgCl2) and stored at 4°C until analysis. Total DIC concentrations were measured by coulometric titration (UIC Model 5014 CO2 coulometer) with an AutoMate automatic acidification system at the University of Florida. TA was determined on filtered samples (0.45-μm cellulose acetate syringe filters). TA samples were analyzed in duplicate by Gran titration with 0.01 N HCl as the acid titrant, made up in a 0.7 M NaCl background to approximate the ionic strength of seawater (Gran, 1952;

Dickson et al., 2007). Certified reference materials supplied by A.G. Dickson (Scripps Institution of

Oceanography) were used to calibrate the acid titrant and to monitor precision. Estimated analytical uncertainties were  0.01 units (pH),  5-15 μmol kg-1 (DIC), and  15 μeq kg-1 (TA).

The program CO2SYS (Lewis and Wallace, 1998) adapted for MATLAB (van Heuven et al.,

2011) was used to calculate carbonate speciation using total DIC, pH, phosphate, salinity, and temperature. The apparent dissociation constants for carbonate and sulfuric acid were from Mehrbach et al. (1973), as refitted by Dickson and Millero (1987), and from Dickson (1990), respectively. Combined uncertainties in calculated [CO2(aq)] due to uncertainties in measured carbonate system parameters were estimated numerically using CO2SYS (following Dickson, 2010; Bevington and Robinson, 2003).

Bulk carbon isotopic analysis

Culture outflow was collected into dark, refrigerated, sterile overflow bottles. Cells were pelleted by centrifugation (4500 rpm) and stored in Falcon tubes at -80°C. Biomass samples were acidified with 1

13 N HCl to remove residual DIC and dried at 50°C prior to analysis. Values of δ Cbiomass were analyzed by an elemental analyzer interfaced to a continuous flow isotope ratio mass spectrometer (EA-IRMS)

(University of California Davis Stable Isotope Facility) relative to laboratory reference materials that had been previously calibrated against NIST Standard Reference Materials, with a long-term standard deviation of 0.2‰.

13 The carbon isotopic composition of DIC (δ CDIC) was measured at the University of Florida Light

Stable Isotope Mass Spec Lab. Measurements were performed using a Thermo Finnigan DeltaPlus XL isotope ratio mass spectrometer paired with a GasBench II, with a long-term precision of < 0.1‰ for sample and reference material measurements. The temperature-dependent fractionation relationship of Mook et al.

13 13 (1974) was used to calculate the isotopic composition of dissolved CO2 (δ CCO2) from δ CDIC and absolute temperature (Eq. 1; Rau et al., 1996; Burkhardt et al., 1999a).

13 13 δ CCO2 = δ CDIC + 23.644 – (9701.5/TK) (1)

All carbon isotopic compositions are reported in δ notation relative to Vienna PeeDee Belemnite

(V-PDB). Values of εP were calculated according to Eq. 2 (Freeman and Hayes, 1992; Goericke et al.,

1994). Standard deviations associated with replicate measurements were propagated following Ellison and

Williams (2012). 13 13 δ CCO2-δ Cbiomass εp= 13 (2) 1+δ Cbiomass/1000

Compound-specific δ13C analyses

Total lipid extracts (TLEs) were obtained using a Bligh and Dyer extraction (Bligh and Dyer, 1959) modified to use 5% trichloroacetic acid in place of water (Nishihara and Koga, 1986). TLEs were divided into two aliquots; one was methylated at 70°C in 0.5 mol l-1 KOH prepared in methanol, while the second was transesterified at 70°C in 5% (v/v) HCl in methanol. A fatty acid mixture containing hexadecanoic acid

(δ13C = -29.5‰) and nonadecanoic acid (δ13C = -31.7‰) was transesterified to fatty acid methyl esters

(FAMEs) in parallel with the TLE, using the same batch of methanol for all reactions.

Transesterified aliquots were dried (N2, 40°C) and separated on 130-270 mesh SiO2 gel, eluting

FAMEs in 9:1 hexane/ethyl acetate (v/v) and sterols in 3:1 hexane/ethyl acetate (v/v). The sterol fractions and the base-hydrolyzed TLEs each were derivatized to acetates (50°C, 1:1 acetic anhydride/pyridine). For carbon isotopic correction of the added acetate group, a mixture containing androstanol (δ13C = -33.4‰) and phytol (δ13C = -33.0‰) was derivatized alongside the samples using the same acetic anhydride/pyridine mixture.

Values of δ13C of individual FAMEs, dinosterol, and phytol were determined by gas chromatography–isotope ratio monitoring mass spectrometry (GC-IRMS: Thermo Scientific Delta V

Advantage interfaced to a Trace GC Ultra via a GC Isolink). FAME samples were injected in hexane and acetylated alcohol fractions were injected in isooctane via a programmable temperature vaporization inlet operated in splitless mode. FAMEs were separated on an Agilent 30 m HP5-MS capillary column according to Close (2012). The program for alcohol fractions was: 60°C for 1.0 minute, ramp to 280°C at 20°C min-

1, then ramp to 310°C at 1.1°C min-1. Dinosterol δ13C values were measured in both alcohol fractions (sterol fractions from transesterified TLE and the entire base-hydrolyzed TLE). Phytol δ13C values were measured

13 only from the base-hydrolyzed TLEs. Samples were co-injected with C32 n-alkane (δ C = -29.5‰) or C38 n-alkane (δ13C = -31.4‰) isotope reference materials (A. Schimmelmann, Indiana University). Each sample was measured 3-6 times and δ13C values were determined from runs with peak responses in the range 0.5V-

5V (m/z 44). The isotopic contribution of the derivative carbons was removed by isotope mass balance, and error was propagated from the 1σ uncertainty in the δ13C values of the derivative carbons. To confirm peak identifications, selected samples were analyzed by gas chromatography-mass spectrometry (GC-MS;

Agilent 6890/5973 GC/mass selective detector) equipped with the same 30 m HP5-MS column. GC conditions were as above.

The isotopic fractionation (εCO2-lipid) of the lipids relative to CO2 in the culture medium was calculated analogously to εP (Eq. 3). The differences in isotopic composition between the lipids and bulk

55 cellular biomass were calculated following Schouten et al. (1998) and Riebesell et al. (2000b) (Eq. 4 and

5). 13 13 δ CCO2-δ Clipid εCO2-lipid= 13 (3) 1+δ Clipids/1000

13 13 13 Δδ Cbiomass-lipid = δ Cbiomass- δ Clipid (4)

13 13 13 Δδ C16:0-lipid = δ C16:0- δ Clipid (5)

Results

Photosynthetic carbon isotope fractionation

Six chemostat culture experiments yielded values of εP ranging from 14.7 to 26.7‰ (Table 3.1).

The variation in εP is a strong linear function of the ratio μ/[CO2(aq)] (Equation. 6; Figure 3.1, solid black line).

μ 2 εp = -558.7 ( ) + 27.4 (r = 0.94, geometric mean regression; Equation 6) [CO2(aq)]

Geometric mean regression analysis was chosen to facilitate comparison with prior chemostat studies (Popp et al., 1998a). The linearity and quality of the data are similar to other straight-line relationships observed in chemostat incubations with the haptophyte E. huxleyi and the diatoms Phaeodactylum tricornutum and

2 Porosira glacialis. These earlier studies had εP-intercepts (εf values) ranging from 24.6 to 25.5‰ and r values of 0.75-0.87 (Popp et al., 1998a; Figure 3.2). By analogy with these previous experiments, the εP- intercept for A. tamarense, 27‰, is inferred to be the value for εf.

Figure 3.1. εp as a function of growth rate and CO2 concentration. Error bars represent ± 1σ propagated error. The geometric mean regression through all A. tamarense experiments (this study) is shown with a 2 solid black line (εp = -558.7(μ/[CO2(aq)]) + 27.4, r = 0.94, n = 6). The black envelope represents the 95% confidence interval for a linear regression through the six experiments. Each data point from Hoins et al. (2016a) corresponds to the mean of duplicate experiments. Growth rates from Hoins et al. (2016a) were converted to instantaneous growth rates following Riebesell et al. (2000a,b) to account for differences in the length of the photoperiod.

Figure 3.2. Comparison with eukaryotic species data from Popp et al. (1998a). (a) εp-μ/[CO2(aq)] relationships measured for 4 eukaryotic species grown in analogous chemostat incubations, modified from Popp et al., 1998a to include the results of this study. The lines represent geometric mean regression analysis (reduced major axis). (b) Plot of εp vs. the product of μ/[CO2(aq)] and volume-to-surface-area ratios for the four species shown in Fig. 3.2a. The line represents the geometric mean regression analysis from Popp et al., 1998a.

We assessed the robustness of this εf prediction in several ways. Omitting the two endpoints (points

1, 6, Figure 3.1) would yield an intercept of 25.3‰. Alternatively, omitting the individual data point that deviates most from the linear fit (point 2, Figure 3.1) would yield an intercept of 28.4‰. The 95% confidence interval for the intercept is 24.9-29.9‰ based on a linear regression through all six experiments

(thin black envelope, Figure 3.1). Thus, regardless of approach, the εf value is (a), not strongly dependent on any of our individual experiments or their associated uncertainties; (b), statistically indistinguishable from the ~25‰ intercept measured for E. huxleyi and the diatoms Phaeodactylum tricornutum and Porosira glacialis; and (c), within the 25-28‰ range classically assumed for marine eukaryotic phytoplankton based on theoretical calculations (Goericke et al., 1994; Raven and Johnston, 1991).

The differences in slope of the relationships between εP and μ/[CO2(aq)] for E. huxleyi, Porosira glacialis, and Phaeodactylum tricornutum (Figure 3.2a) can be explained by cell sizes and geometries

(Popp et al., 1998a). Cells with low surface area to volume (S/V) ratios are more sensitive to changes in growth rate and CO2 availability, displaying a steeper slope than cells with higher S/V ratios. The experiments of Popp et al. (1998a) can be represented with a single linear relationship by plotting εP versus the product of μ/[CO2(aq)] and volume/surface area (Popp et al., 1998a; solid line Figure 3.2b). A. tamarense can be adequately described as a rotational ellipsoid to calculate S/V (Olenina et al., 2006).

Correcting for measured S/V ratios (Table B.3) yields a relationship that plots above the Popp et al. (1998a) linear fit (Figure 3.2b). The cellular carbon budget of this taxon may be less sensitive to S/V than the diatoms and haptophytes from earlier studies due to a different cell volume-to-carbon relationship.

However, dinoflagellates typically have a greater carbon content for a given size relative to diatoms, possibly due to differences in vacuole size or the presence of organic vs. silica cell walls (Menden-Deuer and Lessard, 2000). This extra carbon density would worsen, rather than ameliorate, the slope discrepancy.

Alternatively, the outer structures of A. tamarense may be more permeable to CO2 than that of diatoms and haptophytes.

All of the εP values measured in this study are larger than in any prior culture experiments with dinoflagellates. εP values reported for eight dinoflagellate species range from approximately -1‰ to 14‰

-1 in nutrient-replete batch culture studies (n = 32; [CO2(aq)] = 0-50 μmol kg ; Burkhardt et al., 1999a,b; Rost

59 et al., 2006; Hoins et al., 2015). In nitrate-limited continuous cultures with two dinoflagellate species (but using an experimental design differing from Laws et al., 1995, see discussion), εP ranged from 9.5-13.2‰ and was insensitive to variable [CO2(aq)] (Figure3.1; Hoins et al., 2016a). The two εP values from this earlier study that were measured at maximum μ/[CO2(aq)] fall within the projected 95% confidence interval of the linear fit to our data, but values measured at smaller μ/[CO2(aq)] (higher [CO2(aq)] or slower growth rate) do not (Figure 3.1).

Lipid δ13C values

Most experiments yielded dinosterol δ13C values that were enriched relative to fatty acids and depleted relative to bulk biomass (Table 3.2). The magnitude of depletion relative to biomass

13 2 (Δδ Cbiomass-dinosterol) increased linearly with the ratio μ/[CO2(aq)], from 0.8 to 11.0‰ (r = 0.73; open squares

13 Figure 3.3a). Dinosterol δ C values displayed no clear CO2-dependent trend across experiments

13 (εCO2-dinosterol = 26.1  1.9‰; filled squares Figure 3.3a). Δδ C16:0-dinosterol also did not display a consistent

13 13 directionality or magnitude, with Δδ C16:0-dinosterol varying from -7‰ (dinosterol relatively enriched in C) to 3.6‰ (dinosterol relatively depleted in 13C). These fractionations fall within the observed range of

13 Δδ C16:0-sterol values from a prior survey of 14 marine and freshwater algal species: sterols from most

13 organisms were enriched in C by 0-8‰ relative to C16:0, with the exception of sterols from the dinoflagellate G. simplex, which were depleted relative to C16:0 by up to 3.9‰ (Schouten et al., 1998).

The isotopic fractionation between CO2 and C16:0 fatty acid (εCO2-C16:0) displayed a robust,

2 decreasing trend with μ/[CO2(aq)], ranging from 35.5 ± 0.8 to 22.5 ± 0.7‰ (r = 0.97; filled triangles Figure

13 3.3b). C16:0 fatty acids were depleted in C by 7.4‰ to 12‰ relative to bulk biomass, with a mean difference

13 of Δδ Cbiomass-C16:0 = 8.9  1.7‰ and no statistically significant trend (Table 3.2; open triangles Figure

3.3b). These fractionations partially fall outside the 0-9‰ range previously observed for n-C14-C18 n-alkyl

13 lipids from algal cultures (Hayes, 2001). The δ C values of C14:0 fatty acids were also analyzed for a subset of samples, and differed by no more than 1.4‰ relative to C16:0 (data not shown). Fatty acids with longer chain lengths are not reported because they were not sufficiently abundant or baseline-resolved to obtain quality measurements.

Table 3.2. Results of compound-specific isotope analysis of A. tamarense lipids (δ13C,‰)a

C16:0 13 13 13 13 13 Expt # Phytol Dinosterol Δδ Cbiomass-16:0 Δδ Cbiomass-dinosterol Δδ Cbiomass-phytol Δδ C16:0-dinosterol Δδ C16:0-phytol Fatty Acid

1 -67.0 ± 0.3 -61 ± 3 -60.1 ± 0.8 7.7 0.8 2.0 -7.0 -5.7 2 -66.5 ± 0.6 -52 ± 2 -60.3 ± 0.9 12.0 5.8 -2.2 -6.2 -14.2 3 -57.4 ± 0.4 -44.5 ± 1.3b -51.7 ± 0.8 8.2 2.5 -4.7 -5.7 -12.9 4 -56.5 ± 0.4 -44.4 ± 1.2 -50.9 ± 0.8 9.9 4.2 -2.2 -5.7 -12.1

5 -50.1 ± 0.4 -42.1 ± 1.2 -50.1 ± 0.9 8.2 8.2 0.2 0.0 -8.0 6 -44.0 ± 0.4 -37 ± 2 -47.6 ± 1.2 7.4 11.0 0.3 3.6 -7.0

Mean 8.9 5.4 -1.1 -3.5 -10.0 SD 1.7 3.8 2.4 4.3 3.5 aValues are mean value ± 1σ propagated error bPhytol peak co-eluted with another compound in this sample.

Figure 3.3. Isotopic fractionations of lipids from three biomarker classes: (a) dinosterol, (b) C16:0 fatty acid, and (c) phytol. Fractionations were calculated relative to CO2 (εCO2-lipid; filled symbols; left axis) and 13 biomass (∆δ Cbiomass-lipid; open symbols; right axis), plotted as a function of μ/[CO2(aq)]. Error bars are mean values ± 1σ. Linear fits are shown when r2 > 0.5.

εCO2-phytol values displayed a weak linear dependence on μ/[CO2(aq)] and decreased from 28.9 ± 3 to

15.1 ± 2‰ (r2 = 0.56; filled circles Figure 3.3c). Phytol δ13C values were offset from biomass δ13C values

13 by -4.7‰ to 2‰ but without any significant trend (Δδ Cbiomass-phytol; Table 3.2; open circles Fiure. 3.3c); the values fell within error of biomass δ13C values for four of the six samples. Phytol partially co-eluted with a minor, isotopically-enriched but unidentified compound during GC- irMS analysis. Incomplete

13 chromatographic separation may account for the lack of a consistent directionality of offset in Δδ Cbiomass-

13 13 phytol. Phytol δ C values were enriched in C relative to C16:0 fatty acid in all experiments by 5.7-14.2‰, exceeding the ~0-5‰ enrichments that have been measured in prior culture studies with other algae

(Schouten et al., 1998; Riebesell et al., 2000b).

Discussion A variety of factors can influence carbon isotope fractionation in phytoplankton, including properties of the carboxylating enzyme, illumination and daylength, [CO2(aq)], growth rate, and the presence or absence of various carbon concentrating mechanisms (e.g., Burkhardt et al., 1999a; Laws et al., 2001;

Rost et al., 2002; Pagani, 2014). In the following, we examine our experimental results for A. tamarense with respect to these controls and their implications for interpreting carbon isotope records. We conclude by discussing the outlook for dinoflagellate-derived pCO2 proxies.

Fractionation by RubisCO

Understanding the fractionation associated with the carbon-fixing enzyme RubisCO is thought to be central to interpreting algal δ13C values, yet information on dinoflagellate Form II RubisCO is relatively scarce. Form II RubisCO has a distinct primary structure (L2; two large subunits) from the Form I RubisCO superfamily (L8S8; eight large and eight small subunits), with only ~30% amino acid sequence identity to

Form I in the large-type subunits (Tabita et al., 2007). Genes encoding dinoflagellate Form II RubisCO are

~65% identical to RubisCO from the proteobacterium Rhodospirillum rubrum and are believed to have been acquired from proteobacteria via lateral gene transfer (Rowan et al., 1996).

The only established catalytic property for dinoflagellate RubisCO is its ability to distinguish between competing CO2 and O2 substrates (Lilley et al., 2010). This characteristic is known as the specificity parameter, Sc/o= VCO2KO2/VO2KCO2, where V and K are maximum reaction rates and half saturation constants, respectively (Jordan and Ogren, 1981; Watson and Tabita 1997). Form II RubisCO from the dinoflagellate Amphidinium carterae has a low specificity of only Sc/o=37 (Whitney and Andrews, 1998), whereas green algae (Form IB) exhibit Sc/o values ranging from 54 to 83, and non-green algae (Form ID) are characterized by values ranging from 57 to 238 (Badger, 1998; Badger and Bek, 2008; Young et al.,

2016). Tcherkez et al. (2006) theorized that differences in CO2-selectivity among RubisCOs are related to differences in transition state structure that also determine their carbon isotope selectivity. RubisCOs with low Sc/o are predicted to have a transition state for CO2 addition that is more reactant-like than product-like, leading to weaker binding of the intermediate, a higher CO2-saturated catalytic turnover rate (kcat), and a smaller intrinsic isotope fractionation (Tcherkez et al., 2006; Tcherkez, 2013). In agreement, a relatively small carbon isotope effect has been confirmed in vitro for proteobacterial Form II RubisCO (εRubisCO = 18- 23‰, Sc/o=9-15; Guy et al., 1993; McNevin et al., 2006; Robinson et al., 2003; Roeske and O’Leary, 1985).

The isotope effect for dinoflagellate Form II RubisCO has not yet been measured because it is notoriously unstable outside of the cell (Rowan et al., 1996, Lilley et al., 2010).

The above kinetic and taxonomic arguments predict that εRubisCO and therefore εf should be significantly smaller in dinoflagellates than the observed εf values of ~ 25‰ for algae containing form ID

RubisCO, e.g., diatoms. However, our experiments do not agree with this prediction: the whole-cell A. tamarense εf value of 27‰ is clearly distinct from the ~ 19-23‰ in vitro fractionation associated with proteobacterial Form II RubisCO. Instead, it is surprisingly similar to or slightly larger than the εf values inferred from chemostat studies with diatoms and haptophytes (Popp et al., 1998a). Future in vitro characterization of RubisCO purified from A. tamarense would clarify the relationships among Sc/o, εRubisCO, and εf in this species and represent an important next step for evaluating existing models.

Inorganic carbon acquisition

RubisCO is specific for CO2 as its carbon substrate, but it is debated whether carbon initially enters

- dinoflagellate cells as CO2 or HCO3 , or as a mixture of the two (Dason et al., 2004; Rost et al., 2006; Ratti et al., 2007; Lapointe et al., 2008, Brading et al., 2013). Although A. tamarense may possess an inducible

CCM, as is common in many algal species, we suggest that the large εP values and CO2-sensitivities in our experiments indicate that CO2 must substantially contribute to total inorganic carbon uptake under the low

µ/[CO2(aq)] conditions achieved in this study. Expressed values of εP as large as 27‰ are unlikely to be

- - compatible with significant HCO3 uptake because HCO3 is isotopically enriched by ~8-10‰ relative to

CO2, implying the cell would require a net enzymatic isotope effect of up to 37‰ for carbon fixation.

Even under existing models of isotope fractionation during carbon transport and fixation (Francois et al., 1993; Laws et al., 1995), and assuming CO2 is the sole inorganic carbon source entering the cell, >

90% would need to leak out again to express values of εP that approach εf. The largest value of εP in our study (26.7‰) nearly equals the inferred magnitude of εf, i.e., the cells approach infinite leakiness. Dominant speciation as CO2 can also help to explain the CO2-related changes in εP. For the three

-1 experiments conducted at a fixed growth rate of 0.24 d but variable [CO2(aq)], values of εp spanned 14.7-

26.7‰. Because all other culture parameters were held constant, this variability must reflect a response to changing CO2 rather than to another parameter.

Despite this apparent CO2-dependence, most marine phytoplankton (including dinoflagellates) are thought to possess some form of CCM to enhance the concentration of CO2 near the active site of RubisCO.

Significant uncertainties remain regarding the nature of the CCM in dinoflagellates (Giordano and Raven,

2005; Reinfelder, 2011). CCMs sometimes include active transport of inorganic carbon into the cell from the external medium. Alternatively, CCMs can be localized to different subcellular compartments and may

- be facilitated by a variety of enzymes, including carbonic anhydrase to rapidly equilibrate CO2 and HCO3 .

Our empirical εP values display a linear correlation with μ/[CO2(aq)], consistent with uncatalyzed, diffusive entry of CO2 into the cell; this suggests that any expression of CCMs in this taxon may be primarily subcellular or inducible under higher [CO2(aq)] conditions. We also note that changing the experimental

-1 growth rate () from 0.14 to 0.35 d at a constant [CO2(aq)] shifted εP values by approximately 3‰; i.e.,

there is a relatively small dependence of εP on µ, in contrast to the larger apparent dependence of εP on

[CO2(aq)] at constant µ. The most parsimonious explanation for these observations is that CO2 is diffusing into the cell under the conditions explored in our study, although the isotopic results alone do not preclude the use of extracellular carbonic anhydrase, an inducible CCM, or active CO2 transport (Laws et al., 1997;

Burkhardt et al., 1999a; Keller and Morel, 1999; Laws et al., 2002). For example, our results would also be compatible with a large leakage of carbon dioxide from an accumulated intracellular pool of inorganic carbon (Sharkey and Berry, 1985; Raven and Beardall, 2016).

Carbon acquisition also may differ between chemostat and batch incubations, complicating comparisons between our study and other physiological investigations into dinoflagellate CCMs. This is apparent isotopically through smaller εP values in dinoflagellate batch cultures (Burkhardt et al., 1999a,b;

Rost et al., 2006; Hoins et al., 2015) compared with our chemostat cultures. Experiments with E. huxleyi and P. tricornutum also consistently have yielded smaller εP values in batch conditions than in chemostat (Riebesell et al., 2000a,b; Rost et al., 2002; Burkhardt et al., 1999a; Popp et al., 1998a; Laws et al., 2001). These isotopic discrepancies have been theorized to reflect differing rates of active carbon uptake that are contingent on the factor controlling growth (i.e., nutrients in chemostat vs. light in batch cultures; Riebesell et al., 2000a,b; Cassar et al., 2006; Hoins et al., 2016a), as well as differences in illumination and daylength across studies (Rost et al., 2002). There is no consensus on which experimental design, if any, better emulates the environmental and ecological stresses of algae in a natural setting (Laws et al., 2001; Pagani,

2014).

Phytoplankton evolutionary succession

This study extends the experimental approach of Popp et al. (1998a) to an ecologically distinct group of eukaryotic phytoplankton with a long geological record. Our results confirm that carbon isotope fractionation negatively correlates with μ/[CO2(aq)] and converges on εf ≈ 25-28‰ for all three of the ecologically prominent modern clades of eukaryotic phytoplankton, providing compelling evidence that carbon isotope fractionation is similarly regulated in non-mineralizing (A. tamarense, E. huxleyi BT6),

calcifying (E. huxleyi B92/11), and silicifying (Porosira glacialis, Phaeodactylum tricornutum) phytoplankton.

Continuity of εf across taxa may explain why there is no coherent signature of phytoplankton evolutionary succession in Phanerozoic carbon isotope records (Popp et al., 1989; Freeman and Hayes,

1992; Hayes et al., 1999; Pagani, 2014; Krissansen-Totten et al., 2015). The sensitivity of εP to CO2 saturates at similar CO2 levels across taxonomic groups (i.e., εf is approximately constant), building confidence that the isotopic composition of bulk sedimentary marine organic matter may be used to coarsely constrain CO2 levels at different points in the geologic past. For example, during periods when the inorganic-organic offset varies with time, this variability may be pointed to as evidence that pCO2 falls below a particular threshold level of sensitivity dictated by εf (Freeman and Pagani, 2005; Pancost et al.,

2013). However, the cell size- and growth-rate-dependence of the εP vs. μ/[CO2(aq)] slope prevents the direct use of bulk carbon isotope records for precise, quantitative pCO2 reconstructions. None of the three eukaryotic clades considered here have undisputed fossil records prior to the Mesozoic. While dinosterane biomarker evidence supports a more primitive origin for dinoflagellates, earlier records must be interpreted with caution. Moreover, these maximum isotope fractionations expressed in vivo contradict recent in vitro measurements of the isotope effect associated with RubisCO from E. huxleyi and the diatom Skeletonema costatum (11.1 and 18.5‰, respectively; Boller et al., 2011;

Boller et al., 2015), indicating that our understanding of the mechanisms underpinning this signal may be incomplete.

Intracellular carbon isotope fractionation

Dinosterol and its diagenetic product, dinosterane, are, with minor exceptions, diagnostic for dinoflagellates (Moldowan and Talyzina, 1998; Robinson et al., 1984; Volkman et al. 1993), yet the 13C content of dinosterol has not been explored in well-characterized chemostat conditions to assess its potential as a paleo-proxy. Our results indicate that dinosterane should not be used as a proxy to reconstruct pCO2

13 because dinosterol is not constantly offset from biomass in each experiment (i.e., Δδ Cbiomass-dinosterol varies;

13 13 Figure 3.3a). Therefore δ Cbiomass and εP cannot be calculated from the biomarker δ C value, in contrast,

13 13 for example, to the practice of using the value of alkenone δ C plus a constant 4.2‰ to determine δ Cbiomass and εP (Popp et al., 1998b; Pagani et al., 2005). However, our experiments show that the difference between

13 dinosterol and total biomass is systematic: as μ/[CO2(aq)] increases, Δδ Cbiomass-dinosterol linearly increases. In systems in which biomass δ13C values can be independently constrained – for example, using fossil organic dinoflagellate cysts (dinocysts) – dinosterol or dinosterane δ13C values may then prove useful to confirm

13 that biomass δ C values are responding to μ/[CO2(aq)] as predicted.

Our experiments also revealed systematic isotopic relationships for C16:0 fatty acids and phytol as a function of μ/[CO2(aq)]. Unfortunately, these compounds and their sedimentary derivatives are ubiquitous and provide little additional diagnostic information with respect to the sedimentary record. Our results caution that fatty acids may sometimes be significantly more depleted relative to biomass than previously recognized (Schouten et al., 1998; Hayes, 2001), and that these offsets are unpredictable even when few variables are adjusted between experiments. The dissimilar fractionation patterns observed between the two isoprenoid lipids, dinosterol and phytol, are surprising because both are derived from the isoprenoid precursor isopentenyl diphosphate (IPP), and transcriptome studies imply a single biosynthetic route to IPP in dinoflagellates (via the MEP/DOXP pathway, Bentlage et al., 2015; Lichtenthaler, 1999; Rohmer, 2003).

The IPP precursors may have different isotopic starting points in A. tamarense due to compartmentalization.

Phytol is synthesized in the chloroplast, while dinosterol may be synthesized from a cytosol pool subject to different divisions of carbon flow (e.g., Cvejić and Rohmer 2000; Hayes, 2001).

