SEA CHANGE. PHYTOPLANKTON CHANGE TOO? THERMAL ADAPTATION OF THE MARINE DIATOM THALASSIOSIRA PSEUDONANA IN RESPONSE TO TEMPERATURE: AN EVOLUTIONARY SCENARIO By Daniel Roy O’Donnell A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Integrative Biology—Doctor of Philosophy Ecology, Evolutionary Biology and Behavior—Dual Major 2018 ABSTRACT SEA CHANGE. PHYTOPLANKTON CHANGE TOO? THERMAL ADAPTATION OF THE MARINE DIATOM THALASSIOSIRA PSEUDONANA: AN EVOLUTIONARY SCENARIO Daniel Roy O’Donnell By Rapid evolution in response to environmental change will likely be a driving force determining the distribution of species across the biosphere in coming decades. This is especially true of microorganisms, many of which may evolve in step with warming, including phytoplankton, the diverse photosynthetic microbes forming the foundation of most aquatic food webs. I tested the capacity of a globally important, model marine diatom Thalassiosira pseudonana, for rapid evolution in response to temperature. Selection at 16 and 31°C for 350