Dinoflagellate-derived pCO2 proxies

Hoins et al. (2015) proposed that the δ13C values of fossil dinocysts may be used as a species-

13 specific pCO2 proxy. Our findings generally support this application. Bulk biomass δ C values display a clear CO2-dependence and a larger range of εp values than previously recognized, indicating that the proxy should be sensitive over a wide range of pCO2 before saturating. Our calibration is especially favorable for the outlook of a dinoflagellate pCO2 proxy because it was obtained under nutrient-limited growth (nitrate limitation). Cyst formation is thought to be triggered by nutrient limitation at the end of blooms (Anderson et al., 1984; Ellegaard et al., 1998), suggesting that fossilized dinocysts may record growth conditions

similar to those of chemostat cultures. A prior continuous culture study with dinoflagellates concluded the opposite: that nitrogen-limitation eliminates the CO2 sensitivity of εP (Hoins et al., 2016a; Figure 3.1). This contradictory outcome likely arose from differences in experimental design (summarized in Table B.4) rather than taxonomy because both studies included dinoflagellates from the Alexandrium tamarense species complex (John et al., 2014). While our experiments used a close analogue of the systems and experimental conditions employed by the Popp et al. (1998a) suite of experiments (Figure B.1), Hoins et al. (2016a) used a modified system, with mixing achieved by rocking on a shaker table and experimental

CO2 levels established through pre-aeration of the medium with tank gas (for more details see Van de Waal et al., 2014; Eberlein et al., 2016). No single difference identified between the two experiments clearly accounts for the lack of CO2-sensitivity measured by Hoins et al. (2016a). Notably, the Hoins et al. (2016a) experiment used a 16:8 hour daylength instead of continuous light, so their system may be more precisely described as a cyclostat (Laws et al., 1995; Laws et al., 2001). Daylength is known to independently influence εP values, even after accounting for changes in instantaneous growth rate (Rost et al., 2002). For batch cultures with E. huxleyi and diatoms, a 16:8 h light:dark cycle yielded εP values that were smaller by

6-8‰ compared to continuous light conditions (Burkhardt et al., 1999b; Rost et al., 2002), or approximately the offset between Hoins et al., (2016a) and our study. It is also possible that chemical gradients formed in the Hoins et al. (2016a) system because it was mixed only by gentle rocking, rather than by the continuous bubbling and stirring employed in our system. Chemical gradients would violate the assumption of steady- state conditions and drive εP to smaller values (e.g., Laws et al., 2002). Alternatively, however, the stirring and bubbling used in our system may have stressed the cells since dinoflagellates are sensitive to turbulence, although the strain used in our study has been shown to be less vulnerable to turbulence than other dinoflagellates (Sullivan and Swift, 2003). Modern field calibrations in different nutrient regimes would help to clarify which observations in the laboratory translate into the context of the natural environment.

Before the dinocyst proxy can be recommended for the reconstruction of pCO2, a variety of outstanding issues need to be resolved. It is currently unknown how the δ13C values of motile cells relate to the δ13C of their associated dinocysts, and whether any of the physiological changes during cyst formation might compromise this signal (Hoins et al., 2015). A constant or predictable isotopic offset between the

two lifecycle stages needs to be demonstrated through controlled culture studies. A potential advantage of the dinocyst approach is that cysts can be identified to the species level and related to motile-stage species with known cellular geometries. Similarly, cysts may be sorted by size prior to analysis to potentially eliminate some of the complications associated with variable cell sizes and carbon contents (Hoins et al.,

2015).

This approach may be hampered by many of the same uncertainties present for other algal pCO2 proxies. Namely, it is challenging to infer the growth rates of cells without an effective nutrient proxy. As one potential way forward, we suggest that the median size of the dinocyst community from a given sediment sample could be measured in parallel with the cyst species used for isotope analyses. Core-top studies in modern environments suggest that the median size of the dinocyst community is correlated with nitrate and phosphate concentrations in surface waters for a given sea-surface temperature (Chen et al.,

2011), and as such might provide clues about this parameter in the geologic past. Application of this proxy will also require independent constraints on carbonate chemistry and sea surface temperature, as with other established algal proxies. We recommend that field studies involving sediment traps, core top sediments, and/or mesocosm experiments be carried out under bloom and non-bloom conditions to assess whether the

13 dinocyst δ C signal monitors atmospheric CO2, bloom dynamics, or another parameter. These questions may also be explored by testing whether fossil dinocysts can accurately reconstruct the pCO2 swings associated with recent atmospheric changes (e.g., Pagani et al., 2002) or glacial-interglacial cycles.

We suggest that dinosterane may provide a complementary tool for paleo-pCO2 reconstructions

13 based on dinocysts. As μ/[CO2(aq)] increases, our experiments predict that Δδ Cbiomass-dinosterol should also increase. These two pools may be accessed in a paleo-oceanographic context via fossil dinocysts (as a proxy

13 13 for δ Cbiomass values) and dinosterane (as a proxy for δ Cdinosterol) recovered from coeval sedimentary deposits. The offset between these two reservoirs through time could provide independent confirmation that

13 dinoflagellate δ C values are responding to μ/[CO2(aq)] rather than some other variable. However, we caution that dinosterol and its diagenetic products may be produced by many dinoflagellate species in a geological sample.

Conclusions

We have used nitrate-limited chemostat culture experiments to infer that εf is 27‰ for the marine dinoflagellate Alexandrium tamarense. This value is larger than theoretical predictions for Form II

RubisCO and is not significantly different from the εf values observed for more recently-evolved taxa employing Form ID RubisCO. Thus, we confirm that the temporal succession of modern eukaryotic phytoplankton (minimally Mesozoic to present) is well represented by an εf value in the range 25-28‰, helping to explain the broad uniformity of carbon isotope fractionation between organic and inorganic pools

13 observed through time. Dinoflagellate bulk biomass δ C values clearly responded to [CO2(aq)] and displayed a robust linear relationship with μ/[CO2(aq)]. These findings support a recent proposal to use fossil

13 dinocyst δ C values for constraining CO2 in the geologic past, although we recommend additional proxy validation studies before this approach can be applied with confidence. We also propose that the offset between dinosterane and dinocyst δ13C values through time may provide an additional, independent confirmation that the dinocyst proxy responds to μ/[CO2(aq)] as expected.

Acknowledgements

This material is based upon work supported by the National Science Foundation Graduate Research

Fellowship under Grant No. DGE1144152 to E.B.W. Any opinion, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. We thank Katie Mabbott for providing assistance with laboratory analyses and culture maintenance, Douglas Richardson for assisting with cell size measurements, Joe

Werne for editorial handling, and four anonymous reviewers for their valuable comments. A.P. and S.J.C. are funded by the Gordon and Betty Moore Foundation and by NASA-NAI CAN6 (PI Roger Summons,

MIT). In memory of Mark Pagani and John M. Hayes.

References

Anderson D. M., Kulis D. M. and Binder B. J. (1984) Sexuality and cyst formation in the dinoflagellate Gonyaulax tamarensis: cyst yield in batch cultures. J. Phycol. 20, 418-425. Arthur M. A., Walter D. E., and Claypool G. E. (1985) Anomalous 13C enrichment in modern marine organic carbon. Nature 315, 216-218. Badger M. R. and Bek E. J. (2008) Multiple Rubisco forms in proteobacteria: their functional significance in relation to CO2 acquisition by the CBB cycle. J. Exp. Bot. 59, 1525-1541. Badger M. R., Andrews T. J., Whitney S. M., Ludwig M., Yellowlees D.C., Leggat W., and Dean Price G. (1998) The diversity and coevolution of Rubisco, plastids, pyrenoids, and chloroplast-based CO2- concentrating mechanisms in algae. Can. J. Bot. 76, 1052-1071. Beardall J. and Raven J. A. (2016) Carbon acquisition by microalgae. In The Physiology of Microalgae (eds. M. A. Borowitzka, J. Beardall, and J. A. Raven). pp. 89-99, Springer, New York. Bentlage B., Rogers T. S., Bachvaroff T. R., and Delwiche C. F. (2015) Complex ancestries of isoprenoid synthesis in dinoflagellates Bevington P. R. and Robinson D. K. (2003) Data reduction and error analysis for the physical sciences, 3rd Edition. McGraw-Hill, New York, pp. 1-65. Bidigare R. R., Fluegge A., Freeman K. H., Hanson K. L., Hayes J. M., Hollander D., Jasper J. P., King L. L., Laws E. A., Milder J., Miller F. J., Pancost R., Popp B. N., Steinberg P. A., and Wakeham S. G. (1997) Consistent fractionation of 13C in nature and in the laboratory: Growth-rate effects in some haptophyte algae. Global Biogeochem. Cy. 11, 279-292. Bligh E. G. and Dyer W. J. (1959) A rapid method of total lipid extraction and purification. Can. J. Biochem. Physiol. 37, 911-919. Boller A. J., Thomas P. J., Cavanaugh C. M., and Scott K. M. (2011) Low stable carbon isotope fractionation by coccolithophore RubisCO. Geochim. Cosmochim. Ac. 75, 7200-7207. Boller A. J., Thomas P. J., Cavanaugh C. M., and Scott K. M. (2015) Isotopic discrimination and kinetic parameters of RubisCO from the marine bloom-forming diatom, Skeletonoma costatum. Geobiology 13, 33-43. Brading P., Warner M. E., Smith D. J., and Suggett D. J. (2013) Contrasting modes of inorganic carbon acquisition amongst Symbiodinium (Dinophyceae) phylotypes. New Phytol. 200, 432-442.

Burkhardt S., Riebesell U., and Zonervan I. (1999a) Effects of growth rate, CO2 concentration, and cell size on the stable carbon isotope fractionation in marine phytoplankton. Geochim. Cosmochim. Acta 63, 3729-3741. Burkhardt S., Riebesell U., and Zonervan I. (1999b) Stable carbon isotope fractionation by marine phytoplankton in response to daylength, growth rate, and CO2 availability. Mar. Ecol. Prog. Ser. 184, 31-41. Cassar N., Laws E. A., and Popp B. N. (2006) Carbon isotopic fractionation by the marine diatom Phaeodactylum tricornutum under nutrient- and light-limited growth conditions. Geochim. Cosmochim. Ac. 70, 5323-5335.

Cassar N. and Laws E. A. (2007) Potential contribution of β-carboxylases to photosynthetic carbon isotope fractionation in a marine diatom. Phycologia 46, 307-314. Chen B., Irwin A. J., and Finkel Z. V. (2011) Biogeographic distribution of diversity and size-structure of organic-walled dinoflagellate cysts. Mar. Ecol. Prog. Ser. 425, 35-45. Close H. G. (2012) Size-related isotopic heterogeneity in lipids from the marine water column. Doctoral dissertation, Harvard University.

Cvejić J. H. and Rohmer M. (2000) CO2 as main carbon source for isoprenoid biosynthesis via the mevalonate-independent methylerythritol 4-phosphate route in the marine diatoms Phaeodactylum tricornutum and Nitzschia ovalis. Phytochemistry 53, 21-28. Dason J. S., Huertas I. E., and Colman B. (2004) Source of inorganic carbon for photosynthesis in two marine dinoflagellates. J. Phycol. 40, 285-292. Degens E. T., Behrendt M., Gotthardt B., and Reppman E. (1968) Metabolic fractionation of carbon isotopes in marine plankton − II. Data on samples collected off the coasts of Peru and Ecuador. Deep-Sea Res. 15, 12-20. Delwiche C. F. (2007) The origin and evolution of dinoflagellates. In Evolution of Primary Producers in the Sea (eds. P. Falkowski and A. Knoll). Elsevier, Boston. pp. 191–205.

Dickson A. G. (1990) Standard potential of the reaction: AgCl(s) + 1/2H2(g)=Ag(s) + HCl(aq), and the - standard aciditiy constant of the ion HSO4 in synthetic seawater from 273.15 to 318.15 K.J. Chem. Thermodyn. 22,113-127. Dickson A. G. (2010) The carbon dioxide system in seawater: equilibrium chemistry and measurements. In Guide to Best Practices for Ocean Acidification Research and Data Reporting (eds. U. Riebesell, V. J. Fabry, L. Hansson and J. -P. Gattuso). Publications Office of the European Union, Luxembourg, pp. 17-40. Dickson A. G. and Millero F. J. (1987) A comparison of the equilibrium constants for the dissociation of carbonic acid in seawater media. Deep-Sea Res. Pt. I 34, 1733-1743.

Dickson A. G., Sabine C. L., and Christian J. R. (2007) Guide to best practices for ocean CO2 measurements. PICES Special Publication 3, 1-191. Eberlein T., Van de Waal D. B., Brandenburg K. M., John U., Voss M., Achterberg E. P., and Rost B. (2016) Interactive effects of ocean acidification and nitrogen limitation on two bloom-forming dinoflagellate species. Mar. Ecol. Prog. Ser. 543, 127-140. Ellegaard M., Kulis D. M., and Anderson D. M. (1998) Cysts of Danish Gymnodinium nolleri Ellegaard et Moestrup sp. Ined. (Dinophyceae): studies on encystment, excystment and toxicity. J. Plankton Res. 20, 1743-1755. Ellison S. L. R. and Williams A. (2012) Quantifying Uncertainty in Analytical Measurement 3rd Ed., EURACHEM/CITAC Working Group, pp. 1-133. Farquhar G. D., O’Leary M. H. B., and Berry J. A. (1982) On the relationship between carbon isotope discrimination and the intracellular carbon dioxide concentration in leaves. Aust. J. Plant Physiol. 9, 121-137.

Farquhar G. D., Ehleringer J. R., and Hubick K. T. (1989) Carbon isotope discrimination and photosynthesis. Annu. Rev. Plant. Phys. 40, 503-537. Falkowski P. G., Katz M. E., Knoll A. H., Quigg, A., Raven J. A., Schofield O., and Taylor F. J. R. (2004) The evolution of modern eukaryotic phytoplankton. Science 305, 354-360. Fensome R. A., Saldarriaga J. F., and Taylor M. F. J. R. (1999) Dinoflagellate phylogeny revisited: reconciling morphological and molecular based phylogenies. Grana 38, 66-80. Francois R., Altabet M. A., and Goericke R. (1993) Changes in the δ13C of surface water particulate organic matter across the subtropical convergence in the SW Indian Ocean. Global Biogeochem. Cy. 7, 627-644. Freeman K. H. and Hayes J. (1992) Fractionation of carbon isotopes by phytoplankton and estimates of ancient CO2 levels. Global Biogeochem. Cy. 6, 185–198.

Giordano M., Beardall J., and Raven J. A. (2005) CO2 concentrating mechanisms in algae: mechanisms, environmental modulation, and evolution. Anu Rev. Plant Biol. 56, 99−131. Goericke R., Montoya J., and Fry B. (1994) Physiology of isotopic fractionation in algae and cyanobacteria. In Stable Isotopes in Ecology and Environmental Science (eds. K. Lajtha and R. Michener). pp. 187–221. Blackwell Scientific Publications, Oxford. Gran G. (1952) Determination of the equivalence point in potentiometric titrations, Part II. Analyst 77, 661- 671. Guillard R. R. L. and Hargraves P. E. (1993) Stichochrysis immobilis is a diatom, not a chrysophyte. Phycologia 32, 234-236. Guy R. D., Fogel M. L., and Berry J. A. (1993) Photosynthetic fractionation of the stable isotopes of oxygen and carbon. Plant Physiol. 101, 37–47. Hackett J. D., Anderson D. M., Erdner D. L. and Bhattacharya D. (2004) Dinoflagellates: a remarkable evolutionary experiment. Am. J. Bot. 91, 1523-1534. Hayes J. M. (1993) Factors controlling 13C contents of sedimentary organic compounds: Principles and evidence. Mar. Geol. 113,111-125. Hayes J. M. (2001) Fractionation of carbon and hydrogen isotopes in biosynthetic processes. Rev. Mineral. Geochem. 43, 225-277. Hayes J. M., Popp B. N., Takigiku, R., and Johnson M. W. (1989) An isotopic study of biogeochemical relationships between carbonates and organic carbon in the Greenhorn Formation. Geochim. Cosmochim. Acta 53, 2961-2972. Hayes J. M., Strauss H., and Kaufman A. J. (1999) The abundance of 13C in marine organic matter and isotopic fractionation in the global biogeochemical cycle of carbon during the past 800 Ma. Chem. Geol. 161,103–125.

Head M. (1996) Modern dinoflagellate cysts and their biological affinities. In Palynology: priniciples and applications (eds. J. Jansonius and D.C. McGregor). Pp. 1197-1248, pp. 1197-1248, Dallas.

Heureux A. M. C. and Rickaby R. E. M. (2015) Refining our estimate of atmospheric CO2 across the Eocene-Oligocene climatic transition. Earth Planet. Sc. Lett. 409, 329-338. Hoins M., Van de Waal D. B., Eberlein T., Reichart G.-J., Rost B., and Sluijs A. (2015) Stable carbon isotope fractionation of organic cyst-forming dinoflagellates: Evaluating the potential for a CO2 proxy. Geochim. Cosmochim. Ac. 160, 267-276. Hoins M., Eberlein T., Groβmann C. H., Brandenburg K., Reichart G.-J., Rost B., Sluijs, A., and Van de Waal, D. B. (2016a) Combined effects of ocean acidification and light or nitrogen availabilities on 13C fractionation in marine dinoflagellates. PLoS ONE 11, e0154370.

Hoins M., Eberlein T., Van de Waal D. B., Sluijs A., Reichart G.-J., Rost B. (2016b) CO2-dependent carbon isotope fractionation in dinoflagellates relates to their inorganic carbon fluxes. J. Exp. Mar. Biol. Ecol. 481, 9-14.

Jasper J. P. and Hayes J. M. (1990) A carbon isotope record of CO2 levels during the late Quaternary. Nature 347, 462-464. John U., Litaker R. W., Montresor M., Murray S., Brosnahan M. L., Anderson D. M. (2014) Formal revision of the Alexandrium tamarense species complex (Dinophyceae) taxonomy: the introduction of five species with emphasis on molecular-based (rDNA) classification. Protist 165, 779-804. Jordan D. B. and Ogren W. L. (1981) Species variation in the specificity of ribulose bisphosphate carboxylase/oxygenase. Nature 291, 513-515. Katz M. E., Wright J. D., Miller K. G., Cramer B. S., Fennel K., and Falkowski P. G. (2005) Biological overprint of the geological carbon cycle. Mar. Geol. 217, 323–338. Katz M. E. Fennel K. and Falkowski P. G. (2007) Geochemical and biological consequences of phytoplankton evolution. In Evolution of Primary Producers in the Sea (eds. P. Falkowski and A. Knoll). pp. 133–163. Elsevier, Boston. Keller K. and Morel, F. M. M. (1999) A model of carbon isotopic fractionation and active carbon uptake in phytoplankton, Mar. Ecol. Prog. Ser. 182, 295-298. Knoll A. H., Summons R. E., Waldbauer J. R., Zumberge J. E. (2007) The geological succession of primary producers in the oceans. In Evolution of Primary Producers in the Sea (eds. P. Falkowski and A. Knoll). pp. 133–163. Elsevier, Boston. Kodner R. B., Pearson A., Summons R. E., and Knoll A. (2008) H. Sterols in red and green algae: Quantification, phylogeny, and relevance for the interpretation of geologic steranes. Geobiology 6, 411–420. Krissansen-Totton J., Buick R., and Catling D. C. (2015) A statistical analysis of the carbon isotope record from the Archean to Phanerozoic and implications for the rise of oxygen. Am. J. Sci. 315, 275-316. Lapointe M., MacKenzie T. D. B., and Morse D. (2008) An external δ-carbonic anhydrase in a free-living marine dinoflagellate may circumvent diffusion-limited carbon acquisition. Plant Physiol. 147, 1427-1436.

Laws E. A., Popp B. N., Bidigare R. R., Kennicutt M. C., and Macko S.A. (1995) Dependence of phytoplankton carbon isotopic composition on growth rate and CO2aq: Theoretical considerations and experimental results. Geochim. Cosmochim. Ac. 59,1131–1138.

Lewis E. and Wallace D. (1998) Program developed for CO2 system calculations: Oak Ridge TN Oak Ridge National Laboratory Environmental Sciences Division, v. 4735. Lichtenthaler H. K. (1999) The 1-deoxy-d-xylulose-5-phosphate pathway of isoprenoid biosynthesis in plants. Annu. Rev. Plant Physiol. Plant Mol. Biol. 50, 47-65. Lilley R. M., Ralph P. J., and Larkum A. W. D. (2010) The determination of activity of the enzyme Rubisco in cell extracts of the dinoflagellate alga Symbiodinium sp. By manganese chemiluminescence and its response to short-term thermal stress of the alga. Plant Cell Environ. 33, 995-1004. McNevin D. B., Badger M. R., Kane H. J., and Farquhar G. D. (2006) Measurement of (carbon) kinetic isotope effect by Rayleigh fractionation using membrane inlet mass spectrometry for CO2- consuming reactions. Funct. Plant Biol. 33, 1115-1128. Mehrbach C., Culberson C. H., Hawley J. E., and Pytkowicz R. M. (1973) Measurement of the apparent dissociation constants of carbonic acid in seawater at atmospheric pressure. Limnol. Oceanogr. 18, 897-907. Menden-Deuer S. and Lessard E. J. (2000) Carbon to volume relationships for dinoflagellates, diatoms, and other protist plankton. Limnol. Oceanogr. 45, 569-579. Moldowan J. M. and Talyzina N. M. (1998) Biogeochemical evidence for dinoflagellate ancestors in the early Cambrian. Science 5380, 1168-1170. Mook W. G., Bommerson J. C., and Staverman W. H. (1974) Carbon isotope fractionation between dissolved bicarbonate and gaseous carbon dioxide. Earth Planet. Sci. Lett. 22, 169-176. Morse D., Salois P., Markovic P., and Hastings J. W. (1995) A nuclear-encoded form II RuBisCO in dinoflagellates. Science 268, 1622-1624. Nishihara, M. and Koga, Y. (1986) Extraction and composition of polar lipids from the archaebacterium, Methanobacterium thermoacutorophicum: Effective extraction of tetraether lipids by an acidified solvent. J. Biochem. 101, 997-1005. Olenina I., Hajdu S., Edler L., Andersson A., Wasmund N., Busch S., Göbel J., Gromisz S., Huseby S., Huttunen M., Jaanus A., Kokkonen P., Ledaine I., and Niemkiewicz E. (2006) Biovolumes and size-classes of phytoplankton in the Baltic Sea HELCOM Balt. Sea Environ. Proc. 106, pp.1-144.

Pagani M., Freeman K. H., Ohkouchi N., and Caldeira K. (2002) Comparison of water column [CO2aq] with sedimentary alkeone-based estimates: A test of the alkenone-CO2 proxy Paleoceanography 17, 1069-1081. Pagani M., Zachos J., Freeman K. H., Bohaty S., and Tipple B. (2005) Marked decline in atmospheric carbon dioxide concentrations during the Paleogene. Science 309, 600-603.

Pagani M. (2014) Biomarker-based inferences of past climate: The alkenone pCO2 Proxy. In Treatise on Geochemistry 2nd Ed. (eds. H. Holland and K. Turekian). pp. 361-378, Elsevier, Amsterdam. Pancost R. D., Freeman K. H., Herrmann A. D., Patzkowsky M. E., Ainsaar L., and Martma T. (2013) Reconstructing Late Ordovician carbon cycle variations. Geochim. Cosmochim. Ac. 105, 433–454. Popp B. N., Takigiku R., Hayes J. M., Louda J. W., and Baker E. W. (1989) The post-Paleozoic chronology and mechanism of 13C depletion in primary marine organic matter. Am. J. Sci. 289, 436-454. Popp B. N., Laws E. A., Bidigare R. R., Dore J. E., Hanson K. L., and Wakeham S.G. (1998a) Effect of phytoplankton cell geometry on carbon isotopic fractionation. Geochim. Cosmochim. Ac. 62, 69– 77. Popp B. N., Kenig F., Wakeham S. G., Laws E. A., and Bidigare R. R. (1998b) Does growth rate affect ketone unsaturation and intracellular carbon isotopic variability in Emiliania huxleyi? Paleoceanography 13, 35-41.

Ratti S., Giordano M., and Morse D. (2007) CO2-concentrating mechanisms of the potentially toxic dinoflagellate Protoceratium reticulatum (Dinophyceae, Gonyaulacales). J. Phycol. 43, 693-701. Rau G. H., Riebesell U., and Wolf-Gladrow D. (1996) A model of photosynthetic 13C fractionation by marine phytoplankton based on diffusive molecular CO2 uptake. Mar. Ecol. Prog. Ser. 133, 275- 285.

Raven J. A. and Beardall J. (2016) The ins and outs of CO2. J. Exp. Bot. 67, 1-13. Raven J. A. and Johnston A. M. (1991) Mechanisms of inorganic-carbon acquisition in marine phytoplankton and their implications for the use of other resources. Limnol. Oceanogr. 36, 1701- 1714. Reinfelder J. R. (2011) Carbon concentrating mechanisms in eukaryotic marine phytoplankton. Annu. Rev. Mar. Sci. 3, 291-315. Rickaby R. E. M. (2015) Goldilocks and the three inorganic equilibria: how Earth’s chemistry and life coevolve to be nearly in tune. Phil. Trans. R. Soc. A 373, 20140188. Riebesell U., Burkhardt S., Dauelsberg A., and Kroon B. (2000a) Carbon isotope fractionation by a marine diatom: dependence on the growth-rate-limiting resource. Mar. Ecol. Prog. Ser. 193, 295-303.

Riebesell U., Revill, A. T., Holdsworth, D. G., and Volkman, J. K. (2000b) The effects of varying CO2 concentration on lipid composition and carbon isotope fractionation in Emiliania huxleyi. Geochim. Cosmochim. Ac. 64, 4179-4192. Robinson J. J., Scott K. M., Swanson S. T., O’Leary M. H., Horken K., Tabita F. R., and Cavanaugh C. M. (2003) Kinetic isotope effect and characterization of form II RubisCO from the chemoautotrophic

endosymbionts of the hydrothermal vent tubeworm Riftia pachyptila. Limnol. Oceanogr. 48, 48- 54. Roeske C. and O’Leary M. H. (1984) Carbon isotope effects on enzyme-catalyzed carboxylation of ribulose bisphosphate. Biochemistry 23, 6275–6284. Roeske C. and O’Leary M. H. (1985) Carbon isotope effect on carboxylation of ribulose bisphosphate catalyzed by ribulosebisphosphate carboxylase from Rhodospirillum rubrum. Biochemistry- US 24,1603–1607. Rohmer M. (2003) Mevalonate-independent methylerythritol phosphate pathway for isoprenoid biosynthesis. Elucidation and distribution. Pure Appl. Chem. 75, 375-387. Rost B., Zondervan I., and Riebesell U. (2002) Light-dependent carbon isotope fractionation in the coccolithophorid Emiliania huxleyi. Limnol. Oceanogr. 47, 120-128. Rost B., Richter K. -U., Riebesell U., and Hansen P. J. (2006) Inorganic carbon acquisition in red tide dinoflagellates. Plant Cell Environ. 29, 810-822. Rowan R., Whitney S. M., Fowler A., and Yellowlees D. (1996) Rubisco in marine symbiotic dinoflagellates: form II enzymes in eukaryotic oxygenic phototrophs encoded by a nuclear multigene family. Plant Cell 8, 539-553. Schnepf E. and Elbrachter M (1999) Dinophyte chloroplasts and phylogeny – A review, Grana 38, 81-97. Schouten S., Klein Breteler W. C. M., Blokker P., Schogt N., Rupstra I. C., Grice K., Baas M., and Damsté J. S. S. (1998) Biosynthetic effects on the stable carbon isotopic compositions of algal lipids: Implications for deciphering the carbon isotopic biomarker record. Geochim. Cosmochim. Ac. 62, 1397-1406. Sharkey T. D. and Berry J. A. (1985) Carbon isotope fractionation of algae as influenced by an inducible CO2 concentrating mechanism. In Inorganic Carbon Uptake by Aquatic Photosynthetic Organisms (eds. W. J. Lucas and J. A. Berry). pp. 389-401. Sluijs, A. Pross J., and Brinkhuis H. (2005) From Greenhouse to icehouse: organic-walled dinoflagellates as paleoenironmental indicators in the Paleogene. Earth Sci. Rev. 68, 281-315. Strickland J. D. H. and Parsons T. R. (1968) Determination of reactive phosphorus. In A practical handbook of seawater analysis, Fisheries Research Board of Canada, Bulletin 167, pp. 49-56. Sullivan J. M. and Swift E. (2003) Effects of small-scale turbulence on net growth rate and size of ten species of marine dinoflagellates. J. Phycol. 39, 83-94. Summons R. E. and Walter M. R. (1990) Molecular fossils and microfossils of prokaryoties and protist from Proterozoic sediments. Am. J. Sci. 290, 212-244. Tabita F. R., Hanson T. E. Li H., Satagopan S., Singh J., and Chan S. (2007) Function, structure, and evolution of the RubisCO-like proteins and their RubisCO homologs. Microbiol. Mol. Biol. Rev. 71, 576-599. Tabita F. R, Satagopan S., Hanson T. E., Kreel N. E., and Scott S. S. (2008) Distinct form I, II, III, and IV Rubisco proteins from the three kingdoms of life provide clues about Rubisco evolution and structure/function relationships. J. Exp. Bot. 59, 1515-1524.

Tcherkez G. (2013) Modelling the reaction mechanism of ribulose-1,5-bisphosphate carboxylase/oxygenase and consequences for kinetic parameters. Plant Cell Environ. 36, 1586- 1596. Tcherkez G. G. B., Farquhar G. D., Andrews T. J. (2006) Despite slow catalysis and confused substrate specificity, all ribulose bisphosphate carboxylases may be nearly perfectly optimized. P. Natl. Acad. Sci. USA 103, 7246-7251. Van de Waal D. B., Eberlein T., Bublitz Y., John U., and Rost B. (2014) Shake it easy: a gently mixed continuous culture system for dinoflagellates. J. Plankton Res. 36, 889-894. Van Heuven S., Pierrot D., Rae J. W. B., Lewis E., and Wallace D. W. R. (2011) MATLAB program developed for CO2 system calculations. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory. Watson G. M. and Tabita F. R. (1997) Microbial ribulose 1,5-bisphosphate carboxylase/oxygenase: a molecule for phylogenetic and enzymological investigation. FEMS Microbiol. Lett. 46, 13-22.

Whitney S. M. and Andrews T. J. (1998) The CO2/O2 specificity of single-subunit ribulose-bisphosphate carboxylase from the dinoflagellate, Amphidinium carterae, Aust. J. Plant Physiol. 25, 131-138. Whitney S. M., Shaw D. C., and Yellowlees D. (1995) Evidence that some dinoflagellates contain a ribulose-1,5-bisphosphate carboxylase/oxygenase related to that of the alpha-proteobacteria. Proc. Biol. Sci. 1356, 271-275. Young J. N., Rickaby R. E. M., Kapralov M. V., and Filatov D. A. (2012) Adaptive signals in algal Rubisco reveal a history of ancient atmospheric carbon dioxide. Phil. Trans. R. Soc. B: Biological Sciences 367, 483-492. Young J. N., Heureux A. M. C., Sharwood R. E., Rickaby R. E. M., Morel F. M. M., and Whitney S. M. (2016) Large variation in the Rubisco kinetics of diatoms reveals diversity among their carbon- concentrating mechanisms. J. Exp. Bot. 67, 3445-3456. Zhang J.-Z. and Fischer C. J. (2006) A simplified resorcinol method for direct spectrophotometric determination of nitrate in seawater. Mar. Chem. 99, 220-226.

Chapter 4

A general model for carbon isotope fractionation in eukaryotic phytoplankton

[To be submitted to Geochimica et Cosmochimica Acta]

Abstract

The carbon isotopic composition of organic matter preserved in marine sediments provides a window into the global carbon cycle through geologic time, including variations in atmospheric CO2 levels. Traditional models for interpreting isotope records of marine phytoplankton assume that these archives primarily reflect kinetic isotope discrimination by the carbon-fixing enzyme RubisCO. However, some in vivo and in vitro measurements appear to contradict this assumption, indicating that significant questions remain regarding the mechanistic underpinning of algal isotopic signatures, including the role of carbon concentrating mechanisms (CCMs). Here, we present a general model for explaining photosynthetic carbon isotope fractionation (εP) in prominent eukaryotic phytoplankton groups. The model proposes that a nutrient- and light-dependent step upstream of RubisCO is a kinetic barrier to carbon acquisition and therefore represents a significant source of isotopic fractionation. The model is able to reproduce existing chemostat and batch culture datasets with a normalized root mean squared error (nRMSE) of 7.6%. The primary implications are that RubisCO is predicted to exert minimal isotopic control in nutrient-limited regimes but becomes influential as growth becomes light-limited. This framework enables both environment-specific and taxon-specific isotopic predictions. By refining the mechanistic understanding of marine photosynthetic carbon isotope fractionation, we may begin to reconcile existing datasets and reexamine Phanerozoic isotope records—including the resulting CO2 reconstructions—by emphasizing the influence of different types of resource limitation on photosynthetic carbon acquisition.

Introduction Accurate interpretations of marine organic carbon isotope records rely on a mechanistic understanding of photosynthetic carbon isotope fractionation (εp). Isotopic models for phytoplankton share the common goal to understand εp in the context of ambient carbon dioxide concentrations [CO2(aq)] and algal physiology (e.g.,Sharkey and Berry, 1985; Laws et al., 1995; Cassar et al., 2006; Schulz et al., 2007;

McClelland et al., 2017), thereby enabling efforts to reconstruct pCO2 (paleobarometry; Jasper and Hayes,

1990; Laws et al., 2002; Pagani et al., 2011). The most widely adopted framework consists of a two-step, passive-diffusive supply model (Figure 4.1a; Rau et al., 1992; Francois et al., 1993; Goericke et al., 1994;

Laws et al., 1995; Rau et al., 1996) that was adapted from studies on land plants (Farquhar et al., 1982;

1989). This model predicts that P depends on the balance between two processes with distinct isotope effects: diffusion of CO2 (< 1‰ in water; O’Leary, 1984) and CO2 fixation by the enzyme RubisCO

(~25−30‰; Table 4.1). When algal growth rates () are low or ambient CO2 concentrations are high, the rate-limiting step is presumed to be carbon fixation by RubisCO; and the isotope effect associated with this process sets the theoretical maximum value of εP, which is denoted εf (for “fixation”). When instead the supply of CO2 is rate-limiting, the fractionation accompanying passive diffusion of CO2 (D) is expressed, defining the minimum value of εP. These endmembers correspondingly define a line that denotes all intermediate conditions (Figure 4.1a).

Until recently, it has been assumed that the value of εf equals the fractionation measured in vitro for RubisCO from higher plants (Table 4.1; Roeske and O’Leary 1984; Guy et al., 1993; Scott et al., 2004;

McNevin et al., 2006), adjusted slightly for the effects of anaplerotic reactions (-carboxylations; Francois et al., 1993). This perspective was reinforced by nitrate-limited chemostat experiments with three species of eukaryotic phytoplankton that yielded εf values of ~25‰ at the limit of infinite CO2 supply or slow growth (Figure 4.2; Popp et al., 1998). However, mounting evidence suggests that this RubisCO-centric framework must be revisited. RubisCO exists in several catalytically and phylogenetically distinct forms in phytoplankton, Forms IA, IB, ID and II (Tabita et al., 2008; Whitney et al., 2011); and in particular, the

value of RubisCO for the Form ID version in the haptophyte alga Emiliania huxleyi is reported to be only

11‰ (Boller et al., 2011). It now seems incorrect to interpret the in vivo f values for algal species having

Form ID RubisCO by analogy to in vitro measurements of Form IB RubisCO from spinach and other higher plants (Rickaby et al, 2015; Wilkes et al., 2017; McClelland et al., 2017).

Figure 4.1. Passive diffusion model vs. the revised model. (a) Schematic view of the traditional passive- diffusive supply model of carbon isotope fractionation in phytoplankton (left) and its associated isotopic consequences (center). Photosynthetic fractionation (P) depends on the balance of two processes with different characteristic isotope effects: diffusion of CO2 in water (D) and carbon fixation (f ≈ RubisCO = 25-27‰). The rate-limiting steps and their isotopic consequences are labeled A and B, respectively. (b) Schematic view of processes influencing P in the proposed model: light-limited growth (left) vs. nutrient- limited growth (right), and associated isotopic consequences (center), assuming HYD = 25‰ and RubisCO = 11‰. (Note that RubisCO varies by taxon.) The rate-limiting steps and their isotopic consequences are labeled B and C, where C represents HYD, and D indicates the model behavior at increasing /[CO2(aq)].

Indeed, other measurements in addition to Boller et al. (2011) support the suggestion that εf values inferred from chemostat experiments do not correspond to εRubisCO values (Table 4.1; Figure 4.2). Form ID

RubisCO purified from the diatom Skeletonema costatum yields a value of εRubisCO of 18.5‰ (Boller et al.,

2015). Similarly, recent chemostat incubations with a dinoflagellate employing Form II RubisCO indicate an εf value of 27‰ in vivo (Alexandrium tamarense; Wilkes et al., 2017). Although consistent within error

estimates with the ~25‰ f values for other eukaryotic algae, this result was surprising given the striking differences in catalytic properties, structures, and amino acid sequences between Form I and II RubisCOs

(Rowan et al., 1996) and also the apparent similarities between dinoflagellate and proteobacterial Form II

RubisCOs (RubisCO ≈18−23‰, Form II; Robinson et al., 2003; McNevin et al., 2007). Collectively, the observations indicate that εf values for eukaryotes from nitrate-limited chemostat experiments are in good agreement with one another, yet they consistently do not equal the kinetic isotope effects measured for purified RubisCOs from the most taxonomically similar algal or bacterial source (Figure 4.2, Table 4.1).

Several additional lines of evidence support a greater diversity in εRubisCO values than previously assumed. McNevin et al. (2007) demonstrated that a single point mutation in the large subunit of Form IB

RubisCO from tobacco had a dramatic effect, lowering the in vitro fractionation from 27.4‰ in the wild- type to 11.2‰ in the mutant. Characterization of Form ID RubisCO kinetic properties from 11 diatoms also uncovered unexpected diversity (Young et al., 2016). Given that RubisCO’s intrinsic isotope discrimination has been linked empirically and mechanistically to the enzyme’s kinetic properties

(Tcherkez et al., 2006), it might be predicted that εRubisCO would also vary among phytoplankton. Such diversity makes the apparently uniform ~25-27‰ εf value for marine phytoplankton surprising, if it is related to εRubisCO; alternatively, it argues that the value of εf primarily reflects some other process.

Here we propose a general theoretical model for εP in eukaryotic phytoplankton to reconcile the apparent contradictions between εf values and εRubisCO measurements, with the aim of unifying existing data and models. Our model is tested against a wide range of experimental datasets – including both chemostat and batch-culture approaches – and is constructed such that the rate-limiting step for photosynthetic carbon fixation varies depending on the balance of nutrient and light availability. This work builds upon a long history of modeling and culturing efforts that demonstrate the importance of nutrient availability, energy sources, and carbon concentrating mechanisms (CCMs) to the expression of εP (e.g., Beardall et al., 1982;

Raven, 1997; Tchernov et al., 1997; Burkhardt et al, 1999a,b; Riebesell et al., 2000a,b; Rost et al., 2002,

2006; Cassar et al., 2006; Schulz et al., 2007; Hopkinson, 2014; Hoins et al., 2016; Holtz et al., 2017).

The primary innovation of our model is that under nutrient-limited, light-replete conditions (e.g., chemostat culture experiments; Popp et al., 1998; Wilkes et al., 2017, 2018), we hypothesize that the rate- limiting step occurs upstream of RubisCO and accompanies a catalyzed, irreversible conversion of CO2 to

- HCO3 (Figure 4.1b) either enzymatically, or through a process of non-enzymatic catalysis (Keller et al.,

2015). Hereafter we will refer to this process as enzymatic but acknowledge that the true mechanism is ambiguous. Because this reaction is taken to be the rate-limiting process, and because it is proposed to be common to all eukaryotic phytoplankton, it can provide a taxonomically-constant ~25‰ discrimination against 13C. Analogous abiotic hydration (or hydroxylation) reactions have been theorized to fractionate carbon isotopes by as much as ~20−39‰ based on quantum chemistry calculations and experiments (Table

C.2; Siegenthaler and Munnich, 1981; Usdowski et al., 1982; Clark and Lauriol, 1992; Zeebe and Wolf-

Gladrow, 2001; Zeebe, 2014). The commonly-cited 13‰ kinetic isotope effect for CO2 hydration (Marlier and O’Leary, 1984; O’Leary et al., 1992) likely is too small, as it is based on experiments in which back- reaction (equilibration) may not have been eliminated (Sade and Halevy, 2017). The presence of a common upstream process would mask the variable isotope effects associated with RubisCO, unless the cells experience alternate conditions in which RubisCO activity becomes the slow step of carbon fixation. This is proposed to occur under nutrient-replete conditions in which the

- photon flux rate becomes growth-limiting and, as we detail below, the catalytic conversion of CO2 to HCO3 slows down. We define this as “light-limited”, by which we mean to imply a status of relative limitation rather than an absolute threshold for a specific photon flux. The implication is that there are two distinct rate-limiting steps for carbon fixation in eukaryotic phytoplankton, with different maximum values for εP at the limit of slow  and high [CO2(aq)] that depend on the resource environment (Figure 4.1b). Here we contextualize how this general framework relates to existing algal physiology models, identify plausible cellular mechanisms and evolutionary drivers, and consider the implications for interpreting marine εP records in the context of paleobarometry.

Table 4.1. Compiled εRubisCO and εf values for different RubisCO forms. Marine phytoplankton are indicated with an asterisk. εRubisCO εf RubisCO Biological Organism (‰) Reference (‰) Reference Form Source Type (in vitro) (in vivo)

Solemya velum IA γ-Proteobacterium 24.5 Scott et al. (2004) − symbiont Prochlorococcus Cyanobacterium 24 Scott et al. (2007) − marinus MIT9313* Popp et al. Synechococcus Sp.* Cyanobacterium − − 17 (1998) Roeske & O’Leary (1984); Guy et al. IB Spinacia oleracea Higher Plant 26−30 (1993); Scott et al. − (2004); McNevin et al. (2006) Gossypium Higher Plant 27.1 Wong et al. (1979) − Nicotiana tabacum Higher Plant 27.4 McNevin et al. (2007) Synechococcus PCC Guy et al. (1993); Cyanobacterium 21−22 - 6301a McNevin et al. (2007) Bidigare et ID Emiliania huxleyi* Coccolithophore 11.1 Boller et al. (2011) 25 al. (1997) Skeletonema Diatom 18.5 Boller et al. (2015) − costatum* Phaeodactylum Laws et al. Diatom − − 25 tricornutum* (1997) Popp et al. Porosira glacialis* Diatom − − 25 (1998) Riftia pachyptila II γ-Proteobacterium 19.5 Robinson et al. (2003) − symbiont Roeske & O’Leary Rhodospirillum (1985); Guy et al. α-Proteobacterium 18−23 − rubrum (1993); McNevin et al. (2007) Alexandrium Peridinin-containing Wilkes et al. − − 27 tamarense* Dinoflagellate (2017) a Freshwater cyanobacterium. PCC 6301 is a strain synonym for Anacystis nidulans and Synechococcus PCC 6301.

Figure. 4.2. Comparison of εRubisCO values measured in vitro with εf values determined in vivo. Plotting εp as a function of μ/[CO2(aq)] for eukaryotic phytoplankton grown in nitrate-limited chemostats yields intercepts (εf values) of 25-27‰. The lines represent geometric mean regression analysis. Black squares indicate the diatom Phaeodactylum tricornutum (Form ID RubisCO; Laws et al., 1997); dark grey triangles, calcifying and non-calcifying clones of the coccolithophore Emiliania huxleyi (Form ID RubisCO; Bidigare et al., 1997); light grey circles, dinoflagellate Alexandrium tamarense (Form II RubisCO; Wilkes et al., 2017); white diamonds, diatom Porosira glacialis (Form ID RubisCO; Popp et al., 1998). The intercepts are compared to εRubisCO ranges listed on the left side of the figure, measured in vitro (purified enzyme) for RubisCO forms IA, IB, ID, and II (data, Table 4.1). Box height and location along the εp-axis depict the ranges of εRubisCO, and the shading corresponds to the most similar chemostat-grown species.

Generic Model for Eukaryotic Phytoplankton Model structure

The model employs a simplified cellular architecture representing a generic, eukaryotic algal cell

(Figure 4.3). The cell consists of a cytosol, a membrane-bound chloroplast, and a pyrenoid where RubisCO is concentrated. The pyrenoid provides diffusive resistance, enabling buildup of CO2 around RubisCO and discouraging entry and buildup of O2. The sites of the light reactions of photosynthesis—the thylakoids— are modeled in close proximity to, or penetrating, the pyrenoid. Evidence for these physical features is broadly distributed across eukaryotic phytoplankton groups (Badger et al., 1998; Tachibana et al., 2011),

with exception of a small number of cases in which the pyrenoid appears to be absent (Ratti et al., 2007;

Darienko et al., 2015; Heureux et al., 2017). The internal volume of the thylakoid is acidic, while the chloroplast volume surrounding the thylakoid is relatively alkaline (Raven, 1997; Höhner et al., 2016).

Figure 4.3. Model structure. Dashed lines indicate passive (diffusive) fluxes while solid lines indicate active transport and/or enzymatic conversions. The invoked enzymes are RubisCO, carbonic anhydrase (CA), and a putative enzyme or other non-enzymatic process catalyzing the active hydr(oxyl)ation of CO2 - to HCO3 (HYD). The interaction of photons with the thylakoid membrane is shown with a jagged arrow near the process of hydration/hydroxylation. Fluxes down the HYD path (HYD) and the diffusive path

(CO2) are shown; the ratio between them defines the new parameter,  (Eq. 1).

The model includes passive diffusion of CO2 plus two mechanisms of active carbon acquisition.

The two active modes are:

- (1) Transport of extracellular HCO3 through the plasmalemma and chloroplast membranes using

membrane-bound transporters, regulated by a CO2-responsive transcription factor. Within the

- pyrenoid in the chloroplast, carbonic anhydrase (CA) converts this accumulated HCO3 to CO2

near RubisCO.

(2) Enhanced diffusion of CO2 via a putative hydr(oxyl)ating process that promotes the kinetic

- conversion of CO2 to HCO3 . We assume that this process is directly coupled to transport of

- HCO3 across the thylakoid membrane to suppress the reverse reaction and permit accumulation

of charged bicarbonate within the thylakoid. This approach would help maintain a CO2

concentration gradient between the extracellular environment and the chloroplast by drawing

down intracellular CO2, thereby promoting continued diffusion into the cell. The ultimate step

- is the quantitative dehydration of accumulated HCO3 to CO2 near RubisCO by a CA contained

within the thylakoid in regions penetrating the pyrenoid.

Relationship to existing CCM definitions

RubisCO is characterized by a slow maximum catalytic turnover rate and a low affinity for CO2

(Badger et al., 1998). These inefficiencies are exacerbated in marine environments by the slow diffusion

- of CO2 in water and by inorganic carbon speciation favoring HCO3 . Phytoplankton actively regulate the

CO2 concentration around RubisCO with biophysical and biochemical CCMs to ensure efficient fixation

(for recent reviews, see Reinfelder, 2011; Griffiths et al., 2017). One implication of this physiology is that the concentration of CO2 around RubisCO rarely would be predicted to reflect the concentration of CO2 outside the cell. Another implication is that in the absence of intracellular substructures, the CO2 diffusive gradient between internal and external environments generally would be inverted (intracellular [CO2(aq)] > extracellular [CO2(aq)]), an impediment to carbon flux unless specific strategies are employed to enhance inward-directed diffusion and/or transport (Hopkinson, 2014; Raven and Beardall, 2015). The model topology used here was designed to reflect these considerations. Our two active carbon acquisition mechanisms synthesize a variety of experimentally-verified CCM components (Table 4.2) into the two categories of active processes detailed above.

Table 4.2. CCM components combined and generalized as active transport mechanisms in the model. CCM Component Localization Taxa Description References Membrane-bound Plasmalemma P. tricornutum SLC4 homolog Romero et al., 2004 - HCO3 transporter E. huxleyi Makinder et al., 2011 Richier et al., 2011 Nakajima et al., 2013 Carbonic anhydrase Pyrenoid P. tricornutum -carbonic anhydrases Satoh et al., 2001 (PtCA1, PtCA2) Tanaka et al., 2005 Tachibana et al., 2011 Kikutani et al., 2012 Carbonic anhydrase Internal volume of the P. tricornutum θ-carbonic anhydrase Spalding, 2008 thylakoid T. weissflogii α-carbonic anhydrase Sinetova et al., 2012 C. reinhardtii Kikutani et al., 2016 Pyrenoid Chloroplast P. tricornutum Badger et al., 1998 E. huxleyi Heureux et al., 2017 A. tamarense Hansen & Moestrup, 1998 CO2-responsive T. pseudonana cyclic AMP (cAMP) Hennon et al., 2015 regulation of CCM P. tricornutum transcription factor Matsuda et al., 2011 components Ohno et al., 2012

- In the commonly-invoked CCMs, CO2 and HCO3 are assumed to only substantially interconvert in the presence of carbonic anhydrase (Hopkinson et al., 2016; Mangan et al., 2016). Our model proposes a category of CCM distinct from those that use CA: namely, a transmembrane, photon-energized or

- photosynthetically-enhanced hydr(oxyl)ation of CO2 to HCO3 (similar to Kaplan and Reinhold, 1999;

- Tchernov et al., 2001; Eichner et al., 2015). We assume that the charged species HCO3 does not diffuse through membranes to any significant extent. This also implies that diffusion is not a significant mode of

- entry for HCO3 into the cell, relative to active transport across the plasmalemma. However, equilibration

- of HCO3 with the uncharged species H2CO3 may supply a minor, passive flux of inorganic carbon through

- membranes (Mangan et al., 2016). Therefore, HCO3 is permitted to passively transit the membrane-bound cytoplasm and chloroplast as H2CO3 (depicted with a dotted line in Figure 4.3).

Isotopic mass balance The model follows the reaction network approach of Hayes (2001) and consists of 8 carbon pools with defined isotopic compositions and masses, 12 distinct fluxes between these pools, and one output flux to prevent buildup of biomass (Figure C.1, Appendix C).

The model is described by a system of equations representing the mass balances for total carbon and 13C of each pool and is solved for steady-state 13C values and relative carbon fluxes using ode23 solver

- in MATLAB r2015a. The steady-state isotopic compositions of CO2 and HCO3 are calculated for each

2- carbon pool within the cell, while H2CO3 and CO3 are implicitly treated as equivalent to (contained within

- the pools of) HCO3 since all three species equilibrate rapidly (Zeebe and Wolf-Gladrow, 2001; Mangan et al., 2016; Sade and Halevy, 2017). We assume that intracellular exchange of CO2 between the chloroplast

- and cytosol is rapid and that the rate of spontaneous (i.e., uncatalyzed) equilibration with HCO3 within

- these compartments is insignificant. Isotopic compositions of the extracellular reservoirs of CO2 and HCO3 are prescribed to match values from experiments or the natural environment.

The kinetic isotope effect (ε) governing carbon transfer between any two pools is approximated as the difference between two δ values (Hayes, 2001). The isotope effect associated with RubisCO (RubisCO) is assigned between 11.1 and 24‰ depending on the taxonomic identity of the modeled cell (Table 4.1;

- Table C.1). The isotope effect adopted for the unidirectional conversion of CO2 to HCO3 (HYD) is 25‰.

This value is larger than commonly has been assumed for hydr(oxyl)ation (~11−13‰1; O’Leary et al.,

1992; Zeebe and Wolf-Gladrow, 2001) but is supported by several lines of experimental and theoretical evidence (Table C.2; see discussion in Appendix C). We do not distinguish between the hydration and hydroxylation pathways.

Flux balance response to variable nutrient and light conditions

Under conditions of excess light energy, we assume that extra membrane potential is directed to

- hydr(oxyl)ating CO2 to HCO3 (HYD; Figure 4.3). Photon fluxes exceeding the requirement to synthesize biomass may intensify the pH gradient across the thylakoid membrane, with the light-induced transfer of

+ H from water accompanied by diversion of a hydroxide for the active hydr(oxyl)ation of CO2. This process

1An  value of 11-13‰, when combined with 20-22‰ in the reverse direction, yields the 9‰ equilibrium isotope effect ( = 1.009; 25°C). Here, 25‰ in the forward direction implies 34‰ for the reverse direction, also within theoretical and experimental boundaries.

might serve or be coupled to a photoprotective function or intracellular pH regulation, with carbon concentration representing an ancillary benefit (see Discussion). Thus, this mechanism is assumed to occur constitutively (i.e., even when CO2 is abundant) under nutrient-limited, high-photon conditions, as its primary role would be to dispose photosynthetically-driven pH gradients. As the balance of energy and nutrients shifts such that light begins to limit growth, we assume that this facilitated CO2 uptake is less induced. Instead, passive diffusion of CO2 into the pyrenoid (CO2; Figure 4.3) will be the dominant mechanism by which CO2 accumulates around RubisCO. Thus, diffusion is predicted to be the sole source in the limit of high [CO2(aq)] and minimum photon flux.

To reflect these processes, we introduce a unitless parameter omega (; Equation 1) to index the balance between energized (HYD) or passive (CO2) entry of CO2 into the pyrenoid. The  parameter allows us to smoothly adjust the model fluxes between the endmember physiological states of relative nutrient limitation vs. relative light limitation. Values of  approaching 1 imply that CO2 uptake occurs entirely via the energized mechanism, while values approaching 0 imply passive diffusion.

 = HYD/(HYD+ CO2); (0 <  < 1) (1)

- Regardless of , active HCO3 uptake (parameter “H” in Table C.3; Figure 4.3) is parameterized to become more prevalent in all cells as extracellular [CO2(aq)] decreases (Rost et al., 2003, 2006; Bach et al.,

- 2013). This process is assumed to be actively regulated, and as such the percent contribution of HCO3 is dependent on , , and [CO2(aq)] rather than being a fixed value. However, in all cases, when  approaches

- 1, proportionally less of the cellular carbon demand will need to be met through HCO3 import, since the

- hydr(oxyl)ation pathway is a form of CCM. However, when  approaches 0, active uptake of HCO3 will need to supply a greater percentage of inorganic carbon because it will be the only available CCM in the cell.

The remaining isotope effects and flux balance parameterizations – including free and dependent parameters – for all the intracellular carbon pools are defined in Tables C.1 and C.3.

Model testing and validation

Culture experiments with diatoms (2 species), coccolithophores (calcifying and non-calcifying strains), and dinoflagellates (2 species within the same genus) were compiled from the literature (Figures

C.3, C.4; Table C.4). These include nutrient-limited chemostat experiments consolidated from seven sources (Bidigare et al., 1997; Laws et al., 1997; Popp et al., 1998; Cassar, 2003; Hoins et al., 2016; Wilkes et al., 2017, 2018), as well as batch cultures of the same species, compiled from an additional seven sources

(Burkhardt et al., 1999a,b; Riebesell et al., 2000a,b; Rost et al., 2002; Hoins et al., 2015, 2016).

Empirical model input parameters were defined according to measurements reported in these studies (Table C.4) and include the instantaneous growth rate (i), corresponding to carbon fixation during the photoperiod (Riebesell et al., 2000a, b), and the particulate organic carbon per cell (POC). Together, these two inputs define the carbon-specific growth rate, C = iPOC, which accounts for differences in diel cycle and cellular carbon content between studies (Rost et al., 2002). Experiment-derived inputs also include cell surface area (SA), the concentration of CO2 in the medium ([CO2(aq)]), and reported εp values.

In some cases, indicated by italics in Table C.4, surface area or POC was calculated from the empirical relationships of Montagnes et al. (1994).

Results

Behavior of the generalized model for P

The model reconciles key features of existing P observations for both nutrient-limited chemostat and light-limited batch cultures grown over a variety of  and CO2 conditions. First we demonstrate the general behavior of the model (Figure 4.4) by examining the three isotopic endmembers (labeled B, C, D in Figure 4.1b) and their underlying physiology.

Under nutrient limitation, the model generates intercepts of 25‰ for all taxa (analogous to intercept

C, Figure 4.1b) because this number reflects the full expression of the isotope effect associated with the active hydr(oxyl)ation mechanism. This intercept manifests when virtually all the CO2 entering the

12 chloroplast leaks back out again, permitting continuous replenishment of CO2 around the putative hydr(oxyl)ating enzyme. Physiologically, this endmember represents slow  and/or high [CO2(aq)], under

- conditions of high photosynthetic activity in tandem with no investment in HCO3 import.

By contrast, the model predicts a diversity in P-intercepts under conditions of light limitation, such as in batch culture experiments. Under slow  corresponding to the lowest photon flux densities (PFD, E m-2 s-1), intercept B (Figure 4.1b) reflects the full expression of the isotope effect associated with each taxon-specific RubisCO (Table 4.1), and different taxa have different intercepts (Figure 4.4). In this scenario, CO2 enters the cell solely by diffusion, and subsequently most leaks out, again due to slow growth

- () and/or high [CO2(aq)]. Like for intercept C, intercept B also reflects no HCO3 import. Under intermediate conditions between C and B, the intercept is dictated by the value of . At intermediate values of , the new maximal in vivo isotopic fractionation (f_) follows equation (2).

ε =ε +ω(ε − ε ) (2) f_ RubisCO HYD RubisCO As the ratio /[CO2(aq)] increases (condition D, Figure 4.1b), the isotopic behavior is dominated by the ratio of passive CO2 leakiness relative to gross inorganic carbon influx through the cell membrane

(parameter “L”; Table C.3). The value of L can vary between 0 (impermeable membrane or vanishingly small surface area; P(SA) = 0; or high carbon-specific growth rate, C → ∞) and 1 (highly permeable membrane P(SA)→ ∞ or no growth). Following Francois et al. (1993) and Cassar et al. (2006), L is derived by mass balance (Equation 3):

(P)(SA)([CO ]) L= 2(aq) (3). (P)(SA)([CO2(aq)])+μc

Here, P is defined as the area-specific mass transport of CO2 out of the chloroplast and cell membranes (kg

m-2 cell-1 d-1). Physiologically, P is expected to reflect concentration gradients and combined (chloroplast and plasmalemma) membrane permeabilities. Given substantial uncertainties in membrane permeabilities and the model’s simplified cellular structure, P is treated as a tunable (fitted) parameter. Adjusting P

influences the trajectory of P values when plotted as a function of i/CO2 (compare dashed vs. solid lines in Figure 4.4); e.g., increasing P leads to a shallower decreasing slope for P.

Figure 4.4. Behavior of the generalized model. Model sensitivity to parameters , P, and empirical inputs (RubisCO , SA, and POC). (a) The dinoflagellate A. tamarense is simulated with RubisCO = 19.5‰, SA = 5100 m2, POC = 2.810-4 mol C kg-1, and P = 410-9 (high) or 110-9 (low). (b) The coccolithophore E. huxleyi 2 -7 -1 -9 is modeled with RubisCO = 11.1‰, SA = 88 m , POC = 6.710 mol C kg , and P = 410 (high) or -9 -1 110 (low). i was uniformly assumed to equal 1, and [CO2(aq)] was varied from 2 to 300 mol kg .

- Additionally, as the ratio i/[CO2(aq)] increases (condition D, Figure 4.1b), HCO3 import increases.

13 This increases the cellular C content (decreasing P). However, this process has a secondary impact when compared to the parameters “L” and “” and is not primarily responsible for the decline in P as a function of i/[CO2(aq)] (Figure 4.4). Because we permit bicarbonate to efflux from the cell as H2CO3, the intracellular dehydration is accompanied by isotope fractionation. Relative to some existing models, this

- reduces the importance of HCO3 uptake to the overall expression of P; other models often assume complete

- - dehydration of imported HCO3 and thus a greater isotopic consequence of the ~9‰ enrichment of HCO3 relative to CO2 (e.g., Burkhardt et al., 1999a,b; Keller and Morel, 1999). We uniformly assume 10% of

- HCO3 entering the cell leaks out again (parameter “E” for “efflux” = 0.1, Table C.3). While this assumption

- is clearly a simplification of the pH- and concentration-dependent efflux of HCO3 as H2CO3 (c.f., Mangan et al., 2016), the model behavior is not significantly sensitive to E when 0 < E < 0.4.

Simulation of chemostat and batch culture data

Figures 4.5−4.7 depict model fits to literature data for four eukaryotic phytoplankton from ecologically prominent clades. Experiments were included only where growth was limited specifically by nutrients (Bidigare et al., 1997; Laws et al., 1997; Popp et al., 1998; Cassar, 2003; Hoins et al., 2016; Wilkes et al., 2017, 2018) or light (Burkhardt et al., 1999a,b; Riebesell et al., 2000a,b; Rost et al., 2002; Hoins et al., 2015, 2016), rather than another resource. Species were selected if both nutrient-limited chemostat and light-limited batch culture data were available for comparison, with the exception of P. glacialis, for which there are no batch culture experiments. The batch culture data were sorted further into groups based on light intensity and diel cycle (Tables C.4 and C.5).

When i, [CO2(aq)], POC, and SA are known from experiments, the only free parameters in the model are  and P. To simulate the data, these free parameters were optimized for each experiment

(nutrient, light intensity, and diel cycle combination) using Monte Carlo simulations. The combination of parameters that minimized the sum of the squared residuals for each group of experiments was selected.

Note that  was optimized over constrained ranges to reflect the relative light availability:  = 0.99 represents the nutrient-limited chemostat end-member, while the batch cultures experiencing the lowest light (intensity and diel cycle) were assumed to approach  values near 0. P was permitted to vary between

510-10 and 510-8; most experiments yielded values of P ~10-9 (Figures 4.5-4.7 and Table C.5).

The greatest quantity of data is available for the diatom P. tricornutum (n = 61) and the haptophyte

E. huxleyi (n = 56). These are also the datasets that yielded the best agreement between modeled and measured P values (data shown with colored symbols, all model results shown with black triangles; Figures

4.4 and 4.5). The model reproduces the curvature of the P values for the P. tricornutum chemostat

experiments with respect to i/[CO2(aq)] (Figure 4.5a) and the relatively linear P responses of the light-

- limited batch cultures (Figure 4.5c). It predicts a maximum gross HCO3 import of ~3% (normalized to c) for the chemostat experiments, and a range of 2.3−76% for the batch cultures (Table C.5).

For E. huxleyi, the model predicts P values in good agreement with all datasets (Figure 4.6), except it fails to reproduce the shallow slope of the lowest light condition (PFD = 30; 16:8; n = 4). The model

- predicts a maximum gross HCO3 import of 0.1% for the E. huxleyi chemostat experiments, and a range of

1.5−75% for the batch cultures (Table C.5). The taxa with smaller data sets (Alexandrium spp., n = 13 and

P. glacialis, n = 7) show poorer agreement between modeled and measured P values (Figures 4.7 and 4.8).

The model effectively reproduces 7 of the 9 data points from nitrate-limited chemostat cultures of

Alexandrium; however, omitting the two outliers (circled in Figure 4.7) improves the agreement between

- modeled and measured values significantly (Figure 4.7b). The model predicts maximum HCO3 import of

0.1% for the Alexandrium chemostat experiments and 11% for the batch cultures. For P. glacialis, gross bicarbonate import is predicted to be essentially negligible, never exceeding 0.1% of c.

Figure 4.5. Modeled vs. measured P values for the diatom P. tricornutum. (a) Nitrate and phosphate- limited chemostat cultures (gold circles; Cassar, 2003; Laws et al., 1997), modeled with P = 310-9. (b)

Modeled vs. measured P values from (a). (c) Batch cultures grown in continuous light conditions (filled squares) and diel cycles (open squares) (Riebesell et al., 2000a; Burkhardt et al., 2000a,b), modeled with P -9 = 3 or 710 (d) Modeled vs. measured P values from (c). Specific results for each dataset are detailed in Table C.5.

Figure 4.6. Modeled vs. measured P values for the haptophyte E. huxleyi (a) Nitrate-limited chemostat cultures (gold circles; Bidigare et al., 1997; Wilkes et al., 2018), modeled with P = 110-9. (b) Modeled vs. measured P values from (a). (c) Batch cultures grown in continuous light conditions (filled squares) and diel cycles (open squares) (Riebesell et al., 2000b; Rost et al., 2002), modeled with P = 1.8−6.0  10-9. (d)

Modeled vs. measured P values from (c). Specific results for each dataset are detailed in Table C.5.

Figure 4.7. Modeled vs. measured P values for Alexandrium dinoflagellate species (a) Nitrate-limited chemostat cultures (gold circles; Wilkes et al., 2017; Hoins et al., 2016), modeled with P = 2.2 10-9. (b)

Modeled vs. measured P values from (a), including linear fits with and without the two outliers circled in (a). (c) Nutrient-replete batch cultures grown in diel cycles (Hoins et al., 2015), modeled with P = 4.5  -9 10 . (d) Modeled vs. measured P values from (c). Specific results for each dataset are detailed in Table C.5.

Figure 4.8. Modeled vs. measured P values for the diatom P. glacialis. (a) Nitrate-limited chemostat -10 cultures (gold circles; Popp et al., 1998), modeled with P = 4.5 10 . (b) Modeled vs. measured P values from (a). Specific results are detailed in Table C.5.

100

Assessing model performance

The model is most effective at predicting P values for E. huxleyi, with a normalized root mean squared error (nRMSE) of 4.8% across all batch and chemostat culture data for this species. The combined

P. tricornutum datasets also are reproduced well (nRMSE of 8.4%). The full set of Alexandrium experiments has a larger error (nRMSE of 21.4%), but omitting the two outliers (circled in Figure 4.7a) brings this error down to 11.3%. The model is least effective at predicting P values for P. glacialis (nRMSE of 20.6%). Among the combined data (n = 135; after excluding the two Alexandrium outliers), the greatest deviation between modeled and empirical P is a ~6‰ overestimation for one of the chemostat experiments with P. glacialis. Overall, the model is able to reproduce the 135 experiments with an average nRMSE of

2 7.6%. Plotting all modeled P values against the original measurements yields a line with slope 0.94 and r

= 0.89 (Figure 4.9). The model is best at predicting P values at i/[CO2(aq)] ratios between zero and intermediate values. It is less effective at predicting P values in the higher ranges of i/[CO2(aq)], or for experiments in which the data have shallow slopes.

Figure 4.9. Modeled vs. measured P values for all taxa and conditions (n = 135). Note: this fit omits the two dinoflagellate experiments circled in Figure 4.7a. Including these two points changes the equation to y = 0.93x + 1.2 (r2 = 0.86).

101

Discussion While molecular and physiological evidence indicates that phytoplankton deploy a wide range of

CCMs to ameliorate the kinetic limitations of RubisCO, the details of these mechanisms, their isotopic expression under natural environmental conditions, and their relevance to geochemical signatures of global carbon cycling are still debated (Reinfelder, 2011; Matsuda et al., 2011; Bolton and Stoll, 2013; Pagani

2014; Hopkinson et al., 2016; McClelland et al., 2017). Here we assume that the dominant processes influencing organic carbon signatures in eukaryotic phytoplankton can be captured by a single, simplified carbon isotope flux-balance model, because in chemostat culture studies, representatives from three eukaryotic clades with different life strategies and RubisCOs display similar isotopic responses to

µ/[CO2(aq)] (Figure 4.2). Our model posits the widespread importance of a catalyzed process capable of drawing down intracellular CO2, thus maintaining an inward-directed diffusional gradient. We further suggest that this mechanism is distinct from a conventional carbonic anhydrase (CA), and instead assume it is intimately coupled to the thylakoid and its associated, photosynthetically-activated membrane potential.

The primary advantage of such a framework is its ability to explain four major observations: (1) f values, inferred from the intercepts of in vivo measurements, do not match in vitro measurements of RubisCO,

(2) P values respond to cellular surface area and volume (SA/V), consistent with a primarily diffusive mode of CO2 entry at the cell boundary, (3) the availability of nutrients and light affects the expression of P, and

(4) virtually all eukaryotic phytoplankton taxa are known to use CCMs. Therefore, it can provide a simple and general mechanistic explanation for why—in aggregate—algal paleobarometry seems to work.

Influence of light and nutrient conditions on εp

A variety of light- and nutrient-dependent mechanisms previously have been invoked to explain the variable responses of εp between chemostat (nutrient-limited) and batch (light-limited) culture conditions. The largest values of εp (25-27‰) occur under nutrient-limited (high light) conditions. It has been suggested that complementary light-limited experiments may converge with the high light data (i.e.,

102

form a single curve) after accounting for variations in instantaneous- and cell-specific growth rates (Rost et al., 2002). However, for most datasets, differences persist even after applying these corrections, and thus additional factors must influence εp (Figures C.4, C.5).

One of these additional factors may be changes in the proportion of cyclic to linear electron flow.

Cyclic electron flow enhances the production of ATP above the ratio of ATP/NADPH required to produce biomass, and therefore increases the amount of chemical energy available to operate CCMs (Riebesell et al., 2000a, Cassar et al., 2006, Hoins et al., 2016). Thus, larger εp values in nitrate-limited chemostats have been hypothesized to result from greater delivery of CO2 to RubisCO, maximizing the expression of RubisCO

(assumed to equal ~27‰; e.g., Riebesell et al., 2000a,b; Cassar et al., 2006). These studies do not stipulate a particular ATP-dependent CO2 uptake mechanism, but Cassar et al. (2006) localizes the mechanism to the chloroplast based on an energy-minimization modeling approach. However, this explanation warrants re-examination due to the recent discovery that cyclic electron flow contributes negligibly to photosynthesis and the regulation of ATP/NADPH in diatoms (Baullier et al., 2015). It also warrants re-examination because RubisCO is not 27‰ for any of the planktonic taxa measured to-date.

Instead, it is known that specific CCM components including CAs are activated by light and thus responsive to the balance of energy and nutrients (e.g., Hopkinson et al., 2011, 2014; Yamano et al., 2010,

Mitchell et al., 2017). Such regulation has been proposed to mediate the expression of εp (e.g., Hopkinson et al., 2011, 2014). Our generalized model adapts and simplifies this idea by introducing the light-dependent parameter  to quantify a postulated, highly 13C-fractionating hydr(oxyl)ation. Its unidirectional nature and physiological association with the thylakoid membrane imply that such a “photo-CCM” directly connects the light reactions of photosynthesis with the dark reactions of CO2 fixation.

Reconciling existing CCM models: cellular mechanisms and evolutionary pressures

The abiotic hydr(oxyl)ation of CO2 is kinetically very slow. The rate constants for the uncatalyzed

- -2 -1 4 -1 -1 conversion of CO2 to HCO3 and the reverse reaction are ~ 4  10 s and 3  10 kg mol s , respectively

103

(298.15 K, Salinity = 35; Schulz et al., 2006), ca. 106 slower than a typical carbonic anhydrase (Bundy,

1986; Hopkinson et al., 2011). Therefore spontaneous reaction cannot be invoked directly as the source of a unidirectional sink for CO2 (Zeebe and Wolf-Gladrow, 2001, Schulz et al., 2006; Mangan et al., 2016).

Instead, our model proposes that the reaction must be catalyzed and coupled to ion transport across the thylakoid membrane. One possibility is that this process occurs via non-enzymatic catalysis (Keller et al.,

2015), a strategy that might be more favorable for the cell when nitrogen for protein synthesis is scarce.

Another possibility is that non-enzymatic catalysis has provided a template for the selective evolution of enzymes to optimize the process, i.e., allow the mechanism to be better regulated. In

Chlamydomonas, the LCIB/LCIC protein complex structurally resembles a -CA (Jin et al., 2016). This

- complex is hypothesized to play a role similar to that in our model: converting CO2 to HCO3 , potentially with active regulation that minimizes subsequent dehydration (Wang and Spalding, 2014; Jin et al., 2016).

Cyanobacteria also have an analogous strategy. The NAD(P)H dehydrogenase (NDH-1) complexes of cyanobacteria are essential for CO2 uptake, are coupled to cyclic electron flow around photosystem I, and are expressed in a light-dependent manner (Ogawa, 1992; Ogawa and Mi, 2007). Some of these complexes

- are thought to catalyze the conversion of CO2 to HCO3 (Kaplan and Reinhold, 1999; Tchernov et al., 2001;

Eichner et al., 2015), potentially rendering the process irreversible through the transport of a proton across the thylakoid membrane (Maeda et al., 2002; Price et al., 2002; Zhang et al., 2004). Unlike in our model,

- in cyanobacteria the HCO3 accumulates in the cytosol (i.e., there appears to be no physical cooperation between the thylakoid and the cyanobacterial analogue of the pyrenoid—the carboxysome). Although specific involvement of NDH in hydr(oxyl)ation remains unknown for red-lineage phytoplankton, a functional homologue of this enzyme or the LCIB protein may be utilized. Our invoked hydr(oxyl)ation step is similar to that proposed for conventional, intracellular CAs in prior models (e.g., the “chloroplast pump model” of Hopkinson, 2014), but with the important distinctions that it would be irreversible and would respond directly to cellular photon flux.

There are several reasons to think that a generalized strategy of photocatalytic hydr(oxyl)ation is both plausible and likely to be widely distributed among eukaryotes. Light-dependent pH gradients within

104

cells provide uniform physiological pressures that could support the convergent evolution of enzymes and/or cofactors catalyzing this process. When cells are illuminated, an electric field and proton gradient are generated across the thylakoid membrane. While these gradients are used by ATP synthase to produce

ATP, a secondary role in carbon concentrating mechanisms also has been speculated (Pronina and Borodin,

1993; Raven, 1997; Thoms et al., 2001; Moroney and Ynalvez, 2007; Kikutani et al., 2016, Holtz et al.,

2017, Matsuda et al., 2017). The alkaline chloroplast stroma is naturally suited for hydr(oxyl)ation reactions, while the acidic thylakoid lumen and its proximity to the pyrenoid would favor protonation and dehydration of accumulated bicarbonate near RubisCO (Raven, 1997; Thoms et al., 2001). This possibility is supported by the recent discovery of lumen-localized, θ-CAs in diatoms, analogous to the α-CA found in the lumen of freshwater green algae (Table 4.2). Moreover, a thylakoid-energized CCM would take advantage of an intrinsic feature of oxygenic photosynthesis: namely, it would allow cells to exploit their access to an unlimited supply of electron donor to simultaneously improve their ability to access carbon. Another driver for the evolution of a photo-CCM may be protection from photodamage. Cells have evolved numerous strategies to deal with excess photon flux, and these strategies may include enhanced

- - active uptake of HCO3 or the energized conversion of HCO3 to CO2 specifically to consume excess ATP or proton gradients (Tchernov et al., 1997, 2001; Kaplan and Reinhold, 1999; Tchernov and Lipschultz,

2008; Eichner et al., 2015). Kaplan and Reinfeld (1999) speculate that these energy disposal mechanisms could have originated early in phytoplankton evolution before decreases in CO2 levels became physiologically limiting, and were only subsequently adapted for carbon acquisition. An ancient origin could explain why this energized mechanism would be widespread within eukaryotic clades.

Available genomic and physiological evidence indicates that the other CCM components invoked in our model—namely the association of the thylakoids with the pyrenoid, plus collocated CAs—have evolved multiple discrete times in phytoplankton. For example, new phytoplanktonic CAs, lacking sequence homology to known forms but with similar activities, are still being discovered (Kikutani et al.,

2016; Jin et al., 2016; Hopkinson et al., 2016; Shen et al., 2017). Yet, collectively, the behavior of P in cultures and in the marine record is surprisingly coherent, indicating that in most cases the similar

105

physiological functions of these convergently-evolved CCMs reduce to a common set of isotopic consequences.

Physiological predictions of the model

In our model, increases in photon flux correspond to an increase in the relative amount of the hydr(oxyl)ation reaction (higher ). This implies an increase in CO2 availability, i.e., a decreased reliance on diffusion. Physiologically, this implies that P will get larger at a given [CO2(aq)], due to greater expression of HYD relative to RubisCO. Such reasoning can explain the apparent conundrum that P values in light-limited batch cultures of E. huxleyi become larger with increasing C (Rost et al., 2002): the directional change in P is an increase due to higher , while the concomitant, -induced increase in internal CO2 supply also increases C (and therefore /[CO2(aq)], since in this term the CO2 being invoked is the external concentration, which has not changed). In P. tricornutum, P also appears to increase with increasing C in some light-limited batch cultures; however, the pattern is less pronounced (Riebesell et al., 2000). This taxonomic difference may be explained by the larger isotope effect associated with diatom RubisCO

(18.5‰) when compared to E. huxleyi (11‰): the diatom RubisCO value is more similar to the HYD endmember (25‰), and so -induced changes in P are more difficult to resolve.

In chemostat cultures, growth rate is controlled by the delivery of nutrients, so excess light can be directed to hydr(oxyl)ation to meet carbon demand (C) at all growth rates. In the model, this is indexed as

ω  1 (a constant), and the slow step of carbon acquisition is always predicted to have εf_ = 25‰. This means that P always will decrease with increasing C, consistent with measurements from chemostat cultures (Popp et al., 1998; Wilkes et al., 2017).

The model also makes predictions about the percent bicarbonate import across the cell membrane.

Consistent with measurements (Burkhardt et al., 2001; Rost et al., 2003, 2006; Eberlein et al., 2014), the

- percent HCO3 import is predicted to increase with declining CO2 in all experiments. Our predictions broadly agree with estimates from earlier models. For example, Hopkinson et al., 2011 conclude that ~30%

106

- of net inorganic carbon uptake by P. tricornutum is from import of HCO3 ; our model results for light- limited batch cultures are similar, except in the lowest CO2 conditions (Table C.5). However, our model disagrees with the results from some studies that used membrane-inlet mass spectrometry (MIMS) and disequilibrium isotope techniques (Rost et al, 2003; Hoins et al., 2016b). The worst agreement is for the

- dinoflagellate Alexandrium: our model predicts < ~10% HCO3 import, while Hoins et al., 2016b report

~50%. Part of the discrepancy may be due to methodological limitations of the analyses. Measurements

- using these approaches can significantly overestimate HCO3 import due to the deactivation of extracellular carbonic anhydrase, and/or due to the constant assay pH value of 8 (Burkhardt et al., 2001; Hoins et al.,

2016b; Chrachri et al., 2018). However, increasing the adjustable parameter “E” in our model also would lessen the discrepancy by increasing the gross flux of bicarbonate across the cell membrane without significantly changing our P predictions.

Conclusions: Implications for interpreting paleoenvironmental conditions

The experimentally determined relationship, P_maximum = εf = 25-27‰, influences both the paleoenvironmental and the evolutionary conclusions drawn from paired organic and inorganic carbon isotope records. This value has been argued to establish an upper limit of sensitivity of P to changes in

CO2 concentrations that is ~8-10 times present atmospheric levels, or ~2200 ppm CO2 (e.g., Freeman, 2001;

Freeman and Pagani, 2005; Pancost et al., 2013). Within this framework, any variability in sedimentary P records would imply that pCO2 is below this level—regardless of the algal taxonomic source or nutrient environment (Freeman, 2001). Conversely, if temporal records of P are approximately 25‰ and do not vary, this would argue that atmospheric pCO2 is above the threshold of sensitivity.

The model presented here suggests another cause for variations in P values, even under high pCO2 atmospheres: a relative limitation of growth by photon flux. In this resource condition, P values would still scale directly with pCO2, but the maximum value of P would be defined by f_. Thus, the maximum expression of P, even at the limit of very high CO2, is influenced by the taxonomic identity of the

107

phytoplankton contributing to carbon export and burial, as well as how and where they grow. This may help to reconcile apparent contradictions in the sedimentary record. For example, it may help to explain time intervals (e.g., the Cambrian through the Devonian, or the late Jurassic) when geochemical models predict pCO2 exceeded 2200 ppm (Berner and Kothavala, 2001), yet reconstructed P values are < 25‰ and show variability (Hayes et al., 1999; Kuhn, 2007; Pagani, 2014). Variations in P during these times might reflect enhanced nutrient availability coinciding with major ecological transitions, including the diversification and radiation of coccolithophores and dinoflagellates during the Jurassic (Wiggan et al.,

2018). Estimates of P for the Phanerozoic consistently do not exceed 25-27‰ (Hayes et al., 1999).

In modern oceans, it is already well-established that nutrient conditions influence p values through associated changes in  (Bidigare et al., 1997; Pagani, 2014). In our model, the distinction is that when CO2 is very high and nutrients are replete, the maximum expression of p becomes dependent on RubisCO, which has taxonomic variability. While accounting for both taxonomy and resource limitation certainly complicates interpretation of the sedimentary record, it also adds a dimension of predictive power with respect to prevailing growth conditions. For example, where P records do reach a maximum of 25‰, this implies both high ambient CO2 levels and a resource environment low in nutrients (i.e., oligotrophy).

Field studies in the modern environment demonstrate these points. For example, our model suggests a revised explanation for why modern coastal diatoms growing in upwelling zones are relatively 13C-

13 enriched—their values of  may be near zero and f_ may be approaching RubisCO, promoting a greater C

- enrichment regardless of the uncertain or variable extent of active HCO3 uptake (Fry and Wainright, 1991;

Pancost et al., 1997; Hansman and Sessions, 2015). Illustrating the nutrient-limited principle, modern haptophyte algae growing in oligotrophic waters produce P values as large as ~17.8‰ (Laws et al., 2001;

Pagani, 2014).

However, the best examples of f_ approaching RubisCO may be the present-day Peru upwelling zone. Here, alkenone-based P values from nitrate-replete waters correlate with [CO2(aq)] and reach a

-1 maximum measured value of 11.2‰ under the highest CO2 condition (~29 mol kg ; Bidigare et al., 1997;

108

Pancost et al. 1997). This is in excellent agreement with the ~11‰ value of RubisCO for E. huxleyi (Boller et al., 2011), which would be predicted to be the dominant isotopic control as →0 in this high nutrient flux setting. Simultaneously, diatom P values (from diatom biomarkers; Pancost et al., 1997), exhibited P values with a less well-defined relationship to CO2, but reaching a maximum value of 19.3‰. This is similar to the in vitro isotope effect for diatom RubisCO (18.5‰). Critically, the diatom P measurements cannot be explained solely in terms of low [CO2(aq)] or fast growth rates: the P values are lower than would be predicted using the classic εP = εf – µ/[CO2(aq)] equation applied to the ambient conditions (Laws et al., 1995;

Pancost et al., 1997). Thus, the Peru upwelling results can be understood in terms of a greater influence of

RubisCO on the expression of P in these high-nutrient conditions.

Contrasting with the Peru data, alkenone-derived εP values from oligotrophic systems are large

(~14−19‰), even though [CO2(aq)] in these systems is lower than in the Peru upwelling zone (Bidigare et al., 1997). Our model predicts that this is due to a greater relative flux through the hydr(oxyl)ation mechanism (higher ), i.e., εP is controlled primarily by  and C, with little influence from [CO2(aq)].

Overall, our generalized phytoplankton model can account for the important features of carbon isotope fractionation in red-lineage eukaryotic phytoplankton, both in cultures and in the environment. It incorporates CO2 levels, algal physiology, and consensus observations from the literature regarding CCMs.

As with other recent models (e.g., Cassar et al., 2006; Hopkinson et al., 2011), uncertainties in membrane permeabilities (parameter “P”) complicate the direct application of our model for quantitative paleo-pCO2 reconstructions. However, our model introduces clear endmember predictions that are independent of these considerations in the limit as /CO2 → 0 (Equation 1), potentially providing information about multiple facets of past environments (CO2, growth conditions, and/or algal community composition). The model also provides a physiological explanation for why pCO2 approaches appear to work in some contexts, yet yield ambiguous results in others.

It will be valuable to explore whether the isotopic patterns and processes invoked here can be extended to green-lineage phytoplankton, picoeukaryotes, and other taxa that are not yet adequately

109

represented by in vivo and in vitro carbon isotope studies. This will permit assessment of P values earlier in Earth history. Additionally, our model underscores the need to know the isotope effects associated with algal RubisCOs in vitro—both to test the hypothesis underpinning our model and to ensure sufficient context for interpreting the marine isotope record in high-nutrient environments. However, overall, the mechanisms for control of p suggested here point to the central importance of understanding the role of carbon uptake, transformation, and intracellular sequestration. These processes may respond directly to photosynthetic activity in aquatic organisms via links between the energetics of photosynthesis and the process of carbon acquisition. Under conditions of high photon flux, the kinetics and taxonomy of

RubisCO—and the associated RubisCO—may be largely irrelevant to p.

Acknowledgements This material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No. DGE1144152 to E.B.W. Any opinion, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. A.P. is funded by the Gordon and Betty Moore Foundation and by

NASA-NAI CAN6 (PI Roger Summons, MIT). We thank David Johnston, Boswell Wing, Sarah Hurley,

Itay Halevy, and Rich Pancost for helpful discussions, and John Kondziolka for assistance with Matlab script development.

110

References

Bach L. T., Mackinder L. C., Schulz K. G., Wheeler G., Schroeder D. C., Brownlee C., and Riebesell U. (2013) Dissecting the impact of CO2 and pH on the mechanisms of photosynthesis and calcification in the coccolithophore Emiliania huxleyi. New Phytol. 199, 121−134. Badger M. R., Andrews T. J., Whitney S. M., Ludwig M., Yellowlees D. C., Leggat W., and Price G. D. (1998) The diversity and coevolution of Rubisco, plastids, pyrenoids, and chloroplast-based CO2- concentrating mechanisms in algae. Can. J. Bot. 76, 1052−1071. Baullier B., Berne N., Murik O., Petroutsos D., Prihoda J., Tanaka A., Villanova V., Bligny R., Flori S., Falconet D., KriegerLiszkay A., Santabarbara S., Rappaport F., Joliot P., Tirichine L., Falkowski P. G., Cardol P., Bowler C., and Finazzi G. (2015) Energetic coupling between plastids and

mitochondria drives CO2 assimilation in diatoms. Nature 524, 366−369. Beardall J., Griffiths H., and Raven J. A. (1982) Carbon isotope discrimination and the CO2 accumulating mechanism in Chlorella emersonii (Chlorophyceae). J. Phycol. 17, 134-141. Benthien A., Zondervan I., Engel A., Hefter J., Terbrüggen A., and Riebesell U. (2007) Carbon isotopic fractionation during a mesocosm bloom experiment dominated by Emiliania huxleyi: Effects of CO2 concentration and primary production. Geochim. Cosmochim. Ac. 71, 1528–1541.

Berner R. and Kothavala Z. (2001) Geocarb III: A revised model of atmospheric CO2 over Phanerozoic time. Am. J. Sci. 301,182–204. Bidigare R. R., Fluegge A., Freeman K. H., Hanson K. L., Hayes J. M., Hollander D., Jasper J. P., King L. L., Laws E. A., Milder J., Miller F. J., Pancost R., Popp B. N., Steinberg P. A., and Wakeham S. G. (1997) Consistent fractionation of 13C in nature and in the laboratory: Growth-rate effects in some haptophyte algae. Global Biogeochem. Cy. 11, 279−292. Boller A. J., Thomas P. J., Cavanaugh C. M., and Scott K. M. (2011) Low stable carbon isotope fractionation by coccolithophore RubisCO. Geochim. Cosmochim. Ac. 75, 7200−7207. Boller A. J., Thomas P. J., Cavanaugh C. M., and Scott K. M. (2015) Isotopic discrimination and kinetic parameters of RubisCO from the marine bloom-forming diatom, Skeletonoma costatum. Geobiology 13, 33−43. Bolton C. T. and Stoll H. M. (2013) Late Miocene threshold response of marine algae to carbon dioxide limitation. Nature 500, 558–562.

Burkhardt S., Riebesell U., Zondervan I. (1999a) Effects of growth rate, CO2 concentration, and cell size on the stable carbon isotope fractionation in marine phytoplankton. Geochim. Cosmochim. Acta 63, 3729–3741. Burkhardt S., Riebesell U., Zondervan I. (1999b) Stable carbon isotope fractionation by marine phytoplankton in response to daylength, growth rate, and CO2 availability. Mar.Ecol. Prog. Ser. 184, 31–41.

Burkhardt S., Amoroso G., Riebesell U., and Sültemeyer D. (2001) CO2 and HCO3 uptake in marine diatoms acclimated to different CO2 concentrations. Limnol. Oceanogr.. 46, 1378–1391. Bundy H. F. (1986) Comparative kinetics and inhibition of a carbonic anhydrase from Chlamydomonas reinhardii. Comp. Biochem. Physiol. 84B, 63−69.

111

Cassar N. (2003) Carbon-concentrating mechanisms and -carboxylation: Their potential contribution to marine photosynthetic carbon isotope fractionation. Doctoral dissertation, University of Hawai’i. Cassar N., Laws E. A., and Popp B. N. (2006) Carbon isotopic fractionation by the marine diatom Phaeodactylum tricornutum under nutrient- and light-limited growth conditions. Geochim. Cosmochim. Ac. 70, 5323-5335. Chrachri A., Hopkinson B. M., Flynn K., Brownlee C., and Wheeler G. L. (2018) Dynamic changes in carbonate chemistry in the microenvironment around single marine phytoplankton cells. Nat. Commun. 9, 74. Clark I. D. and Lauriol B. (1992) Kinetic enrichment of stable isotopes in cryogenic calcites. Chem. Geol. 102, 217-228. Darienko T., Gustavs L., Eggert A., Wolf W., and Pröschold T. (2015) Evaluating the species boundaries of green microalgae (Coccomyxa, Trebouxiophyceae, Chlorophyta) using integrative taxonomy and DNA barcoding with further implications for the species identification in environmental samples. PLoS One 10, e0127838. Eberlein T., Van de Waal D. B., and Rost B. (2014) Differential effects of ocean acidification on carbon acquisition in two bloom-forming dinoflagellate species. Physiol. Plant. 151, 468−479. Eichner M., Thoms S., Kranz S. A., and Rost B. (2015) Cellular inorganic carbon fluxes in Trichodesmium: a combined approach using measurements and modelling. J. Exp. Bot. 66, 749−759. Farquhar G. D., O’Leary M. H., and Berry J. A. (1982) On the relationship between carbon isotope discrimination and the intercellular carbon dioxide concentration in leaves. Aust. J. Plant Physiol. 9, 121−137. Farquhar G. D., Ehleringer J. R., and Hubick K. T. (1989) Carbon isotope discrimination and photosynthesis. Annu. Rev. Plant. Phys. 40, 503-537. Francois R., Altabet M. A., Goericke R., McCorkle D. C., Brunet C., and Poisson A. (1993) Changes in the 13C of surface water particulate organic matter across the sub-tropical convergence in the S.W. Indian Ocean. Global Biogeochem. Cycles 7, 627−644. Freeman K. H. (2001) Isotopic biogeochemistry of marine organic carbon. Rev. Mineral. Geochem. 43, 579–605.

Freeman K. H. and Pagani M. (2005) Alkenone-based estimates of past CO2 levels: a consideration of their utility based on an analysis of uncertainties. In A History of Atmospheric CO2 and Its Effects on Plants, Animals, and Ecosystems (eds. J. R. Ehleringer, T. E. Cerling, and M. D. Dearing). pp. 35– 61. Springer, New York. Fry B. and Wainwright S. C. (1991) Diatom sources of 13C-rich carbon in marine food webs. Mar. Ecol. Prog. Ser. 76, 149–15. Goericke R., Montoya J., and Fry B. (1994) Physiology of isotopic fractionation in algae and cyanobacteria. In Stable Isotopes in Ecology and Environmental Science (eds. K. Lajtha and R. Michener). pp. 187–221. Blackwell Scientific Publications, Oxford. Griffiths H., Meyer M. T., and Rickaby R. E. M. (2017) Overcoming adversity through diversity: aquatic carbon concentrating mechanisms. J. Exp. Bot. 68, 3689−3695.

112

Guy R. D., Fogel M. L., and Berry J. A. (1993) Photosynthetic fractionation of the stable isotopes of oxygen and carbon. Plant Physiol. 101, 37–47. Hansen G. and Moestrup Ø. (1998) Fine-structural characterization of Alexandrium catanella (Dinophyceae) with special emphasis on the flagellar apparatus. Eur. J. Phycol. 33, 281−291. Hansman R. L. and Sessions A. L. (2015) Measuring the in situ carbon isotopic composition of distinct marine plankton populations sorted by flow cytometry. Limnol. Oceanogr-Meth. 14, 87−99. Hayes J. M. (2001) Fractionation of carbon and hydrogen isotopes in biosynthetic processes. Rev. Mineral. Geochem. 43, 225-277. Hayes J. M., Strauss H., and Kaufman A. J. (1999) The abundance of 13C in marine organic matter and isotopic fractionation in the global biogeochemical cycle of carbon during the past 800 Ma. Chem. Geol. 161,103–125. Hennon, G. M. M., Ashworth, J., Groussman, R. D., Berthiaume, C., Morales, R. L., Baliga, N. S., et al. (2015). Diatom acclimation to elevated CO2 via cAMP signalling and coordinated gene expression. Nat. Clim. Chang. 5, 761–765. doi: 10.1038/nclimate2683 Heureux, A. M. C., Young, J. N., Whitney S. M., Eason-Hubbard M. R., Lee R. B. Y., Sharwood, R. E., Rickaby, R. E. M. (2017) The role of Rubisco kinetics and pyrenoid morphology in shaping the CCM of haptophyte microalgae. J. Exp. Bot. 68, 3959−3969. Höhner R., Aboukila A., Kunz, H-H., and Venema K. (2016) Proton gradients and proton-dependent transport processes in the chloroplast. Front. Plant Sci. 7, 218. Hoins M., Van de Waal D. B., Eberlein T., Reichart G.-J., Rost B., and Sluijs A. (2015) Stable carbon isotope fractionation of organic cyst-forming dinoflagellates: Evaluating the potential for a CO2 proxy. Geochim. Cosmochim. Ac. 160, 267−276. Hoins M., Eberlein T., Groβmann C. H., Brandenburg K., Reichart G.-J., Rost B., Sluijs, A., and Van de Waal, D. B. (2016a) Combined effects of ocean acidification and light or nitrogen availabilities on 13C fractionation in marine dinoflagellates. PLoS ONE 11, e0154370.

Hoins M., Eberlein T., Van de Waal D. B., Sluijs A., Reichart G-J., and Rost B. (2016b) CO2-dependent carbon isotope fractionation in dinoflagellates relates to their inorganic carbon fluxes. J. Exp. Mar. Biol. Ecol. 481, 9−14. Holtz L-M., Wolf-Gladrow D., and Thoms S. (2017) Stable carbon isotope signals in particulate organic and inorganic carbon of coccolithophores – A numerical model study for Emiliania huxleyi 420, J. Theor. Biol. 117−127.

Hopkinson B. M., Dupont C. L., Allen A. E., Morel F. M. M. (2011) Efficiency of the CO2-concentrating mechanism of diatoms. Procoeedings of the National Academy of Sciences USA 108:3830–3837.

Hopkinson B. M. (2014) A chloroplast pump model for the CO2 concentrating mechanism in the diatom Phaeodactylum tricornutum. Photosynth Res 121, 223−233.

Hopkison B. M., Dupont C. L., and Matsuda Y. (2016) The physiology and genetics of CO2 concentrating mechanisms in model diatoms. Curr. Opin. Plant Biol. 31, 51−57.

Jasper J. P. and Hayes J. M. (1990) A carbon isotope record of CO2 levels during the late Quaternary. Nature 347, 462−464.

113

Jin S., Sun J., Wunder T., Tang D., Cousins A. B., Sze S. K., Mueller-Cajar O., and Gao Y-G. (2016) Structural insights into the LCIB protein family reveals a new group of -carbonic anhydrases. Proc. Natl. Acad. Sci. 113, 14716−14721. Kaplan, A. and Reinhold, L. (1999) CO2 concentrating mechanisms in photosynthetic microorganisms. Annu. Rev. Plant Physiol. Plant Mol. Biol. 50, 539−570. Keller M. A., Piedrafita G., and Ralser M. (2015) The widespread role of non-enzymatic reactions in cellular metabolism. Curr. Opin. Biotech.34, 153−161. Kikutani, S., Tanaka, R., Yamazaki, Y., Hara, S., Hisabori, T., Kroth, P.G., and Matsuda, Y. (2012) Redox regulation of carbonic anhydrases via thioredoxin in chloroplast of the marine diatom Phaeodactylum tricornutum. J. Biol. Chem. 287, 20689−20700. Kikutani S., Nakajima K., Nagasato C., Tsuji Y., Miyatake A., and Matsuda Y. (2016) Thylakoid luminal θ-carbonic anhydrase critical for growth and photosynthesis in the marine diatom Phaeodactylum tricornutum Proc. Nat. Acad. Sci.113, 9828−9833.

Kuhn T. (2007) The evolution of the photosynthetic carbon isotope fractionation εp of marine phytoplankton during the Devonian to Permian time interval Ph.D. diss., University Erlangen-Nuremberg. Laws E. A., Popp B. N., Bidigare R. R., Kennicutt M. C., and Macko S.A. (1995) Dependence of phytoplankton carbon isotopic composition on growth rate and CO2aq: Theoretical considerations and experimental results. Geochim. Cosmochim. Ac. 59,1131–1138.

Laws E. A., Bidigare R. R., and Popp B. N. (1997) Effect of growth rate and CO2 concentration on carbon isotope fractionation by the marine diatom Phaeodactylum tricornutum. Limnol. Oceanogr., 42, 1552−1560. Laws E. A., Popp B. N., Bidigare R. R., Riebesell U., Burkhardt S., and Wakeham S. G. (2001) Controls on the molecular distribution and carbon isotopic composition of alkenones in certain haptophyte algae. Geochem. Geophys. Geosyst. 2, 2000GC000057. Laws E. A., Popp B. N., Cassar N., and Tanimoto J. (2002) 13C discrimination patterns in oceanic phytoplankton: likely influence of CO2 concentrating mechanisms, and implications for palaeoreconstructions. Funct. Plant Biol. 29, 323-333. Mackinder L., Wheeler G., Schroeder D., von Dassow P., Riebesell U., and Brownlee C. (2011) Expression of biomineralization-related ion transport genes in Emiliania huxleyi. Environ. MIcrobiol. 13, 3250−3265. Maeda S., Badger M. R., and Price G. D. (2002) Novel gene products associated with NdhD3/D4- containing NDH-1 complexes are involved in photosynthetic CO2 hydration in the cyanobacterium Synechococcus sp. PCC7942. 43, 425−435. Mangan N. M., Flamholz A., Hood R. D., Milo R., and Savage D. F. (2016) pH determines the energetic efficiency of the cyanobacterial CO2 concentrating mechanism. Proc. Nat. Acad. Sci. 113, E5354−E5362. Marlier J. F. and O’Leary M. H. (1984) Carbon kinetic isotope effects on the hydration of carbon dioxide and the dehydration of bicarbonate ion. J. Am. Chem. Soc. 106, 5054-5057. Matsuda Y., Nakajima K., and Tachibana M. (2011) Recent progresses on the genetic basis of the regulation of CO2 acquisition systems in response to CO2 concentration. Photosynth. Res. 109, 191−203.

114

Matsuda Y., Hopkinson B. M., Nakajima K., Dupont C. L., and Tsuji Y. (2017) Mechanisms of carbon dioxide acquisition and CO2 sensing in marine diatoms: a gateway to carbon metabolism. Philos. Trans. R. Soc. Lond. B. 372, 20160403. McClelland H. L. O., Bruggeman J., Hermoso M, and Rickaby R. E. M. (2017) The origin of carbon isotope vital effects in coccolith calcite. Nat. Comun. 8, 14511. McNevin D. B., Badger M. R., Kane H. J., and Farquhar G. D. (2006) Measurement of (carbon) kinetic isotope effect by Rayleigh fractionation using membrane inlet mass spectrometry for CO2- consuming reactions. Funct. Plant Biol. 33, 1115−1128. McNevin D. B., Badger M. R., Whitney S. M., von Caemmerer S., Tcherkez G. G. B., and Farquhar G. D. (2007) Differences in carbon isotope discrimination of three variants of D-Ribulose-1,5- bisphosphate Carboxylase/Oxygenase reflect differences in their catalytic mechanisms. J. Biol. Chem. 282, 36068−36076. Mitchell M. C., Metodieva G., Metodiev M. V., Griffiths H. and Meyer M. T. (2017) Pyrenoid loss ipairs carbon-concentrating mechanism induction and alters primary metabolism in Chlamydomonas reinhardtii 68, 3891−3902. Montagnes D.J.S., Berges J.A., Harrison P.J., and Taylor F.J.R. (1994) Estimating carbon, nitrogen, protein, and chlorophyll a from volume in marine phytoplankton. Limnol. Oceanogr. 39, 1044- 1060. Moroney J. V. and Ynalvez R. A. (2007) Proposed carbon dioxide concentrating mechanism in Chlamydomonas reinhardtii . Eukaryot. Cell 6, 1251−1259. Morse D., Salois P., Markovic P., and Hastings J. W. (1995) A nuclear-encoded form II RuBisCO in dinoflagellates. Science 268, 1622−1624. Nakajima K., Tanaka A., and Matsuda Y. (2013) SLC4 family transporters in a marine diatom directly pump bicarbonate from seawater. Proc Natl Acad Sci USA 110,1767−1772. Ogawa T. (1992) Identification and characterization of the ictA/ndhL gene product essential to inorganic carbon transport of Synechocystis PCC 6803. Plant Physiol. 99, 1604−1608. Ogawa T. and Mi H. (2007) Cyanobacterial NADPH dehydrogenase complexes. Photosynth. Res. 93, 69−77. Ohkawa H., Sonoda M., Shibata M., and Ogawa T. (2001) Localization of NAD(P)H dehydrogenase in the cyanobacterium Synechocystis sp. Strain PCC 6803. J. Bacteriol. 183, 4938−4939. O’Leary M. H. (1984) Measurement of the isotope fractionation associated with diffusion of carbon dioxide in aqueous solution. J. Phys. Chem. 88, 823−825. O’Leary M. H., Madhaven S., and Paneth P. (1992) Physical and chemical basis of carbon isotope fractionation in plants. Plant, Cell Environ. 15, 1099−1104.

115

Pancost R. D., Freeman K. H., Wakeham S. G., and Robertson C. Y. (1997) Controls on carbon isotope fractionation by diatoms in the Peru upwelling region. Geochim. Cosmochim. Ac. 61, 4983−4991. Pancost R. D., Freeman K. H., Herrmann A. D., Patzkowsky M. E., Ainsaar L., and Martma T. (2013) Reconstructing Late Ordovician carbon cycle variations. Geochim. Cosmochim. Ac. 105, 433–454. Popp B. N., Laws E. A., Bidigare R. R., Dore J. E., Hanson K. L., and Wakeham S.G. (1998a) Effect of phytoplankton cell geometry on carbon isotopic fractionation. Geochim. Cosmochim. Ac. 62, 69– 77. Price G. D., Maeda S., Omata T., and Badger M. R. (2002) Modes of active carbon uptake in the cyanobacterium, Synechococcus sp. PCC7942. Funct. Plant Biol. 29, 131−149. Pronina N. A and Borodin V. V. (1993) CO2 stress and CO2 concentration mechanism: investigation by means of photosystem-deficient and carbonic anhydrase-deficient mutants of Chlamydomonas reinhardtii. Photosynthetica 28, 515−522.

Ratti S., Giordano M., and Morse D. (2007) CO2-concentrating mechanisms of the potentially toxic dinoflagellate Protoceratium reticulatum (Dinophyceae, Gonyaulacales). J. Phycol. 43, 693−701.

Raven JA (1997) CO2-concentrating mechanisms: A direct role for thylakoid lumen acidification? Plant Cell Environ 20(2):147−154.

Raven J. A. and Beardall J. (2015) The ins and outs of CO2. J. Exp. Bot. 67, 1−13. Rau G. H., Takahashi T., Des Marais D. J., Repeta D. J., and Martin J.H. (1992) The relationship between 13  C of organic matter and [CO2(aq)] in ocean surface water: Data from a JGOFS site in the northeast Atlantic Ocean and a model. Geochim. Cosmochim. Ac. 56, 1413−1419. Rau G.H., Riebesell U. and Wolf-Gladrow D. (1996) A model of photosynthetic 13C fractionation by marine phytoplankton based on diffusive molecular CO2 uptake. Mar. Ecol. Prog. Ser. 133, 275−285. Reinfelder J. R. (2011) Carbon concentrating mechanisms in eukaryotic marine phytoplankton. Annu. Rev. Mar. Sci. 3, 291−315. Richier S., Fiorini S., Kerros M-E., von Dassow P., and Gattuso J-P. (2011) Response of the calcifying coccolithophore Emiliania huxleyi to low pH/high pCO2: from physiology to molecular level. Mar. Biol. 158, 551−560. Rickaby R. E. M., Heureux A. M. C., Eason Hubbard M. R., Chan I., McClelland H. L.O, Young J. N., Lee R. B. Y. and Hermoso M. (2015) Everything is not spinach: A history of evolving carbon affinity and isotopic fractionation in marine algal Rubisco. Goldschmidt Abstracts 2634. Riebesell U., Burkhardt S., Dauelsberg A., and Kroon B. (2000a) Carbon isotope fractionation by a marine diatom: Dependence on the growth-rate-limiting resource. Mar. Ecol. Prog. Ser. 193, 295−303.

116

Roeske C. and O’Leary M. H. (1984) Carbon isotope effects on enzyme-catalyzed carboxylation of ribulose bisphosphate. Biochemistry 23, 6275–6284. Roeske C. and O’Leary M. H. (1985) Carbon isotope effect on carboxylation of ribulose bisphosphate catalyzed by ribulosebisphosphate carboxylase from Rhodospirillum rubrum. Biochemistry- US 24,1603–1607.

- Romero M.F., Fulton C.M., and Boron W.F. (2004) The SLC4 family of HCO3 transporters. Pflug Arch Eur J Physiol 447, 495−509. Rost B., Zondervan I., and Riebesell U. (2002) Light-dependent carbon isotope fractionation in the coccolithophorid Emiliania huxleyi. Limnol. Oceanogr. 47, 120−128. Rost B., Riebesell U., and Burkhardt S. (2003) Carbon acquisition of bloom-forming marine phytoplankton. Limnol. Oceanogr. 48, 55−67. Rowan R., Whitney S. M., Fowler A., and Yellowlees D. (1996) Rubisco in marine symbiotic dinoflagellates: form II enzymes in eukaryotic oxygenic phototrophs encoded by a nuclear multigene family. Plant Cell 8, 539−553. Rost B., Riebesell U., Sültemeyer D. (2006) Carbon acquisition of marine phytoplankton: Effect of photoperiod length. Limnol. Oceanogr. 51, 12−20.

Sade Z. and Halevy I. (2017) New constraints on kinetic isotope effects during CO2(aq) hydration and hydroxylation: Revisiting theoretical and experimental data. Geochim. Cosmochim. Ac. 214, 246−265. Satoh D., Hiraoka Y., Colman B., and Matsuda Y. (2001) Physiological and molecular biological characterization of intracellular carbonic anhydrase from the marine diatom Phaeodactylum tricornutum. Plant Physiol. 126, 1459−1470. Schulz K. G., Riebesell U., Rost B., Thoms S., and Zeebe R. E. (2006) Determination of the rate constants for the carbon dioxide to bicarbonate inter-conversion in pH-buffered seawater systems. Marine Chemistry 100, 53−65. Schulz K. G., Rost B., Burkhardt S., Riebesell U., Thoms S., Wolf-Gladrow D. A. (2007) The effect of iron availability on the regulation of carbon acquisition in the coccolithophore Emiliania huxleyi and the significance of cellular compartmentation for stable carbon isotope fractionation. Geochim. Cosmochim. Ac. 71, 5301−5312. Scott K. M., Schwedock J., Schrag D. P., and Cavanaugh C. M. (2004) Influence of form IA RubisCO and environmental dissolved inorganic carbon on the 13C of the chlam-chemoautotroph symbiosis Solemya velum. Environ. Microbiol. 6, 1210−1219. Scott K. M., Henn-Sax M., Harmer T. L., Longo D. L., Frame C. H., and Cavanaugh C. M. (2007) Kinetic isotope effect and biochemical characterization of form IA RubisCO from the marine cyanobacterium Prochlorococcus marinus MIT9313. Limnol. Oceanogr. 52, 2199−2204. Sharkey T. D. and Berry J. A. (1985) Carbon isotope fractionation of algae as influenced by an inducible CO2 concentrating mechanism. In Inorganic Carbon Uptake by Aquatic Photosynthetic Organisms (eds. W. J. Lucas and J. A. Berry). pp. 389−401.

Shen C., Dupont C. L., and Hopkinson B. M. (2017) The diversity of CO2-concentrating mechanisms in marine diatoms as inferred from their genetic content. 68, 3937−3948.

117

13 12 Siegenthalar U. and Münich K. O. (1981) C/ C fractionation during CO2 transfer from air to sea. In Carbon Cycle Modelling (ed. B. Bolin), Wiley, New York, 249−257. Sinetova, M.A., Kupriyanova, E., Markelova, A.G., Allakheverdiev, S.I., Pronina, NA. (2012) Identification and functional role of the carbonic anhydrase Cah3 in thylakoid membranes of pyrenoid of Chlamydomonas reinhardtii. Biochim. Biophys. Ac. 1817, 1248−1255. Spalding M. H. (2008) Microalgal carbon dioxide-concentrating mechanisms: Chlamydomonas inorganic carbon transporters. J Exp Bot 59, 1463–1473. Tabita F. R, Satagopan S., Hanson T. E., Kreel N. E., and Scott S. S. (2008) Distinct form I, II, III, and IV Rubisco proteins from the three kingdoms of life provide clues about Rubisco evolution and structure/function relationships. J. Exp. Bot. 59, 1515−1524. Tachibana M., Allen A. E., Kikutani S., Endo Y., Bowler C., Matsuda Y. (2011) Localization of putative carbonic anhydrases in two marine diatoms, Phaeodactylum tricornutum and Thalassiosira pseudonana. Photosynth. Res. 109, 205−221. Tanaka Y., Nakatsuma D., Harada H., Ishida M., and Matsuda Y. (2005) Localization of soluble beta- carbonic anhydrase in the marine diatom Phaeodactylum tricornutum. Sorting to the chloroplast and cluster formation on the girdle lamellae. Plant Physiol. 138, 207−217.

Tanaka A., Ohno N., Nakajima K., and Matsuda Y. (2015) Light and CO2/cAMP signal crosstalk on the promoter elements of chloroplastic -carbonic anhydrase genes in the marine diatom Phaeodactylum tricornutum. Plant Physiol.170,1105-1116.

Tchernov D., Hassidim M., Luz B., Sukenik A., Reinhold L., and Kaplan A. (1997) Sustained net CO2 evolution during photosynthesis by marine microorganism. Curr. Biol. 7, 723−728. Tchernov D., Helman Y., Keren N., Luz B., Ohad I., Reinhold L., Ogawa T., and Kaplan A. (2001) - Passive entry of CO2 and its energy-dependent intracellular conversion to HCO3 in cyanobacteria are driven by a photosystem I-generated H+. J. Biol. Chem. 276, 23450−23455. Tchernov D., Silverman J., Reinhold L. B., and Kaplan A. (2003) Massive light-dependent cyclingof inorganic carbon between oxygenic photosynthetic microorganisms and their surroundings. Photosynth. Res. 77, 95−103. Tchernov D. and Lipschultz F. (2008) Carbon isotopic composition of Trichodesmium spp. Colonies off Bermuda: effects of colony mass and season. J. Plankton Res. 30, 21−31. Tcherkez G. G. B., Farquhar G. D., Andrews T. J. (2006) Despite slow catalysis and confused substrate specificity, all ribulose bisphosphate carboxylases may be nearly perfectly optimized. P. Natl. Acad. Sci. USA 103, 7246−7251. Thoms S., Pahlow M., and Wolf-Gladrow D. A. (2001) Model of the carbon concentrating mechanism in chloroplasts of eukaryotic algae. J. theor. Biol. 208, 295−313. Usdowski E., Menschel G., and Hoefs J. (1982) Kinetically controlled partitioning and isotopic equilibrium 13 12 of C and C in the system CO2-NH3-H2O. Z. Phys. Chem. 130, 13−21.

Wang Y., and Spalding M. H. (2014) Acclimation to very low CO2: contribution of limiting CO2 inducible proteins, LCIB and LCIA, to inorganic carbon uptake in Chlamydomonas reinhardtii. 166, 2040−2050.

118

Whitney S. M. and Andrews T. J. (1998) The CO2/O2 specificity of single-subunit ribulose-bisphosphate carboxylase from the dinoflagellate, Amphidinium carterae, Aust. J. Plant Physiol. 25, 131−138. Whitney S. M., Houtz R. L., and Alonso H. (2011) Advancing our understanding and capacity to engineer nature’s CO2-sequestering enzyme, Rubisco. Plant Physiol. 155, 27−35. Whitney S. M., Shaw D. C., and Yellowlees D. (1995) Evidence that some dinoflagellates contain a ribulose-1,5-bisphosphate carboxylase/oxygenase related to that of the alpha-proteobacteria. Proc. Biol. Sci. 1356, 271−275. Wiggan NJ, Riding JB, Fensome RA, Mattioli E. (2018) The Bajocian (Middle Jurassic): A key interval in the early Mesozoic phytoplankton radiation. Earth Sci. Rev. doi:10.1016/j.earscirev.2018.03.009.

Wilkes E. B., Carter S. J., and Pearson A. (2017) CO2-dependent carbon isotope fractionation in the dinoflagellate Alexandrium tamarense. Geochim. Cosmochim. Ac., 212, 48−61. Wilkes E. B., McClelland H. L. O., Rickaby R. E. M., and Pearson A. (2018) Carbon isotope ratios of coccolith-associated polysaccharides of Emiliania huxleyi as a function of growth rate and CO2 concentration. Org. Geochem.119, 1−10. Wong, W. W., Benedict, C. R., and Kohel, R. J. (1979) Enzymic fractionation of the stable carbon isotopes of carbon dioxide by Ribulose-1,5-bisphosphate Carboxylase Plant Physiol. 63, 852−856. Yamano T., Tsujikawa T., Hatano K., Ozawa S., Takahashi Y., and Fukuzawa H. (2010) Light and low- CO2-dependent LCIB-LCIC complex localization in the chloroplast supports the carbon- concentrating mechanism in Chlamydomonas reinhardtii. Plant Cell Physiol. 51, 1453−1468. Young J. N., Heureux A. M. C., Sharwood R. E., Rickaby R. E. M., Morel F. M. M., and Whitney S. M. (2016) Large variation in the Rubisco kinetics of diatoms reveals diversity among their carbon- concentrating mechanisms. J. Exp. Bot. 67, 3445-3456.

Young J. N. and Hopkinson B.M. (2017) The potential for co-evolution of CO2-concentrtating mechanisms and Rubisco in diatoms. J. Exp. Bot. 68, 3751-3762. Zeebe R. E. (2014) Kinetic fractionation of carbon and oxygen isotopes during hydration of carbon dioxide. Geochim. Cosmochim. Ac. 2014, 540-552.

Zeebe R. E. and Wolf-Gladrow, D. (2001) CO2 in Seawater: Equilibrium, Kinetics, Isotopes. Elsevier Oceanography Series, Amsterdam. Zhang P., Battchikova N., Jansen T., Appel J., Ogawa T., and Aro E. M. (2004) Expression and functional roles of the two distinct NDH-1 complexes and the carbon acquisition complex NdhD3/NdhF3/CupA/SII1735 in Synechocystis sp. PCC 6803.

119

Chapter 5

Ongoing work and future directions

Introduction

The work presented throughout this dissertation has focused on understanding the mechanistic underpinning of P in prominent eukaryotic phytoplankton groups for application to the marine isotope record. In Chapter 2 I presented experimental work with E. huxleyi, supporting the use of coccolith- associated polysaccharides (CAPs), coccolith calcite, and alkenones from the sedimentary record to

13 reconstruct the ratio /[CO2(aq)] with fewer assumptions than current approaches. Assuming  CDIC can be independently estimated, this work also paves the way for pCO2 reconstructions deeper in geologic time

(~180 Ma) than presently possible using alkenones (~45 Ma; Pagani, 2005). The work in Chapter 3 establishes that P in the marine dinoflagellate A. tamarense responds to /[CO2(aq)], with similar sensitivity

(f value) to diatoms and coccolithophores, bolstering interpretations of bulk and dinoflagellate-derived P values into the Mesozoic. Importantly, this work also indicates that f is unlikely to reflect the kinetic isotope effect associated with RubisCO. In Chapter 4 I assert a nutrient- and light-dependent model reconciling the heterogeneity in available RubisCO measurements (as inferred from in vitro measurements) with the apparent uniformity in f values (as inferred from in vivo chemostat culture experiments). This model also helps to explain the different isotopic responses measured in nutrient-limited experiments and oligotrophic waters when compared to nutrient-replete batch cultures and upwelling zones.

However, these studies do not explicitly account for contributions to isotope records by taxa that were important prior to the Mesozoic. Early eukaryotes gave rise to two major plastid lineages: one distinguished by its use of chlorophyll c as an accessory pigment (red algae) and another which uses chlorophyll b (green algae). The taxa that were the focus of Chapters 1−4 all belong to the red algal lineage and grew to ecological prominence in the Meso- and Cenozoic. By contrast, green algae were likely prominent in the later Neoproterozoic and Early Paleozoic and continued to play an important role until at least the Triassic (Falkowski et al., 2004; Knoll et al., 2007; Brocks et al., 2017). Thus, in order to extend

120

the interpretations from my thesis to other critical intervals of geologic time, it will be essential to explore whether the patterns observed for diatoms, coccolithophores, and dinoflagellates extend to taxa with distinct evolutionary histories and longer geological records.

To this end, I am presently conducting a series of chemostat culture experiments with the green alga Micromonas pusilla which may help to answer these questions. Micromonas has a cosmopolitan distribution in modern oceans (Whitney and Lomas, 2016), and belongs to the prasinophytes which are thought to retain genome characteristics of the ancestral alga that gave rise to plants (Worden, 2009;

Rynearson and Palenik, 2011). Like land plants, Micromonas employs Form IB RubisCO. While this form of the enzyme is well-characterized in higher plants, with an RubisCO value of 26-30‰, its discrimination against 13C has not yet been characterized in vitro from any green algal source. Micromonas is also one of the smallest known eukaryotes, falling into the picoplankton size class (cell diameters < 2 m); thus, cell size and geometry may be significant controls on isotope fractionation in this taxon.

Chemostat culture experiments will help to establish whether green algae fractionate carbon similarly to red-lineage species or require special consideration in the geologic record. These experiments will inform biogeochemical interpretations of the pre-Mesozoic carbon isotope record and will provide an additional test of my theoretical model presented in Chapter 4.

Materials and Methods

Chemostat culture methods

Axenic cultures of Micromonas pusilla (CCMP 1545, isolated from temperate coastal waters near

Plymouth, England in 1950) were obtained from the National Center for Marine Algae and Microbiota

(East Boothbay, Maine). M. pusilla was grown in a 4-l chemostat culture with nitrate as the growth-limiting nutrient. Temperatures were maintained at 19 C by circulating water through a chiller and through the glass jacket of the growth chamber. The vessel was stirred at 50 rpm. Continuous light was provided by cool white fluorescent daylight tubes (~150 μmol photons m-2 s-1, 400−700 nm radiation). Temperature was monitored continuously using an in situ probe (Sartorius Stedim). Growth rates of 0.23 − 0.48 d-1 were

121

set by adjusting the rate of dilution with fresh medium and outflow of effluent using peristaltic pumps

(Table 5.1).

Table 5.1. Steady-state algal characteristics under three CO2 and growth rate combinations.

Parameter Experiment 1 Experiment 2 Experiment 3

[CO2(aq)] (mol kg-1) 30.6  0.9 12.8  0.4 21.1  0.5

Growth Rate (d-1) 0.23  0.01 0.48  0.04 0.36  0.01

Cell Abundance (x106 ml-1) 3.0  0.2 2.9  0.2 2.8  0.2

Chl a (fg cell -1) 15  5 6.4  1.3 14  4

Cell diameter (m ) 1.5  0.8 1.7  0.9 1.7  0.6

Cells were grown in 0.2 m sterile-filtered natural seawater from the Gulf of Maine, which also was autoclaved for Experiments #1 and #3. The seawater medium was amended with macronutrients, trace metals, vitamins (L1 medium except for silicate, which was omitted; Guillard and Hargraves, 1993), and nitrate, which was added to achieve a final concentration of ~20 M. The seawater medium was prepared with Kanamycin (12.5 g ml-1), Neomycin (20 g ml-1), and Penicillin (25 g ml-1) to maintain axenic conditions throughout the experiments. Cultures were checked routinely for bacterial contamination by streaking marine broth agar plates; all tests were negative.

Steady state conditions for each experiment were confirmed by monitoring cell density, carbonate system parameters, residual phosphate and nitrate concentrations, chlorophyll content, and cell size characteristics. Cell growth, motility, and morphology were monitored routinely using a light microscope.

Daily cell counts were performed on cells fixed with Lugol’s iodine solution (4% v/v) using a hemocytomer

122

counting chamber and a microscope. Additional samples were preserved in paraformaldehyde (0.8% final concentration) and stored at -80 C for cell number and size measurements using a Multisizer 3 Coulter

Counter (Beckman Coulter), yielding reasonable agreement with the microscope counts (Figure D.1).

Residual nitrate and phosphate concentrations were measured by spectrophotometry on 0.22 m filtered and refrigerated samples, using the resorcinol (Zhang and Fischer, 2006) and mixed molybdate

(Strickland and Parsons, 1968) methods, respectively. The concentration of chlorophyll a was monitored by spectrophotometry: 40-100 ml samples were gently filtered onto 0.45 m nitrocellulose filters, extracted overnight in the dark (-20C, 90% acetone), centrifuged, and quantified according to Equation (1) by measuring the absorbance at  = 645, 663, and 750 nm.

[11.87 × E - 1.54 × E -0.08 × E ] × Volume Chl a (μg l-1) = 663-750 645-750 630-750 acetone (1) Volumesample

Carbonate system chemistry Three distinct CO2(aq) concentrations (Table 5.1) were achieved by bubbling the chemostat reservoir

13 with 0.2 m filtered mixtures of compressed 4:1 N2:O2 and CO2 ( CCO2 = -38.58  0.03‰). The pH of the growth medium was monitored continuously using an EasyFerm Plus pH probe (Hamilton) and ranged from 7.81 to 8.16 across experiments (Table 5.2). Total dissolved inorganic carbon (DIC) and alkalinity samples were measured routinely during the last days of approach to, and throughout the duration of, each steady state condition. Samples for DIC were collected with no headspace, poisoned with saturated mercuric chloride, and stored in darkness at 4C. DIC concentrations were measured by coulometric titration (UIC Model 5014 CO2 coulometer with an AutoMate automatic acidification system) at the

University of Florida.

Total alkalinity was determined on filtered samples (0.45-m cellulose acetate syringe filters) and analyzed by Gran titration with 0.01 N HCl solution prepared in a 0.7 M NaCl background to approximate the ionic strength of seawater (Gran, 1952; Dickson et al., 2007). Certified reference materials supplied by

A. G. Dickson (Scripps Institution of Oceanography) were routinely titrated to monitor precision. The

123

carbonate system was calculated from DIC, pH, phosphate, temperature, and salinity using CO2SYS (Lewis and Wallace, 1998; van Heuven et al., 2011; Table 5.2) and the K1and K2 dissociation constants of

Mehrbach et al. (1973) as refitted by Dickson and Millero (1987), and Dickson (1990). Combined uncertainties in calculated inorganic carbon species concentrations were estimated numerically following

Bevington and Robinson, 2003.

Table 5.2. Steady-state carbonate system parameters. The number of measurements for each parameter is reported in parentheses. Parameter Experiment 1 Experiment 2 Experiment 3

[DIC] 1920  9 (8) 1861  3 (7) 1885  7 (10) (mol kg-1)

Total Alkalinity (mol kg-1) 2029  3 (3) 2133  15 (4) 2090  20 (5)

pHa 7.81  0.01 8.16  0.01 7.96  0.01

Temperature (C) 19.03  0.01 18.96  0.02 18.96  0.01

Phosphate (mol l-1) 27.3  1.1 (12) 28.7 1.0 (8) 25.8  1.4 (10) apH measurements are reported on the NBS (IUPAC) scale.

Isotopic analysis

Samples were not harvested for isotopic analysis until cells had completed at least seven doubling times for a given dilution rate and [CO2(aq)] combination. Spent medium and cells were removed through an overflow port, and the effluent was collected at 4C in the dark. Cells from the effluent were harvested by gentle filtration onto combusted glass fiber filters (GF/F) and stored at -80C. Filter punches were

13 thawed, acidified by wet HCl addition (1N) to remove residual DIC, and dried at 50C. Values of  Cbiomass were measured using an elemental analyzer interfaced to a continuous flow isotope ratio mass spectrometer

(EA-IRMS; UC Davis Stable Isotope Facility).

124

13 The carbon isotopic composition of DIC ( CDIC) was measured at the University of Florida Light

Stable Isotope Mass Spec Lab. Samples were analyzed using a Thermo Finnigan DeltaPlus XL isotope ratio mass spectrometer paired with a GasBench II. Temperature-dependent fractionation (Mook et al., 1974)

13 13 was used to calculate the isotopic composition of dissolved CO2 ( CCO2) from  CDIC and absolute temperature (Eq (2); Burkhardt et al., 1999).

13 13 δ CCO2 = δ CDIC + 23.644 – (9701.5/TK) (2)

Preliminary Results and Discussion

Steady-state characteristics

Each experiment reached steady-state between 11 and 24 days after setting the initial conditions

-1 (Figures D. 2–D.4). Steady-state growth rates of 0.23, 0.36 and 0.48 d were achieved, and dissolved CO2 concentrations were 31, 13, and 21 mol kg-1, respectively (Table 5.1). Final cell densities were approximately 3 x 106 cells ml-1 in the three experiments (Table 5.1). Residual N:P ratios were drawn down below 0.17 in each case. Cell diameters were identical (within error) across experiments (~1.6m; Table

5.1). Chlorophyll a quotas were low, ~6.4−15 fg cell-1(Table 5.1), similar to ranges observed in phosphate- limited chemostat cultures (~9.8 −21.5 fg cell-1; Maat et al., 2014) and smaller than ranges observed in light-limited, nutrient-rich batch cultures (21.1-28.6 fg cell-1; Hoppe et al., 2018).

Carbon isotopic compositions

The P values for the first three experiments are 20.5  0.4‰ (Experiment #1), 19.2  0.5‰

(Experiment #2), and 20.6  0.5‰ (Experiment #3) (Figure 5.1). Experiments 1 and 3 are not statistically different from one another. A linear fit through the three experiments implies an intercept of 21‰, which is distinctly lower than the consensus ~25-27‰ value converged on by red-lineage phytoplankton. While intriguing, this result is based on limited data and will require investigation over a wider range of /[CO2] conditions. Any additional scatter in the data could significantly change the predicted intercept. In particular, it will be necessary to verify the P value at a combination of both higher [CO2(aq)] and slower

125

growth rates (i.e., closer to /[CO2] = 0), as well as farther along the /[CO2] axis to confidently characterize

-1 f for this taxon. An experiment at  = 0.15 d and pH = 7.57 is currently in progress towards the first goal.

An additional 1-2 experiments at faster  and/or lower CO2 will also be completed towards the second goal.

Unfortunately, preliminary trials with Micromonas suggest that this organism does not grow well at pH values > 8.25 in the chemostat, so there is a limited accessible range of growth conditions that can be used for the remaining experiments.

Figure 5.1. Photosynthetic carbon isotope fractionation of Micromonas pusilla, plotted as a function of growth rate and CO2 concentration. Experiments #1−3 are labeled and indicate the order in which the experiments were performed.

Comparison with prior studies

In general, relatively little is known about picoeukaryotic contributions to carbon export and burial, or how such contributions would affect the sedimentary carbon isotopic signal. This size-related distinction might be expected to matter because chemostat culture results published for Synechococcus, a picoplanktonic prokaryote, suggest that extremely small microalgae have virtually no isotopic dependence on ambient [CO2(aq)] under nitrogen limitation (Figure 5.2). The relatively flat relationship between P and

/[CO2] implied by the three available Micromonas experiments may be more similar to the results obtained

126

for Synechococcus than to the larger, red-lineage species (Figure 5.2), suggesting that it could be challenging to discriminate taxonomic effects from large vs. small cell physiological effects.

Figure 5.2. Compilation of nitrate-limited chemostat culture experiments (Popp et al., 1998, Wilkes et al., 2017, this study). Micromonas pusilla is plotted in green circles (right). The cyanobacterium Synechococcus (Popp et al., 1998) is plotted in purple diamonds (right).

If additional experiments with Micromonas corroborate the trends implied by the preliminary chemostat data, this finding may help to explain some additional (unpublished) observations from the modern ocean. For example, it may help to explain findings from a modern oceanographic study2 along a cruise transect in a productive region of the Northeast Pacific Ocean. This region is dominated by diatoms, cyanobacteria, and picoeukaryotes (Figure D.5; Ribalet et al., 2010). Biomass samples were collected from the water column (4 stations along the transect, 3 depths per station) and separated into three size fractions by pumping water through filters with sequentially smaller pore sizes. Fatty acids derived from intact glyco- and phospho- lipids in these samples were analyzed by GC-irMS to obtain the carbon isotopic compositions

(δ13C values) of individual lipids. Mass-weighted-averages of the δ13C values of fatty acids are displayed

2 I performed the total lipid extractions, purifications, derivatizations, and a portion of the instrumental analyses for this study.

127

in Figure 5.3. The resulting isotopic compositions were all surprisingly 13C-depleted, especially for a highly productive regime. For the majority of natural phytoplankton in contemporary oceans, εp falls in the range of 10-16‰ (Bidigare et al., 1997; Killops and Killops, 2005; Kuhn 2007). Some fatty acids in this study, however, had mass-weighted εp values of approximately 20‰. (Figure 5.4). These isotopically depleted fatty acids were collected from stations 6 and 8, which are dominated by Synechococcus and Ostreococcus

(a picoeukaryotic green algal species). By contrast, the nearer-shore stations with relatively 13C-enriched lipids are characterized by a more prominent pennate diatom population. Given that the maximal isotopic fractionation in Synechococcus has been established as approximately 17‰, and that P values for eukaryotic organisms would be expected be ≤ this value at typical [CO2(aq)] and growth rates, εp values of approximately 20‰ or larger are not easily explained. It is possible that picoeukaryotic green algae are responsible for the anomalously 13C-depleted lipids. These cells, growing in oligotrophic conditions, may consistently fractionate carbon at or near ~20-21‰ (the value observed in chemostat cultures), regardless of ambient [CO2(aq)]. This may drive εp to relatively large values in regions where these organisms dominate.

It should be emphasized that the preliminary intercept from the Micromonas chemostat culture experiments differs from the isotope effect associated with Form IB RubisCO from land plants (~26-30‰ in vitro). The experiments with Synechococcus intercept at a value of 17‰, which is also significantly different from the in vitro measurement of 24‰ for RubisCO from the marine cyanobacterium

Prochlorococcus. Thus, the full suite of available chemostat culture experiments uphold a central premise of the model presented in Chapter 4: f ≠ RubisCO. However, if further experiments confirm that Micromonas approaches an f value of ~21‰ rather than 25‰, then specific controls on carbon isotope expression in picoplanktonic species will require closer inspection. Comparisons with light-limited batch cultures of

Micromonas, in vitro measurements of RubisCO from green algae, or chemostat cultures using green algal species of larger cell size would significantly clarify and advance the work presented in this dissertation.

128

Figure 5.3. Weighted-mean δ13C values of fatty acids derived from glycolipids from the 1.6-53μm size class, Line P, depicted as a function of station number and depth. Station 1 is nearest to shore. Glycolipids primarily reflect phytoplankton contributions. Station 1 is dominated by diatoms and exhibits the most enriched δ13C values. The most depleted samples are from Station 6, which is dominated by picoeukaryotes and prokaryotes.

13 Figure 5.4. Average εp values were estimated for three particle size classes. Values of δ CDIC were not 13 available in this study for calculating εp, so the above values reflect estimates. δ CCO2 was calculated according to the equation of Mook et al., 1974, ε = 24.12-9866/T, where T is the absolute temperature in - Kelvin, and ε is the carbon isotope fractionation of CO2(aq) with respect to HCO3 . The reference value of 13 - δ C of HCO3 in sea surface water was taken to be +1.4‰ based on the average isotopic composition of - HCO3 measured along another transect in the Northeast Pacific (averaged from Schmittner et al., 2013). 13 Values of δ Corg for eukaryotic organisms were calculated by assuming a constant isotopic offset of 4‰ between photosynthetic lipids and algal biomass.

129

References

Bevington P. R. and Robinson D. K. (2003) Data reduction and error analysis for the physical sciences, 3rd Edition. McGraw-Hill, New York, pp. 1–65. Brocks J. J., Jarrett A. J. M., Sirantoine E., Hallmann C., Hoshino Y. and Liyanage T. (2017) The rise of algae in Cryogenian oceans and the emergence of animals. Nature 548, 578−581.

Dickson A. G. (1990) Standard potential of the reaction: AgCl(s) + 1/2H2(g)=Ag(s) + HCl(aq), and the - standard aciditiy constant of the ion HSO4 in synthetic seawater from 273.15 to 318.15 K.J. Chem. Thermodyn. 22,113-127. Dickson A. G. and Millero F. J. (1987) A comparison of the equilibrium constants for the dissociation of carbonic acid in seawater media. Deep-Sea Res. Pt. I 34, 1733-1743.

Dickson A. G., Sabine C. L., and Christian J. R. (2007) Guide to best practices for ocean CO2 measurements. PICES Special Publication 3, 1–191. Falkowski P. G., Katz M. E., Knoll A. H., Quigg A., Raven J. A., Schofield O. and Taylor F. (2004) The evolution of modern eukaryotic phytoplankton. Science 305, 354–360. Gran G. (1952) Determination of the equivalence point in potentiometric titrations, Part II. Analyst 77, 661– 671. Guillard R. R. L. and Hargraves P. E. (1993) Stichochrysis immobilis is a diatom, not a chrysophyte. Phycologia 32, 234–236. Hoppe C. J. M., Flintrop C. M., and Rost B. (2018) The arctic picoeukaryote Micromonas pusilla benefits synergistically from warming and ocean acidification. Biogeosciences doi:10.5194/bg-2018-28. Knoll A. H., Summons R. E., Waldbauer J. R. and Zumberge, J. E. (2007) The geological succession of primary producers in the oceans In Falkowski P, Knoll A, editors, Evolution of Primary Producers in the Sea, pp. 133–163. Boston, Elsevier.s

Lewis E. and Wallace D. (1998) Program developed for CO2 system calculations: Oak Ridge TN Oak Ridge National Laboratory Environmental Sciences Division, v. 4735.

Maat D. S., Crawfurd K. J., Timmermans K. R., and Brussaard C. P. D. (2014) Elevated CO2 and phosphate limitation favor Micromonas pusilla through stimulated growth and reduced viral impact. Appl. Environ. Microb. 80, 3119−3127. Mehrbach C., Culberson C. H., Hawley J. E., and Pytkowicz R. M. (1973) Measurement of the apparent dissociation constants of carbonic acid in seawater at atmospheric pressure. Limnol. Oceanogr. 18, 897–907. Mook W. G., Bommerson J. C., and Staverman W. H. (1974) Carbon isotope fractionation between dissolved bicarbonate and gaseous carbon dioxide. Earth Planet. Sci. Lett. 22, 169–176.

130

Ribalet F., Marchetti A., Hubbard K. A., Brown K., Durkin C. A., Morales R., Robert M., Swalwell J. E., Tortell P. D., ad Armbrust E. V. (2010) Unveiling a phytoplankton hotspot at a narrow boundary between coastal and offshore waters. Proc. Natl. Acad. Sci. 107, 16571–16576. Rynearson T. A. and Palenik B. (2011) Learning to read the oceans genomics of marine phytoplankton. Adv. Mar. Biol. 60, 1–39. Schmittner A., Gruber N., Mix A., Key R., Tagliabue A., and Westberry T. (2013) Biology and air-sea gas exchange controls on the distribution of carbon isotope ratios 13C in the ocean. Biogeosciences 10. Strickland J. D. H. and Parsons T. R. (1968) Determination of reactive phosphorus. In A practical handbook of seawater analysis, Fisheries Research Board of Canada, Bulletin 167, pp. 49–56. Van Heuven S., Pierrot D., Rae J. W. B., Lewis E., and Wallace D. W. R. (2011) MATLAB program developed for CO2 system calculations. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory. Whitney L. A. P. and Lomas M. W. (2016) Growth on ATP elicits a P-stress response in the picoeukaryote Micromonas pusilla. PLOS one 11, e0155158. Worden A. Z., Lee J-H., Mock T., Rouzé P., Simmons M. P., and Aerts A. L. (and 45 other authors) (2009) Green evolution and dynamic adaptations revealed by genomes of the marine picoeukaryotes Micromonas. Science 324, 5924. Zhang, J.-Z. and Fischer, C.J. (2006) A simplified resorcinol method for direct spectrophotometric determination of nitrate in seawater. Marine Chemistry 99, 220–226.

131

Page intentionally left blank

132

Appendix A

Supporting Information for Chapter 2

Table A.1. Carbonate system parameters Expt pH DIC TA Temperature [CO ] [HCO -] [CO 2-] 2(aq) 3 3 # (NBS) (µmol/kg) (µmol/kg) (°C) (µmol/kg) (µmol/kg) (µmol/kg) 1 8.01 1761 2188 17.98 17.6 1643 101 2 8.10 1728 n.d. 17.94 14.0 1595 119 3 8.20 1867 2190 17.94 11.9 1697 158 4 8.21 1744 2181 17.99 10.7 1580 153

Table A.2. Steady-state cell densities and residual nutrient concentrations Expt Cell densit(x106 NO - + NO - [PO 3-] 3 2 4 nd # cells/ml) (µmol/l)a (µmol/l) c

1 2.6 ± 0.3 4 ± 1 18.9 ± 1.3 6 2 3.3 ± 0.2 3 ± 1 24.8 ± 1.9 4 3 2.9 ± 0.2 b 25.7 ± 0.6 4 4 3.3 ± 0.2 2 ± 2 27.8 ± 0.4 3 aThe estimated 1σ error associated with the concentration measurements ranged from 0.9 to 2 mol/l (±4–9mol/l, 95% confidence limits). bConcentrations were indistinguishable from the blank. cThe estimated 1σ error associated with the concentration measurements was 1mol/l (±2.45mol/l, 95% confidence limits). dNumber of samples averaged for cell density and residual nutrient concentrations. Daily samples from the stable steady state served as replicates.

Table A.3. Isotopic fractionation between cellular pools (‰).

Expt # εcalcite-CAP εCAP-biomass εbiomass-alkenone εcalcite-biomass εalkenone-CAP

1 23.6 10.6 3.5 34.4 –13.9 2 25.3 8.8 4.4 34.3 –13.1 3 25.8 7.9 3.7 33.9 –11.4 4 27.7 4.7 3.1 32.5 –7.7

Mean 25.6 8 3.7 33.8 –11.5 SD 1.7 2 0.6 0.9 2.7

133

Page intentionally left blank

134

Table A.4. Three carbon allocation scenarios associated with different flip values. % carbon

Relative  Expt # δbiomass flip δlip fprot δprot fcarb δcarb fCAP δCAP fsacc δsacc allocation to CAPs

Prota 0 1 –48.2 0.15 –52.2 0.70 –48.2 0.15 –44.2 0.14 –38.2 0.86 –45.2 2.1 Carba 4 2 –45.7 0.15 –49.7 0.70 –45.7 0.15 –41.7 0.37 –37.3 0.63 –44.3 5.6 Lipa –4 3 –42.9 0.15 –46.9 0.70 –42.9 0.15 –38.9 0.50 –35.4 0.50 –42.4 7.5

a CAP-sacc 7 4 –39.5 0.15 –43.5 0.70 –39.5 0.15 –35.5 0.91 –34.9 0.09 –41.9 13.7 Prota 0 1 –48.2 0.2 –52.2 0.60 –48.2 0.2 –44.2 0.14 –38.2 0.86 –45.2 2.8

135 a

Carb 4 2 –45.7 0.2 –49.7 0.60 –45.7 0.2 –41.7 0.37 –37.3 0.63 –44.3 7.4 Lipa –4 3 –42.9 0.2 –46.9 0.60 –42.9 0.2 –38.9 0.50 –35.4 0.50 –42.4 10.0

a CAP-sacc 7 4 –39.5 0.2 –43.5 0.60 –39.5 0.2 –35.5 0.91 –34.9 0.09 –41.9 18.2 Prota 0 1 –48.2 0.33 –52.2 0.33 –48.2 0.33 –44.2 0.14 –38.2 0.86 –45.2 4.6 Carba 4 2 –45.7 0.33 –49.7 0.33 –45.7 0.33 –41.7 0.37 –37.3 0.63 –44.3 12.2 Lipa –4 3 –42.9 0.33 –46.9 0.33 –42.9 0.33 –38.9 0.50 –35.4 0.50 –42.4 16.5

a CAP-sacc 7 4 –39.5 0.33 –43.5 0.33 –39.5 0.33 –35.5 0.91 –34.9 0.09 –41.9 30.0

Figure. A.1. The CAP extraction method used in the present study was identical to that of Lee et al. (2016). The CAP extracts from that study were subjected to native polyacrylamide gel electrophoresis (PAGE) and stained with Alcian blue, a dye that selectively stains the carboxyl groups of acidic polysaccharides. The PAGE bands shown in this example gel migrated according to mass and verify the presence of a single polysaccharide for each of the strains listed in the legend. This gel also indicates that the CAP band is reproducible for a given strain from different types of cultures (semicontinuous batch vs. bulk). Note that RCC1216 is a strain synonym for E. huxleyi CCMP3266, the strain used in the present study.

136

Figure. A.2. Comparison with E. huxleyi isotopic data from the nitrate-limited chemostat incubations of Bidigare et al. (1997). All values were obtained in nitrate-limited cultures with similar feed media compositions and were maintained at 18 C in continuous light. Data are included for a non-calcifying clone of E. huxleyi (BT6, Bidigare et al., 1997, white circles), a calcifying clone (PLY B92/11, Bidigare et al., 1997, black circles), and the calcifying strain used in this study (CCMP3266, grey circles), plotted as a function of the ratio /[CO2(aq)]. (a) The calculated P value for Experiment #1 (21.4 ± 1.5‰), superimposed on the nine P values reported in Bidigare et al. (1997). The dashed line corresponds to the geometric mean 13 regression reported in Bidigare et al. (1997). (b)  Cbiomass values, with dashed lines corresponding to geometric mean regressions. Note that the vertical axis is plotted with 13C values growing progressively lighter moving up-axis. The strongly 13C-depleted composition of the seawater medium in Experiment 1 13 (this study) helps to explain the offset in the δ Cbiomass values between the two experiments.

137

Figure. A.3. Cellular carbon allocation in E. huxleyi if prot and carb are assumed to be 1 and 3‰ enriched relative to biomass, respectively. Arrow widths and shading indicate fractional fluxes and isotopic compositions. The main differences from the scenario depicted in Figure. 2.3a are a relatively larger flux to proteins and a smaller flux to bulk carbohydrates, resulting in total % carbon allocations to CAPs that are approximately half those predicted in the main text.

138

Page intentionally left blank

139

Appendix B Supporting Information for Chapter 3

Table B.1. Three experiments were conducted under approximately equivalent CO2 conditions ([CO2(aq)] = 23.7  1.1 μmol kg-1) with growth rates varied from 0.14 to 0.35 d-1, and four experiments were carried out -1 at a growth rate of 0.24 d with [CO2(aq)] varied six-fold.

μ (d-1)

)

1 - 0.14 0.24 0.35

10 ✓

] (μmol kg (μmol ] 13 ✓ 2(aq)

~ 24 ✓ ✓ ✓ [CO 63 ✓

140

Table B.2. Summary of the carbonate chemistry parameters and isotope measurements for each experiment

3- 13 13 [DIC] [PO4 ] Total Alkalinity δ CDIC δ Cbiomass Expt Description pHa (µmol kg-1) (µmol l-1) (µeq kg-1) (‰) (‰)

Mean value ± 2338 ± 14 7.59 ± 0.03 30 ± 3 2230 ± 19 -24.5 ± 0.4 -59.3 ± 0.2 1σ 1 N 8 6624 17 21 8 6 Days 28,33,40,42, 24,26,28,33,38,42,44,45 24-45 22-39 22,24,26,28,33,38,42,44,45 24,26,28,33,38,42,44,45 combinedb 44,45 Mean value ± 1989 ± 16 7.95 ± 0.01 26 ± 1.8 2120 ± 30 -23.7 ± 0.5 -54.5 ± 0.6 1σ 2 N 10 6621 25 26 10 3 Days 19,21,25,27,31,35,37,39 19,21,25,27,31,35,37,39,41,43 19-43 19-43 19-43 (every other day) 28,37,41 combined ,41,43 Mean value ± 2113 ± 19 7.98 ± 0.02 28.3 ± 1.1 2260 ± 25 -19.3 ± 0.3 -49.2 ± 0.6 1σ 3 N 8 2880 11 17 8 4 141 Days 26,29,33,36 26-36 26-36 20-36 (every other day) 26,29,33,36 29,30,32,36

combined Mean value ± 2091± 7 7.94 ± 0.02 28.8 ± 0.8 2237 ± 37 -18.7 ± 0.3 -46.6 ± 0.4 1σ 4 N 6 2300 11 12 5 2 Days 8,12,16 8-16 8-18 8,10,12,14,16,18,19 8,12,16 16,17 combined Mean value ± 1975 ± 12 8.19 ± 0.01 27.3 ± 2.5 2143 ± 23 -15.7 ± 0.14 -41.9 ± 0.5 1σ 5 N 4 1746 12 14 4 5 Days 16,18,19,20, 17,19,21,23 17-23 11-22 11-23 (every other day) 17,19,21,23 combined 21 Mean value ± 1924 ± 26 8.27 ± 0.02 30.1 ± 1.5 2132 ± 30 -12.7 ± 0.6 -36.6 ± 1.4 1σ 6 N 4 2140 8 6 4 3 Days 14,16,18,22 14-22 14-22 14-22 (every other day) 14,16,18,22 14,15,18 combined areported on the IUPAC (NBS) scale bsampling days included in the mean value, with day 1 corresponding to the beginning of each experimental condition

TableTable B.3. S3. Cell Cell dimensions dimensions

Expt. # Diameter (µm) Height (µm) Volume (µm3)a SA/V (µm)b n Est. POC quota (pgC/cell)c 1 n/a n/a n/a n/a n/a n/a 2 33.4 ± 2.5 37.0 ± 2.7 21852 ± 4753 0.175 ± 0.012 41 2947 ± 523 3 35.1 ± 2.9 39.1 ± 4.1 25685 ± 6655 0.167 ± 0.014 117 3363 ± 712 4 35.1 ± 3.0 38.6 ± 3.3 25439 ± 6446 0.167 ± 0.013 82 3336 ± 690 5 34.8 ± 2.5 38.9 ± 3.2 24928 ± 5563 0.168 ± 0.012 66 3282 ± 598 6 36.2 ± 3.4 41.0 ± 4.4 28840 ± 7871 0.161 ± 0.015 30 3697 ± 824 aVolume computed for each cell using the formula for a rotational ellipsoid (V=(π/6)*(d 2)*(h); Olenina et al. , 2006) bSurface area computed for each cell using the approximate formula SA = 4π[((d/2)p(h/2)p+(d/2)p(h/2)p+(d/2)p(d/2)p)/3]1/P where p=1.6075 c Estimated POC quota was calculated according to the equation for photoautotrophic dinoflagellates of Menden-Deuer and Lessard (2000)

Table B.4. Experimental design comparison with Hoins et al. (2016b)

A. fundyensea S. trochoideaa A. tamarenseb μ (d-1) 0.15 0.2 0.14, 0.24, or 0.35 Instantaneous μ (d-1)c 0.24 0.32 0.14, 0.24, or 0.35 -1 [CO2(aq)] (μmol kg ) 8.4-36 11.2-30 10.2-63 Light:Dark Cycle 16:8 16:8 24:0

Light Intensity (μmol 250 250 150 photons m-2 s-1) shaker table, shaker table, bubbled and stirred, Mixing 16 rpm 16 rpm 50 rpm Temperature (°C) 15 15 18 Salinity 34 34 32.5 Initial Phosphate (μM) 6.25 6.25 36 Initial Nitrate ( μM) 16 8 100 Cell Density < 103 < 103 104 Reservoir volume (L) 2.1 2.1 4 pH monitoring frequency 2-3 days 2-3 days Continuously aHoins et al. (2016a). Experimental details described in Eberlein et al. (2016) bThis study cCalculated following Riebesell et al. (2000a,b) and Burkhardt et al. (1999a)

142

Figure. B.1. The chemostat system and its associated components. Cellular growth rate at steady state was set by the pumping rate of nitrate-limited medium divided by the 4-l culture volume. The chemostat reservoir was continuously bubbled with a regulated mixture of tank CO2 and tank N2/O2 and was maintained at a constant temperature of 18.00  0.01 °C by a recirculating chiller connected to the double jacketed culture reservoir. Full carbonate system parameters were characterized, including the 13 concentration and δ C of CO2

143

Appendix C Supporting Information for Chapter 4 Model Topology The model employs a simplified cellular architecture representing a generic, eukaryotic algal cell

(Figure C.1). The processes described in the main text could plausibly have different localizations within the cell. One variation is depicted in Figure C.2.

Figure C.1. Model structure. Dashed lines indicate passive fluxes while solid lines indicate active transport processes and/or enzymatic conversions. Steps involving enzymes are labeled and represent RubisCO, - carbonic anhydrase, and a putative enzyme catalyzing the active hydr(oxyl)ation of CO2 to HCO3 .

Figure C.2. Alternative model structure capable of explaining broad isotopic patterns. Note that this version contains one fewer carbonic anhydrase enzyme than depicted in Figure C.1.

144

Steady-state isotope flux balance models The isotopic flux balances of 13C for all the intracellular carbon pools are defined below and illustrated in Figure C.1.

(1) extracellular CO2 𝑑𝛿𝑐푖 = 0 𝑑𝑡 𝑑𝑚1 = 휑 − 휑 𝑑𝑡 2 1

(2) chloroplast CO2 𝑑𝛿𝑐푖푖 = (휑 (𝛿 − 𝜀 ) + 휑 (𝛿 − 𝜀 ) − 휑 (𝛿 − 𝜀 ) − 휑 (𝛿 − 𝜀 ) − 휑 (𝛿 − 𝜀 ))/𝑚 𝑑𝑡 1 1 𝐷 7 5 𝐷 2 2 𝐷 3 2 퐻푌𝐷 6 2 𝐷 2 𝑑𝑚2 = 휑 + 휑 −휑 − 휑 − 휑 𝑑𝑡 1 7 2 3 6

- (3) intrathylakoid HCO3 𝑑𝛿𝑏푖푖푖 = (휑 (𝛿 − 𝜀 ) − 휑 (𝛿 − 𝜀 ))/𝑚 𝑑𝑡 3 2 퐻푌𝐷 4 3 𝐶퐴푖 3 𝑑𝑚3 = 휑 − 휑 𝑑𝑡 3 4

(4) intrathylakoid CO2 𝑑𝛿𝑐푖푖푖 = (휑 (𝛿 − 𝜀 ) − 휑 (𝛿 − 𝜀 ))/𝑚 𝑑𝑡 4 3 𝐶퐴푖 5 4 𝐷 4 𝑑𝑚4 = 휑 −휑 𝑑𝑡 4 5

(5) pyrenoid CO2 𝑑𝛿𝑐푖푣 = (휑 (𝛿 − 𝜀 ) + 휑 (𝛿 − 𝜀 )+휑 (𝛿 − 𝜀 ) − 휑 (𝛿 − 𝜀 ) − 휑 (𝛿 − 𝜀 ) 𝑑𝑡 6 2 𝐷 5 4 𝐷 10 7 𝐶퐴푖 7 5 𝐷 11 5 𝐶퐴푖푖

− 휑12(𝛿5 − 𝜀푅푢𝑏푖푠𝐶푂))/𝑚5 𝑑𝑚5 = 휑 + 휑 +휑 − 휑 − 휑 − 휑 𝑑𝑡 6 5 10 7 11 12

- (6) extracellular HCO3 𝑑𝛿𝑏푖 = 0 𝑑𝑡

145

𝑑𝑚6 = 휑 − 휑 𝑑𝑡 9 8

- (7) pyrenoid HCO3 𝑑𝛿𝑏푖푖 = (휑 (𝛿 − 𝜀 ) + 휑 (𝛿 − 𝜀 )−휑 (𝛿 − 𝜀 ) − 휑 (𝛿 − 𝜀 ))/𝑚 𝑑𝑡 8 6 푇 11 5 𝐶퐴푖푖 9 7 𝐷 10 7 𝐶퐴푖 7 𝑑𝑚7 = 휑 + 휑 − 휑 −휑 𝑑𝑡 8 11 9 10

(8) organic matter 𝑑𝛿표𝑟𝑔 = (휑 (𝛿 − 𝜀 ) − 휑 (𝛿 ))/𝑚 𝑑𝑡 12 5 푅푢𝑏푖푠𝐶푂 13 8 8 𝑑𝑚8 = 휑 − 휑 𝑑𝑡 12 13

Empirical Parameters (fixed inputs, from experiments)

i

POC

[CO2(aq)]

ci

bi SA

Tunable Parameters 휔 P C E

Dependent Parameters

𝐶 = 푖𝑃𝑂𝐶 (1 − 휔) 퐻 = [𝐶𝑂2(𝑎푞)]

146

𝑃 푆퐴 [𝐶𝑂2(𝑎푞)] 퐿 = 𝑃 푆퐴 [𝐶𝑂2(𝑎푞)] + 휇𝐶

𝐷 = 퐿 + 퐻

Flux Parameterization (units of mol C cell-1 d-1)

휑12 = 𝐶

휑13 = 휑12 (80%)퐾 휑 퐻 휑 : 푖𝑓 [𝐶𝑂 ] < 푚 , 휑 = 12 , 𝑒𝑙𝑠𝑒 휑 = 0 10 2(𝑎푞) (100−80%) 10 (1−𝐶) 10

휑11 = 𝐶휑10

(휑12 + 휑11 − 휑10) 휑6 = 휔 (1 + 1 − 휔 − 𝐷)

휑7 = 𝐷휑6 휔휑 휑 = 6 5 (1 − 휔)

휑4 = 휑5

휑3 = 휑4 (휑 − 휑 ) 휑 = 10 11 8 (1 − 𝐸)

휑9 = 𝐸휑8 (휑 + 휑 ) 휑 = 9 12 2 1 (퐿 − 1) 휑 휑 = 2 − 휑 1 퐿 8

147

Isotope Effects Table C.1. Assumed kinetic isotope effects used for model calibration.  Associated Process Value RubisCO Carboxylation by RubisCO, measured in vitro Varies (11−24‰) - HYD Kinetic conversion of CO2 to HCO3 25‰

D Passive transport in water (diffusion) 0.7‰ T Active transport of bicarbonate 0.7‰ - Cai Carbonic anhydrase-catalyzed hydration of CO2 to HCO3 1.1‰ - CAii Carbonic anhydrase-catalyzed dehydration of HCO3 to CO2 10.1‰

- Table C.2. Measured and calculated kinetic isotope effects for the conversion of CO2 to HCO3 .

Mechanism Method  Temperature (°C) Source Experimental study 13a NR [1] O’Leary et al. 1992 Calculated from HCO3- dehydration experimental [2] Clark and Lauriol, study and 19.7 0 1992 Hydration of CO2 equilibriumisotopic + - CO2+H2O→H + HCO3 fractionation Calculated based on transition state theory and 23-33 25 [3] Zeebe, 2014 quantum chemistry calculations O’Leary personal communication, 1998, Experimental 11 24 reported in [4] Zeebe and Hydroxylation of CO 2 Wolf-Gladrow, 2001 CO +OH-→HCO - 2 3 [5] Siegenthaler and Experimental 27 20 Munnich, 1981 Experimental 39 18 [6] Usdowski et al., 1982 aPreviously cited by the same group with a different value (6.9‰, [7] Marlier and O’Leary 1984)

References for Table C.2:

[1] O’Leary M. H., Madhaven S., and Paneth P. (1992) Physical and chemical basis of carbon isotope fractionation in plants. Plant, Cell Environ. 15, 1099-1104. [2] Clark I. D. and Lauriol B. (1992) Kinetic enrichment of stable isotopes in cryogenic calcites. Chem. Geol. 102, 217-228. [3] Zeebe R. E. (2014) Kinetic fractionation of carbon and oxygen isotopes during hydration of carbon dioxide. Geochim. Cosmochim. Ac. 2014, 540-552. [4] Zeebe R. E. (2001) CO2 in Seawater: Equilibrium, Kinetics, Isotopes. Elsevier Oceanography Series, Amsterdam. 13 12 [5] Siegenthalar U. and Münich K. O. (1981) C/ C fractionation during CO2 transfer from air to sea. In Carbon Cycle Modelling (ed. B. Bolin), Wiley, New York, 249-257. [6] Usdowski E., Menschel G., and Hoefs J. (1982) Kinetically controlled partitioning and isotopic 13 12 equilibrium of C and C in the system CO2-NH3-H2O. Z. Phys. Chem. 130, 13-21. [7] Marlier J. F. and O’Leary M. H. (1984) Carbon kinetic isotope effects on the hydration of carbon dioxide and the dehydration of bicarbonate ion. J. Am. Chem. Soc. 106, 5054-5057.

148

Summary of model parameters

Table C.3. Model parameter definitions and units.

Parameters Description Units Empirical -1 i Instantaneous growth rate d POC Average particulate organic carbon per cell mol C cell-1 -1 Km Michaelis-Menten half-saturation constant for RubisCO carboxylation mol C kg -1 [CO2(aq)] Concentration of CO2 in seawater medium mol C kg SA Cellular surface area m2 ci Isotopic composition of CO2 in seawater medium ‰ - bi Isotopic composition of HCO3 in seawater medium ‰ Tunable 휔 Relative influx of CO2 to pyrenoid via the hydration/hydroxylation Unitless pathway relative to passive entry. A function of relative nutrient and light availability. -2 -1 -1 P Area-specific mass transport of CO2 through chloroplast and cell kg m cell d membranes—a function of concentration gradients and membrane permeabilities C CA-catalyzed hydration flux in pyrenoid relative to dehydration Unitless - E Passive efflux of HCO3 from cell (as H2CO3) relative to uptake. Unitless Dependent Variables -1 -1 C Carbon-specific growth rate corresponding to carbon fixation during the mol C cell d photoperiod H Proportionality constant governing active uptake of bicarbonate into the Unitless chloroplast. A function of CO2 availability and the relative balance of light and nutrients L Passive efflux of CO2 from cell relative to influx Unitless D Passive efflux of CO2 from pyrenoid relative to influx. Enhanced with Unitless increasing cell leakiness and active bicarbonate uptake into chloroplast.

149

Kinetic isotope effect associated with active hydr(oxyl)ation: HYD

- We assume that the unidirectional formation of HCO3 from CO2 is well-represented by an εHYD value of 25‰. Several experimental and theoretical studies have attempted to determine the uncatalyzed,

- kinetic (abiotic) fractionation during hydration or hydroxylation of CO2 to HCO3 , which might serve as a guideline or bound on our assumption (Table C.2). Unfortunately, the available data are inconsistent. For example, the most widely-adopted, experimentally-determined value of 13‰ was originally reported as

6.9‰ in the primary reference for the study (Marlier and O’Leary, 1984). The same group subsequently cited this value as 13‰ with no explanation for the contradiction (see also discussion in Zeebe, 2014). We assume that this 13‰ value, and the 11‰ value reported by the same group for the analogous hydroxylation reaction, may reflect a problem with back-reactions during their experiments, since reversibility is difficult to prevent and would tend to reduce the expression of the isotope effect (Sade and Halevy, 2017). Once these two numbers are taken out of consideration, we select 25‰ as the most parsimonious value, consistent with theorized hydration values and measured hydroxylation values. A similar value (30‰) was employed in a carbon-flux study for Trichodesmium as a potential fractionation accompanying the cyanobacterial

NHD-14 complex, which is also thought to hydrate or hydroxlate CO2 through a potentially irreversible mechanism (Eichner et al., 2015). A selection of 27‰ could slightly improve our model agreement with existing data sets but is not necessary to explain most available data.

Other model inputs assumed from the literature

Km values were assumed from the literature when measurements were available and are reported in

Table C.5. Km measurements for Form II RubisCO from -proteobacteria (80-89; Jordan and Ogren, 1981) were averaged to estimate the Km associated with peridinin-containing dinoflagellates since the only established catalytic parameter for dinoflagellate RubisCO is its specificity for CO2 relative to O2 (Lilley et al., 2010). This parameter is only used for calculating the point at which a threshold response is induced in the model. We assume that the cell will not invest any energy in active bicarbonate uptake until the concentration of CO2 outside the cell falls below the concentration of CO2 needed in the pyrenoid to saturate

150

RubisCO by 80%. Sensitivity analysis suggests that the choice of Km has little bearing on the patterns observed for the tested data sets, as most do not investigate CO2 levels exceeding this concentration threshold. However, this parameter is included for completeness as it does not make energetic sense for the cell to invest in bicarbonate uptake in the presence of very high CO2 levels, and many CCM components have been shown to be inducible as extracellular CO2 drops from very high levels (i.e., several % CO2) to low levels (analogous to present atmospheric conditions; e.g., Mitchell et al., 2017).

RubisCO values were also assumed from the literature when measurements were unavailable (see

Table 1). For the dinoflagellate Alexandrium sp., RubisCO was assumed to equal 19.5, falling within the ranges measured for R. rubrum (19-24; Guy et al., 1993) and Riftia pachyptila endosymbiont (19.5 ± 1;

Robinson et al., 2003)). For all other experiments, RubisCO was treated as a fixed value constrained by the literature and equaling the values reported in Table 1.

151

References:

Eichner M., Thoms S., Kranz S. A., and Rost B. (2015) Cellular inorganic carbon fluxes in Trichodesmium: a combined approach using measurements and modelling. J. Exp. Bot. 66, 749−759. Guy R. D., Fogel M. L., and Berry J. A. (1993) Photosynthetic fractionation of the stable isotopes of oxygen and carbon. Plant Physiol. 101, 37–47. Jordan D. B. and Ogren W. L. (1981) Species variation in the specificity of ribulose bisphosphate carboxylase/oxygenase. Nature 291, 513-515. Lilley R. M., Ralph P. J., and Larkum A. W. D. (2010) The determination of activity of the enzyme Rubisco in cell extracts of the dinoflagellate alga Symbiodinium sp. By manganese chemiluminescence and its response to short-term thermal stress of the alga. Plant Cell Environ. 33, 995-1004. Marlier J. F. and O’Leary M. H. (1984) Carbon kinetic isotope effects on the hydration of carbon dioxide and the dehydration of bicarbonate ion. J. Am. Chem. Soc. 106, 5054-5057. Montagnes D.J.S., Berges J.A., Harrison P.J., and Taylor F.J.R. (1994) Estimating carbon, nitrogen, protein, and chlorophyll a from volume in marine phytoplankton. Limnol. Oceanogr. 39, 1044- 1060. Riebesell U., Burkhardt S., Dauelsberg A., and Kroon B. (2000a) Carbon isotope fractionation by a marine diatom: Dependence on the growth-rate-limiting resource. Mar. Ecol. Prog. Ser. 193, 295-303.

Riebesell U., Revill, A. T., Holdsworth D. G., and Volkman J. K. (2000b) The effects of varying CO2 concentration on lipid composition and carbon isotope fractionation in Emiliania huxleyi. Geochim. Cosmochim. Ac. 64, 4179-4192 Robinson J. J., Scott K. M., Swanson S. T., O’Leary M. H., Horken K., Tabita F. R., and Cavanaugh C. M. (2003) Kinetic isotope effect and characterization of form II RubisCO from the chemoautotrophic endosymbionts of the hydrothermal vent tubeworm Riftia pachyptila. Limnol. Oceanogr. 48, 48- 54. Rost, B., Zondervan, I., Riebesell, U., 2002. Light-dependent carbon isotope fractionation in the coccolithophorid Emiliania huxleyi. Limnology and Oceanography 47, 120–128. Zeebe R. E. (2014) Kinetic fractionation of carbon and oxygen isotopes during hydration of carbon dioxide. Geochim. Cosmochim. Ac. 2014, 540-552.

152

Figure C.3. Experimental datasets used for model testing, corresponding to the experiments and values reported in Table C.4.

153

Figure C.4. Experimental datasets used for model testing, corresponding to the experiments and values reported in Table C.4.

154

Table C.4. Literature data used to test model. Surface PFD [CO 2(aq) ] Instantaneous   /  Temp. Taxon Species Limitation L:D Cycle POC/cell Area i P Strain Refs. -2 -1 -1 -1 (E m s ) (mol kg ) (i ; d ) (pg) [CO2(aq)] (‰) (°C) (SA; m2) diatom P. tricornutum light 16:8 150 2.1 2.21 8.68 140.0 1.052 9.4 CCAP 1052/1A 15 [1] [2] diatom P. tricornutum light 16:8 150 3.2 2.23 8.35 140.0 0.695 11.6 CCAP 1052/1A 15 [3] [4] diatom P. tricornutum light 16:8 150 4.4 2.21 8.07 140.0 0.507 14.0 CCAP 1052/1A 15 [3] [4] diatom P. tricornutum light 16:8 150 7.2 2.21 8.91 140.0 0.306 15.6 CCAP 1052/1A 15 [3] [4] diatom P. tricornutum light 16:8 150 14.5 2.30 9.57 140.0 0.159 15.7 CCAP 1052/1A 15 [3] [4] diatom P. tricornutum light 16:8 150 37.7 2.27 9.70 140.0 0.060 16.0 CCAP 1052/1A 15 [1] [2] diatom P. tricornutum light 24:0 15 1.8 0.47 13.76 140.0 0.261 9.3 CCAP 1052/1A 15 [4] diatom P. tricornutum light 24:0 15 3.9 0.50 15.86 140.0 0.130 11.5 CCAP 1052/1A 15 [4] diatom P. tricornutum light 24:0 15 8.0 0.51 16.91 140.0 0.064 12.9 CCAP 1052/1A 15 [4] diatom P. tricornutum light 24:0 15 14.9 0.49 18.02 140.0 0.033 14.8 CCAP 1052/1A 15 [4] diatom P. tricornutum light 24:0 30 22.4 1.06 9.29 140.0 0.047 15.1 CCAP 1052/1A 15 [4]

155 diatom P. tricornutum light 24:0 30 2.3 0.98 8.80 140.0 0.432 8.7 CCAP 1052/1A 15 [4] diatom P. tricornutum light 24:0 30 2.8 1.02 8.82 140.0 0.366 9.8 CCAP 1052/1A 15 [4]

diatom P. tricornutum light 24:0 30 3.9 1.00 9.20 140.0 0.260 13.5 CCAP 1052/1A 15 [4] diatom P. tricornutum light 24:0 30 6.9 0.90 8.75 140.0 0.130 13.8 CCAP 1052/1A 15 [4] diatom P. tricornutum light 24:0 30 11.6 1.03 9.04 140.0 0.089 14.3 CCAP 1052/1A 15 [4] diatom P. tricornutum light 24:0 150 1.2 1.33 12.97 140.0 1.147 7.2 CCMP1327 15 [4] diatom P. tricornutum light 24:0 150 2.1 1.64 9.01 140.0 0.781 10.6 CCAP 1052/1A 15 [3] [4] diatom P. tricornutum light 24:0 150 3.2 1.63 8.68 140.0 0.508 17.0 CCAP 1052/1A 15 [3] [4] diatom P. tricornutum light 24:0 150 4.4 1.64 8.40 140.0 0.376 12.6 CCAP 1052/1A 15 [3] [4] diatom P. tricornutum light 24:0 150 6.7 1.26 12.41 140.0 0.188 14.8 CCMP1327 15 [4] diatom P. tricornutum light 24:0 150 7.2 1.41 9.24 140.0 0.195 14.5 CCAP 1052/1A 15 [3] [4] diatom P. tricornutum light 24:0 150 7.6 1.18 15.07 140.0 0.155 14.7 CCMP1327 15 [4] diatom P. tricornutum light 24:0 150 14.5 1.39 9.90 140.0 0.096 15.6 CCAP 1052/1A 15 [3] [4] diatom P. tricornutum light 24:0 150 30.2 1.19 15.86 140.0 0.039 16.2 CCMP1327 15 [4] diatom P. tricornutum light 24:0 150 37.7 1.63 10.03 140.0 0.056 16.7 CCAP 1052/1A 15 [3] [4]

Pageintentionall

156

yleft blank

Table C.4 (continued). Literature data used to test model. diatom P. tricornutum nitrate 24:0 250 0.6 0.77 4.95 100.6 1.280 10.5 UTEX 642 15 [5] diatom P. tricornutum nitrate 24:0 250 0.6 1.38 4.95 100.6 2.163 7.4 CCMP1327 15 [6] diatom P. tricornutum nitrate 24:0 250 0.9 1.04 5.50 100.6 1.134 11.1 CCMP1327 15 [6] diatom P. tricornutum nitrate 24:0 250 1.0 0.50 5.82 100.6 0.509 10.7 CCMP1327 15 [6] diatom P. tricornutum nitrate 24:0 250 1.3 0.98 5.82 100.6 0.754 12.7 UTEX 642 15 [5] diatom P. tricornutum nitrate 24:0 250 1.7 0.98 6.99 100.6 0.576 12.8 UTEX 642 15 [5] diatom P. tricornutum nitrate 24:0 250 2.4 0.75 6.99 100.6 0.318 11.8 CCMP1327 15 [6] diatom P. tricornutum nitrate 24:0 250 2.9 0.75 6.21 100.6 0.256 16.8 CCMP1327 15 [6] diatom P. tricornutum nitrate 24:0 250 3.6 0.50 7.24 100.6 0.140 20.6 CCMP1327 15 [6] diatom P. tricornutum nitrate 24:0 250 6.5 0.50 7.94 100.6 0.077 19.9 CCMP1327 15 [6] diatom P. tricornutum nitrate 24:0 250 10.3 1.40 11.00 100.6 0.136 18.4 CCMP1327 15 [6] diatom P. tricornutum nitrate 24:0 250 10.7 1.25 9.15 100.6 0.117 18.9 CCMP1327 15 [6] diatom P. tricornutum nitrate 24:0 250 12.0 1.00 7.53 100.6 0.083 20.6 CCMP1327 15 [6] diatom P. tricornutum nitrate 24:0 250 15.6 0.75 8.20 100.6 0.048 22.0 CCMP1327 15 [6] diatom P. tricornutum nitrate 24:0 250 22.2 0.50 9.57 100.6 0.022 24.4 CCMP1327 15 [6] diatom P. tricornutum nitrate 24:0 250 34.7 0.50 6.87 100.6 0.014 25.7 CCMP1327 15 [6]

157 diatom P. tricornutum nitrate 24:0 250 0.4 1.00 4.95 100.6 2.500 6.7 CCMP1327 15 [5] diatom P. tricornutum phosphate 24:0 250 0.4 1.02 4.95 100.6 2.550 4.8 CCMP1327 15 [5]

diatom P. tricornutum nitrate 24:0 250 0.7 0.18 4.95 100.6 0.257 21.4 NR 15 [5] diatom P. tricornutum nitrate 24:0 250 1.1 0.93 5.82 100.6 0.846 13.0 NR 15 [5] diatom P. tricornutum nitrate 24:0 250 2.3 1.42 6.99 100.6 0.620 13.8 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 3.6 1.00 7.24 100.6 0.280 16.2 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 4.2 1.00 7.24 100.6 0.240 17.4 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 4.8 1.00 7.24 100.6 0.210 17.8 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 7.7 0.23 7.94 100.6 0.030 26.3 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 8.3 0.25 7.94 100.6 0.030 26.6 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 8.7 1.04 7.94 100.6 0.120 16.7 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 11.6 0.93 7.53 100.6 0.080 19.7 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 12.8 0.51 7.53 100.6 0.040 23.5 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 16.7 1.00 8.20 100.6 0.060 17.8 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 30.0 0.30 6.87 100.6 0.010 24.2 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 30.0 0.60 6.87 100.6 0.020 22.7 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 50.0 1.00 6.87 100.6 0.020 20.2 CCMP1327 15 [5] diatom P. tricornutum phosphate 24:0 250 70.1 1.02 6.87 100.6 0.015 22.4 CCMP1327 15 [5] diatom P. tricornutum nitrate 24:0 250 96.4 0.93 6.87 100.6 0.010 22.5 NR 15 [5]

Pageintentionall

158 yleft blank

Table C.4 (continued). Literature data used to test model. diatom P. glacialis nitrate 24:0 250 18.7 0.22 2015 3886 0.012 13.0 CCMP980 2 [7] diatom P. glacialis nitrate 24:0 250 23.0 0.09 2015 3886 0.004 15.7 CCMP980 2 [7] diatom P. glacialis nitrate 24:0 250 23.1 0.21 2015 3886 0.009 18.2 CCMP980 2 [7] diatom P. glacialis nitrate 24:0 250 23.4 0.27 2015 3886 0.016 6.0 CCMP980 2 [7] diatom P. glacialis nitrate 24:0 250 25.8 0.32 2015 3886 0.012 9.8 CCMP980 2 [7] diatom P. glacialis nitrate 24:0 250 26.0 0.27 2015 3886 0.010 15.3 CCMP980 2 [7] diatom P. glacialis nitrate 24:0 250 79.9 0.17 2015 3886 0.002 22.2 CCMP980 2 [7] coccolithophore E. huxleyi light 16:8 30 5.3 1.06 6.29 74.0 0.200 7.7 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 30 10.9 1.10 7.48 83.2 0.101 6.7 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 30 16.2 1.20 9.15 95.2 0.074 7.6 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 30 23.8 1.08 5.66 68.9 0.045 6.7 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 30 32.1 1.07 5.90 70.9 0.033 7.7 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 80 5.5 1.59 8.29 89.1 0.289 8.7 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 80 12.5 1.61 11.03 108.0 0.129 9.7 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 80 16.3 1.71 10.74 106.1 0.105 7.9 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 80 24.2 1.55 14.67 130.8 0.064 10.3 PML B92/11 15 [8]

159 coccolithophore E. huxleyi light 16:8 80 32.2 1.52 13.42 123.2 0.047 9.7 PML B92/11 15 [8]

coccolithophore E. huxleyi light 16:8 150 5.6 1.77 8.28 89.0 0.316 9.2 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 150 12.4 1.86 8.48 90.5 0.150 10.8 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 150 18.4 1.78 8.59 91.3 0.097 11.9 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 150 21.3 1.78 9.69 99.0 0.084 12.0 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 150 27.2 1.75 9.07 94.7 0.064 11.9 PML B92/11 15 [8] coccolithophore E. huxleyi light 16:8 150 1.1 0.85 5.30 66.0 0.770 7.0 PML B92/11 16 [9] coccolithophore E. huxleyi light 16:8 150 2.7 1.32 6.00 71.7 0.490 7.5 PML B92/11 16 [9] coccolithophore E. huxleyi light 16:8 150 4.0 1.49 9.20 95.6 0.370 8.1 PML B92/11 16 [9] coccolithophore E. huxleyi light 16:8 150 5.3 1.57 10.20 102.5 0.300 8.5 PML B92/11 16 [9] coccolithophore E. huxleyi light 16:8 150 7.6 1.56 9.00 94.2 0.210 9.8 PML B92/11 16 [9] coccolithophore E. huxleyi light 16:8 150 14.5 1.46 10.40 103.8 0.100 11.9 PML B92/11 16 [9] coccolithophore E. huxleyi light 16:8 150 14.6 1.46 9.40 97.0 0.100 12.4 PML B92/11 16 [9] coccolithophore E. huxleyi light 16:8 150 15.0 1.32 11.40 110.4 0.090 12.3 PML B92/11 16 [9] coccolithophore E. huxleyi light 16:8 150 16.3 1.26 10.60 105.1 0.080 12.3 PML B92/11 16 [9] coccolithophore E. huxleyi light 16:8 150 36.4 1.27 10.00 101.1 0.030 12.7 PML B92/11 16 [9] coccolithophore E. huxleyi light 16:8 150 53.5 1.23 11.70 112.4 0.020 13.8 PML B92/11 16 [9]

Pageintentionall

160

yleft blank

Table C.4 (continued). Literature data used to test model. coccolithophore E. huxleyi light 24:0 15 13.2 0.53 6.63 76.7 0.040 15.9 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 15 19.2 0.53 6.74 77.5 0.028 14.7 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 15 19.9 0.54 6.73 77.5 0.027 14.6 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 15 30.4 0.54 6.60 76.4 0.018 14.9 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 15 33.8 0.54 7.23 81.3 0.016 15.2 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 30 11.9 0.77 7.21 81.1 0.065 16.3 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 30 16.9 0.83 8.16 88.2 0.049 15.8 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 30 18.6 0.81 7.79 85.5 0.044 15.6 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 30 23.2 0.76 9.18 95.4 0.033 16.0 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 30 27.3 0.81 9.91 100.5 0.030 15.3 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 80 11.7 1.02 10.30 103.1 0.087 16.3 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 80 19.3 1.08 10.63 105.3 0.056 16.8 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 80 19.5 1.03 12.65 118.4 0.053 17.8 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 80 27.3 1.12 12.27 116.0 0.041 17.1 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 80 29.2 1.16 14.55 130.1 0.040 16.9 PML B92/11 15 [8]

161 coccolithophore E. huxleyi light 24:0 150 11.7 1.09 10.20 102.5 0.093 16.9 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 150 16.8 1.15 10.99 107.7 0.068 17.5 PML B92/11 15 [8]

coccolithophore E. huxleyi light 24:0 150 17.9 1.10 12.81 119.4 0.061 17.1 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 150 22.6 0.98 12.04 114.5 0.043 17.1 PML B92/11 15 [8] coccolithophore E. huxleyi light 24:0 150 30.0 0.93 14.30 128.6 0.031 17.7 PML B92/11 15 [8] coccolithophore E. huxleyi nitrate 24:0 150 9.6 0.60 8.30 87.6 0.063 17.2 B92/11 18 [10] coccolithophore E. huxleyi nitrate 24:0 150 11.3 0.40 8.30 87.6 0.035 18.5 B92/11 18 [10] coccolithophore E. huxleyi nitrate 24:0 150 12.1 0.20 8.30 87.6 0.017 21.5 B92/11 18 [10] coccolithophore E. huxleyi nitrate 24:0 150 12.2 0.40 8.30 87.6 0.033 19.4 B92/11 18 [10] coccolithophore E. huxleyi nitrate 24:0 150 17.6 0.20 8.30 87.6 0.011 21.4 CCMP3266 18 [11] coccolithophore E. huxleyi nitrate 24:0 150 20.6 0.50 8.30 87.6 0.024 20.9 BT6 18 [10] coccolithophore E. huxleyi nitrate 24:0 150 20.8 0.60 8.30 87.6 0.029 21.3 BT6 18 [10] coccolithophore E. huxleyi nitrate 24:0 150 21.4 0.50 8.30 87.6 0.023 22.2 BT6 18 [10] coccolithophore E. huxleyi nitrate 24:0 150 29.1 0.40 8.30 87.6 0.014 22.7 BT6 18 [10] coccolithophore E. huxleyi nitrate 24:0 150 274.1 0.50 8.30 87.6 0.002 24.9 BT6 18 [10]

Pageintentionall

162 yleft blank

Table C.4 (continued). Literature data used to test model. dinoflagellate A. fundyense light 16:8 250 6.1 0.75 3169 4866 0.126 9.0 Alex5 15 [12] [13] dinoflagellate A. fundyense light 16:8 250 11.8 0.75 3620 5322 0.065 10.2 Alex5 15 [12] [13] dinoflagellate A. fundyense light 16:8 250 26.5 0.78 3455 5157 0.030 12.7 Alex5 15 [12] [13] dinoflagellate A. fundyense light 16:8 250 37.3 0.73 3461 5163 0.020 12.1 Alex5 15 [12] [13] dinoflagellate A. tamarense nitrate 16:8 250 8.4 0.24 3930 5624 0.029 13.2 Alex5 15 [13] dinoflagellate A. tamarense nitrate 16:8 250 30.1 0.24 2709 4379 0.008 13.2 Alex5 15 [13] dinoflagellate A. tamarense nitrate 16:8 250 37.9 0.24 3544 5246 0.007 12.6 Alex5 15 [13]

163 dinoflagellate A. tamarense nitrate 24:0 150 63.0 0.24 2947 4634 0.004 26.7 CCMP1771 18 [14] dinoflagellate A. tamarense nitrate 24:0 150 23.2 0.14 2947 4634 0.006 22.3 CCMP1771 18 [14] dinoflagellate A. tamarense nitrate 24:0 150 22.9 0.24 3336 5037 0.010 21.3 CCMP1771 18 [14] dinoflagellate A. tamarense nitrate 24:0 150 25.0 0.35 3336 5037 0.014 19.4 CCMP1771 18 [14] dinoflagellate A. tamarense nitrate 24:0 150 12.8 0.24 3282 4982 0.019 17.2 CCMP1771 18 [14] dinoflagellate A. tamarense nitrate 24:0 150 10.2 0.24 3697 5398 0.024 14.7 CCMP1771 18 [14] 1Bolded numbers indicate assumed values when measurements were not reported. Italicized numbers indicate parameters calculated from other values reported in the reference. 2NR = not reported

References for Table C.4: [1] Burkhardt S., Riebesell U., and Zondervan I. (1999a) Stable carbon isotope fractionation by marine phytoplankton in response to daylength, growth rate, and CO2 availability. Mar. Ecol. Prog. Ser. 184, 31-41. [2] Burkhardt S., Riebesell U., and Zondervan I. (1999b) Seawater carbonate chemistry, stable carbon isotope fractionation and growth rate during experiments with marine phytoplankton community. PANGAEA, https://doi.pangaea.de/10.1594/PANGAEA.718103.

[3] Burkhardt S., Riebesell U., and Zondervan I. (1999c) Effects of growth rate, CO2 concentration, and cell size on the stable carbon isotope fractionation in marine phytoplankton. Geochim. Cosmochim. Ac. 63, 3729-3741. [4] Riebesell U., Burkhardt S., Dauelsberg A., and Kroon B. (2000a) Carbon isotope fractionation by a marine diatom: Dependence on the growth-rate-limiting resource. Mar. Ecol. Prog. Ser. 193, 295- 303.

[5] Cassar N. (2003) Carbon-concentrating mechanisms and -carboxylation: Their potential contribution to marine photosynthetic carbon isotope fractionation. Doctoral dissertation, University of Hawai’i.

[6] Laws E. A., Bidigare R. R., and Popp B. N. (1997) Effect of growth rate and CO2 concentration on carbon isotope fractionation by the marine diatom Phaeodactylum tricornutum. Limnol. Oceanogr. 42, 1552-1560. [7] Popp B. N., Laws E. A., Bidigare R. R., Dore J. E., Hanson K. L., and Wakeham S. G. (1998) Effect of phytoplankton cell geometry on carbon isotopic fractionation. Geochim. Cosmochim. Ac. 62, 69- 77. [8] Rost B., Zondervan I., and Riebesell U. (2002) Light-dependent carbon isotope fractionation in the coccolithophorid Emiliania huxleyi. Limnol. Oceanogr. 47, 120-128.

[9] Riebesell U., Revill, A. T., Holdsworth D. G., and Volkman J. K. (2000b) The effects of varying CO2 concentration on lipid composition and carbon isotope fractionation in Emiliania huxleyi. Geochim. Cosmochim.. Ac. 64, 4179-4192. [10] Bidigare R. R., Fluegge A., Freeman K. H., Hanson K. L., Hayes J. M., Hollander D., Jasper J. P., King L. L., Laws E. A., Milder J., Miller F. J., Pancost R., Popp B. N., Steinberg P. A., and Wakeham S. G. (1997) Consistent fractionation of 13C in nature and in the laboratory: Growth-rate effects in some haptophyte algae. Global Biogeochem. Cy. 11, 279-292. [11] Wilkes E. B., McClelland H. L. O., Rickaby R. E. M., and Pearson A. (2018) Carbon isotope ratios of coccolith-associated polysaccharides of Emiliania huxleyi as a function of growth rate and CO2 concentration (Organic Geochemistry, in press). [12] Hoins M., Van de Waal D. B., Eberlein T., Reichart G.-J., Rost B., and Sluijs A. (2015) Stable carbon isotope fractionation of organic cyst-forming dinoflagellates: Evaluating the potential for a CO2 proxy. Geochim. Cosmochim. Ac. 160, 267-276.

164

[13] Hoins M., Eberlein T., Groβmann C. H., Brandenburg K., Reichart G.-J., Rost B., Sluijs, A., and Van de Waal, D. B. (2016) Combined effects of ocean acidification and light or nitrogen availabilities on 13C fractionation in marine dinoflagellates. PLoS ONE 11, e0154370.

[14] Wilkes E.B., Carter S.J., and Pearson A. (2017) CO2-dependent carbon isotope fractionation in the dinoflagellate Alexandrium tamarense. Geochimica et Cosmochimica Ac., 212, 48-61. [15] Bartual A., Gálvez J.A., and Ojeda F. (2008) Phenotypic response of the diatom Phaeodactylum tricornutum Bohlin to experimental changes in the inorganic carbon system. Bot. Mar. 51, 350- 359. [16] Montagnes D.J.S., Berges J.A., Harrison P.J., and Taylor F.J.R. (1994) Estimating carbon, nitrogen, protein, and chlorophyll a from volume in marine phytoplankton. Limnol. Oceanogr. 39, 1044- 1060.

165

Table C.5. Input parameters and model outcomes accompanying Figures 4.4-4.7.a Net Gross - - Modeled Measured % Ci HCO3 HCO3 Other Input Fig. Species Limitation L:D PFD [CO2] i POC SA  P  P Efflux Import Import Parameters (%) (%)

4.4 a,b P. tricornutum nitrate 24:0 250 0.6 0.77 4.95 100.6 9.3 10.5 36.5 1.7 1.8  RubisCO 18.5

P. tricornutum nitrate 24:0 250 0.6 1.38 4.95 100.6 6.7 7.4 25.5 1.6 1.7 Km 41 P. tricornutum nitrate 24:0 250 0.9 1.04 5.50 100.6 9.5 11.1 36.9 1.1 1.2 ω 0.99 P. tricornutum nitrate 24:0 250 1.0 0.50 5.82 100.6 13.8 10.7 55.0 1.0 1.1 P 3.0E-09 P. tricornutum nitrate 24:0 250 1.3 0.98 5.82 100.6 11.5 12.7 45.3 0.8 0.8 P. tricornutum nitrate 24:0 250 1.7 0.98 6.99 100.6 12.0 12.8 47.4 0.6 0.6 P. tricornutum nitrate 24:0 250 2.4 0.75 6.99 100.6 15.6 11.8 62.0 0.4 0.5 P. tricornutum nitrate 24:0 250 2.9 0.75 6.21 100.6 17.4 16.8 69.5 0.3 0.4 P. tricornutum nitrate 24:0 250 3.6 0.50 7.24 100.6 19.5 20.6 78.1 0.3 0.3 P. tricornutum nitrate 24:0 250 6.5 0.50 7.94 100.6 21.4 19.9 85.6 0.2 0.2 P. tricornutum nitrate 24:0 250 10.3 1.40 11.00 100.6 17.8 18.4 70.7 0.1 0.1 P. tricornutum nitrate 24:0 250 10.7 1.25 9.15 100.6 19.4 18.9 77.2 0.1 0.1 P. tricornutum nitrate 24:0 250 12.0 1.00 7.53 100.6 21.3 20.6 85.2 0.1 0.1 P. tricornutum nitrate 24:0 250 15.6 0.75 8.20 100.6 22.5 22.0 90.2 0.1 0.1 P. tricornutum nitrate 24:0 250 22.2 0.50 9.57 100.6 23.6 24.4 94.4 0.0 0.0 P. tricornutum nitrate 24:0 250 34.7 0.50 6.87 100.6 24.3 25.7 97.3 0.0 0.0 P. tricornutum nitrate 24:0 250 0.4 1.00 4.95 100.6 5.9 6.7 22.9 2.5 2.8 P. tricornutum phosphate 24:0 250 0.4 1.02 4.95 100.6 5.9 4.8 22.5 2.5 2.8 P. tricornutum nitrate 24:0 250 0.7 0.18 4.95 100.6 18.3 21.4 74.0 1.4 1.6 P. tricornutum nitrate 24:0 250 1.1 0.93 5.82 100.6 10.8 13.0 42.5 0.9 1.0 P. tricornutum nitrate 24:0 250 2.3 1.42 6.99 100.6 11.6 13.8 45.5 0.4 0.5 P. tricornutum nitrate 24:0 250 3.6 1.00 7.24 100.6 16.1 16.2 64.1 0.3 0.3 P. tricornutum nitrate 24:0 250 4.2 1.00 7.24 100.6 17.0 17.4 67.6 0.2 0.3 P. tricornutum nitrate 24:0 250 4.8 1.00 7.24 100.6 17.7 17.8 70.4 0.2 0.2 P. tricornutum nitrate 24:0 250 7.7 0.23 7.94 100.6 23.4 26.3 93.8 0.1 0.1 P. tricornutum nitrate 24:0 250 8.3 0.25 7.94 100.6 23.4 26.6 93.8 0.1 0.1 P. tricornutum nitrate 24:0 250 8.7 1.04 7.94 100.6 19.8 16.7 79.2 0.1 0.1 P. tricornutum nitrate 24:0 250 11.6 0.93 7.53 100.6 21.4 19.7 85.7 0.1 0.1 P. tricornutum nitrate 24:0 250 12.8 0.51 7.53 100.6 23.0 23.5 92.3 0.1 0.1 P. tricornutum nitrate 24:0 250 16.7 1.00 8.20 100.6 22.0 17.8 88.0 0.1 0.1 P. tricornutum nitrate 24:0 250 30.0 0.30 6.87 100.6 24.5 24.2 98.1 0.0 0.0 P. tricornutum nitrate 24:0 250 30.0 0.60 6.87 100.6 24.0 22.7 96.3 0.0 0.0 P. tricornutum nitrate 24:0 250 50.0 1.00 6.87 100.6 24.0 20.2 96.3 0.0 0.0 P. tricornutum phosphate 24:0 250 70.1 1.02 6.87 100.6 24.3 22.4 97.3 0.0 0.0 P. tricornutum nitrate 24:0 250 96.4 0.93 6.87 100.6 24.5 22.5 98.2 0.0 0.0

4.4 c,d P. tricornutum light 24:0 150 1.2 1.33 12.97 140.0 6.1 7.2 48.1 69.0 75.9  RubisCO 18.5

P. tricornutum light 24:0 150 2.1 1.64 9.01 140.0 10.4 10.6 64.1 38.1 41.9 Km 41 P. tricornutum light 24:0 150 3.2 1.63 8.68 140.0 12.8 17.0 73.5 24.9 27.4 ω 0.2 P. tricornutum light 24:0 150 4.4 1.64 8.40 140.0 14.3 12.6 79.2 18.3 20.2 P 7.0E-09 P. tricornutum light 24:0 150 6.7 1.26 12.41 140.0 15.3 14.8 83.6 12.0 13.2 P. tricornutum light 24:0 150 7.2 1.41 9.24 140.0 16.3 14.5 86.9 11.1 12.2 P. tricornutum light 24:0 150 7.6 1.18 15.07 140.0 15.2 14.7 83.6 10.5 11.6 P. tricornutum light 24:0 150 14.5 1.39 9.90 140.0 17.8 15.6 92.6 5.5 6.1 P. tricornutum light 24:0 150 30.2 1.19 15.86 140.0 18.4 16.2 95.0 2.6 2.9 P. tricornutum light 24:0 150 37.7 1.63 10.03 140.0 18.8 16.7 96.5 2.1 2.3

4.4 c,d P. tricornutum light 16:8 150 2.1 2.21 8.68 140.0 8.5 9.4 58.1 38.1 41.9  RubisCO 18.5

P. tricornutum light 16:8 150 3.2 2.23 8.35 140.0 11.0 11.6 67.9 24.9 27.4 Km 41 P. tricornutum light 16:8 150 4.4 2.21 8.07 140.0 12.8 14.0 74.7 18.3 20.2 ω 0.2 P. tricornutum light 16:8 150 7.2 2.21 8.91 140.0 14.6 15.6 81.4 11.1 12.2 P 7.0E-09 P. tricornutum light 16:8 150 14.5 2.30 9.57 140.0 16.5 15.7 88.6 5.5 6.1 P. tricornutum light 16:8 150 37.7 2.27 9.70 140.0 18.4 16.0 95.3 2.1 2.3

166

Table C.5 (Continued). Input parameters and model outcomes accompanying Figures 4.4-4.7.a Net Gross - - Modeled Measured % Ci HCO3 HCO3 Other Input Fig. Species Limitation L:D PFD [CO2] i POC SA  P  P Efflux Import Import Parameters (%) (%)

4.4 c,d P. tricornutum light 24:0 30 22.4 1.06 9.29 140.0 92.0 16.9 15.1 4.0 4.4  RubisCO 18.5

P. tricornutum light 24:0 30 2.3 0.98 8.80 140.0 58.8 8.9 8.7 39.6 43.6 Km 41 P. tricornutum light 24:0 30 2.8 1.02 8.82 140.0 62.3 9.6 9.8 32.3 35.5 ω 0.1 P. tricornutum light 24:0 30 3.9 1.00 9.20 140.0 68.7 11.0 13.5 23.4 25.7 P 3.0E-09 P. tricornutum light 24:0 30 6.9 0.90 8.75 140.0 81.8 14.5 13.8 13.0 14.3 P. tricornutum light 24:0 30 11.6 1.03 9.04 140.0 86.4 15.4 14.3 7.8 8.5

4.4 c,d P. tricornutum light 24:0 15 1.8 0.47 13.76 140.0 60.6 10.6 9.3 50.0 55.0  RubisCO 18.5

P. tricornutum light 24:0 15 3.9 0.50 15.86 140.0 71.7 12.0 11.5 23.4 25.7 Km 41 P. tricornutum light 24:0 15 8.0 0.51 16.91 140.0 82.5 14.5 12.9 11.3 12.4 ω 0.1 P. tricornutum light 24:0 15 14.9 0.49 18.02 140.0 89.5 16.3 14.8 6.0 6.6 P 3.0E-09

4.5 a,b E. huxleyi nitrate 24:0 150 9.6 0.60 8.30 87.6 16.8 17.2 67.0 0.1 0.1  RubisCO 11.1

E. huxleyi nitrate 24:0 150 11.3 0.40 8.30 87.6 19.5 18.5 78.2 0.1 0.1 Km 70 E. huxleyi nitrate 24:0 150 12.1 0.20 8.30 87.6 22.0 21.5 88.5 0.1 0.1 ω 0.99 E. huxleyi nitrate 24:0 150 12.2 0.40 8.30 87.6 19.9 19.4 79.4 0.1 0.1 P 1.0E-09 E. huxleyi nitrate 24:0 150 17.6 0.20 8.30 87.6 22.8 21.4 91.8 0.1 0.1 E. huxleyi nitrate 24:0 150 20.6 0.50 8.30 87.6 20.9 20.9 83.9 0.0 0.1 E. huxleyi nitrate 24:0 150 20.8 0.60 8.30 87.6 20.4 21.3 81.5 0.0 0.1 E. huxleyi nitrate 24:0 150 21.4 0.50 8.30 87.6 21.1 22.2 84.4 0.0 0.1 E. huxleyi nitrate 24:0 150 29.1 0.40 8.30 87.6 22.5 22.7 90.2 0.0 0.0 E. huxleyi nitrate 24:0 150 274.1 0.50 8.30 87.6 24.5 24.9 98.6 0.0 0.0

4.5 c,d E. huxleyi light 24:0 150 11.7 1.09 10.20 102.5 16.5 16.9 88.6 3.4 3.8  RubisCO 11.1

E. huxleyi light 24:0 150 16.8 1.15 10.99 107.7 17.2 17.5 91.2 2.4 2.6 Km 70 E. huxleyi light 24:0 150 17.9 1.10 12.81 119.4 17.3 17.1 91.6 2.2 2.5 ω 0.6 E. huxleyi light 24:0 150 22.6 0.98 12.04 114.5 17.9 17.1 94.1 1.8 1.9 P 6.0E-09 E. huxleyi light 24:0 150 30.0 0.93 14.30 128.6 18.2 17.7 95.4 1.3 1.5

4.5 c,d E. huxleyi light 24:0 80 11.7 1.02 10.30 103.1 15.3 16.3 89.3 4.3 4.7  RubisCO 11.1

E. huxleyi light 24:0 80 19.3 1.08 10.63 105.3 16.2 16.8 92.7 2.6 2.8 Km 70 E. huxleyi light 24:0 80 19.5 1.03 12.65 118.4 16.2 17.8 92.8 2.6 2.8 ω 0.5 E. huxleyi light 24:0 80 27.3 1.12 12.27 116.0 16.6 17.1 94.3 1.8 2.0 P 6.0E-09 E. huxleyi light 24:0 80 29.2 1.16 14.55 130.1 16.6 16.9 94.2 1.7 1.9

4.5 c,d E. huxleyi light 24:0 30 11.9 0.77 7.21 81.1 14.8 16.3 92.6 5.0 5.5  RubisCO 11.1

E. huxleyi light 24:0 30 16.9 0.83 8.16 88.2 15.2 15.8 94.1 3.6 3.9 Km 70 E. huxleyi light 24:0 30 18.6 0.81 7.79 85.5 15.4 15.6 94.8 3.2 3.5 ω 0.4 E. huxleyi light 24:0 30 23.2 0.76 9.18 95.4 15.6 16.0 95.8 2.6 2.8 P 6.0E-09 E. huxleyi light 24:0 30 27.3 0.81 9.91 100.5 15.7 15.3 96.1 2.2 2.4

4.5 c,d E. huxleyi light 24:0 15 13.2 0.53 6.63 76.7 14.3 15.9 95.4 5.3 5.8  RubisCO 11.1

E. huxleyi light 24:0 15 19.2 0.53 6.74 77.5 14.6 14.7 96.8 3.6 4.0 Km 70 E. huxleyi light 24:0 15 19.9 0.54 6.73 77.5 14.6 14.6 96.8 3.5 3.9 ω 0.3 E. huxleyi light 24:0 15 30.4 0.54 6.60 76.4 14.8 14.9 97.9 2.3 2.5 P 6.0E-09 E. huxleyi light 24:0 15 33.8 0.54 7.23 81.3 14.8 15.2 98.1 2.1 2.3

4.5 c,d E. huxleyi light 16:8 150 5.6 1.77 8.28 89.0 8.7 9.2 71.4 13.4 14.7  RubisCO 11.1

E. huxleyi light 16:8 150 12.4 1.86 8.48 90.5 11.2 10.8 83.8 6.0 6.7 Km 70 E. huxleyi light 16:8 150 18.4 1.78 8.59 91.3 12.2 11.9 88.8 4.1 4.5 ω 0.25 E. huxleyi light 16:8 150 21.3 1.78 9.69 99.0 12.5 12.0 89.8 3.5 3.9 P 6.0E-09 E. huxleyi light 16:8 150 27.2 1.75 9.07 94.7 12.9 11.9 92.1 2.8 3.0 E. huxleyi light 16:8 150 1.1 0.85 5.30 66.0 5.4 7.0 57.0 68.2 75.0 E. huxleyi light 16:8 150 2.7 1.32 6.00 71.7 7.2 7.5 64.9 27.8 30.6 E. huxleyi light 16:8 150 4.0 1.49 9.20 95.6 7.8 8.1 67.4 18.8 20.6 E. huxleyi light 16:8 150 5.3 1.57 10.20 102.5 8.7 8.5 71.4 14.2 15.6 E. huxleyi light 16:8 150 7.6 1.56 9.00 94.2 10.2 9.8 78.8 9.9 10.9 E. huxleyi light 16:8 150 14.5 1.46 10.40 103.8 12.0 11.9 87.8 5.2 5.7 E. huxleyi light 16:8 150 14.6 1.46 9.40 97.0 12.1 12.4 88.2 5.1 5.7 E. huxleyi light 16:8 150 15.0 1.32 11.40 110.4 12.3 12.3 88.9 5.0 5.5 E. huxleyi light 16:8 150 16.3 1.26 10.60 105.1 12.6 12.3 90.3 4.6 5.1 E. huxleyi light 16:8 150 36.4 1.27 10.00 101.1 13.6 12.7 95.4 2.1 2.3 E. huxleyi light 16:8 150 53.5 1.23 11.70 112.4 13.9 13.8 96.8 1.4 1.5 167

4.5 c,d E. huxleyi light 16:8 80 5.5 1.59 8.29 89.1 7.4 8.7 71.6 16.4 18.0  RubisCO 11.1

E. huxleyi light 16:8 80 12.5 1.61 11.03 108.0 9.5 9.7 83.5 7.2 7.9 Km 70 E. huxleyi light 16:8 80 16.3 1.71 10.74 106.1 10.0 7.9 86.2 5.5 6.1 ω 0.1 E. huxleyi light 16:8 80 24.2 1.55 14.67 130.8 10.7 10.3 90.2 3.7 4.1 P 5.5E-09 E. huxleyi light 16:8 80 32.2 1.52 13.42 123.2 11.2 9.7 92.8 2.8 3.1

4.5 c,d E. huxleyi light 16:8 30 5.3 1.06 6.29 74.0 4.4 7.7 56.8 17.9 19.7  RubisCO 11.1

E. huxleyi light 16:8 30 10.9 1.10 7.48 83.2 6.6 6.7 70.7 8.7 9.6 Km 70 E. huxleyi light 16:8 30 16.2 1.20 9.15 95.2 7.4 7.6 75.4 5.9 6.5 ω 0.05 E. huxleyi light 16:8 30 23.8 1.08 5.66 68.9 9.1 6.7 85.4 4.0 4.4 P 1.8E-09 E. huxleyi light 16:8 30 32.1 1.07 5.90 70.9 9.7 7.7 88.7 3.0 3.3

4.6 a,b A. tamarense nitrate 24:0 150 63.0 0.24 2947 4634 22.9 26.7 91.6 0.0 0.0  RubisCO 19.5

A. tamarense nitrate 24:0 150 23.2 0.14 2947 4634 21.8 22.3 87.3 0.0 0.0 Km 84.5 A. tamarense nitrate 24:0 150 22.9 0.24 3336 5037 19.9 21.3 79.2 0.0 0.0 ω 0.99 A. tamarense nitrate 24:0 150 25.0 0.35 3336 5037 18.6 19.4 74.0 0.0 0.0 P 2.2E-09 A. tamarense nitrate 24:0 150 12.8 0.24 3282 4982 17.2 17.2 68.1 0.1 0.1 A. tamarense nitrate 24:0 150 10.2 0.24 3697 5398 15.7 14.7 62.1 0.1 0.1 A. tamarense nitrate 16:8 250 8.4 0.24 3930 5624 14.4 13.2 56.9 0.1 0.1 b A. tamarense nitrate 16:8 250 30.1 0.24 2709 4379 21.1 13.2 84.3 0.0 0.0 b A. tamarense nitrate 16:8 250 37.9 0.24 3544 5246 21.5 12.6 86.1 0.0 0.0

4.6 c,d A. fundyense light 16:8 250 6.1 0.75 3169 4866 6.1 9.0 41.1 9.8 10.8  RubisCO 19.5

A. fundyense light 16:8 250 11.8 0.75 3620 5322 9.4 10.2 55.9 5.1 5.6 Km 84.5 A. fundyense light 16:8 250 26.5 0.78 3455 5157 13.8 12.7 73.4 2.3 2.5 ω 0.4 A. fundyense light 16:8 250 37.3 0.73 3461 5163 15.8 12.1 80.5 1.6 1.8 P 4.5E-09

4.7 a,b P. glacialis nitrate 24:0 250 18.7 0.22 2015 3886 12.0 13.0 47.0 0.1 0.1  RubisCO 18.5

P. glacialis nitrate 24:0 250 23.0 0.09 2015 3886 18.3 15.7 72.7 0.0 0.0 Km 41 P. glacialis nitrate 24:0 250 23.1 0.21 2015 3886 13.6 18.2 53.4 0.0 0.0 ω 0.99 P. glacialis nitrate 24:0 250 23.4 0.27 2015 3886 12.1 6.0 47.4 0.0 0.0 P 4.5E-10 P. glacialis nitrate 24:0 250 25.8 0.32 2015 3886 11.7 9.8 45.6 0.0 0.0 P. glacialis nitrate 24:0 250 26.0 0.27 2015 3886 12.8 15.3 50.1 0.0 0.0 P. glacialis nitrate 24:0 250 79.9 0.17 2015 3886 20.8 22.2 83.0 0.0 0.0 a -1 - POC is reported here in units of Pg C cell . All other units are as reported in Table C.3. Ci efflux represents the percentage of inorganic carbon (CO2+HCO3 ) entering the cell that leaks back out again as CO2 or H2CO3. bThese data points are clear outliers when compared with the rest of the nitrate-limited data. Throughout the text and Figure 4.6, we consider scenarios where these data points are and are not included.

168

Appendix D

Supporting Information for Chapter 5

6.0E+06

4.0E+06

2.0E+06

Multisizer Count Multisizer y = 1.25x - 199,062.64 R² = 0.77 0.0E+00 0.0E+00 2.0E+06 4.0E+06 6.0E+06 Microscopy Count

Figure D.1. Multisizer cell count agreement with microscopy count (including data from 3 experiments).

169

Figure D.2. Temporal data showing the approach to steady state conditions and the stability of experimental parameters at steady state for Experiment #1. Plotted error bars represent 95% confidence limits, and the grey rectangle highlights the days considered to be at steady state.

170

Figure. D.3. Temporal data showing the approach to steady state conditions and the stability of the steady state experimental conditions for Experiment #2. Plotted error bars represent 95% confidence limits, and the grey rectangle highlights the days considered to be at steady state.

171

Figure. D.4. Temporal data showing the approach to steady state conditions and the stability of the steady- state experimental conditions for Experiment #3. Plotted error bars represent 95% confidence limits, and the grey rectangle highlights the days considered to be at steady state.

172

Figure D.5. Line P transect, with stations 1, 2, 4, and 8 emphasized by the black rectangle. Figure adapted from Ribalet et al. (2010).

173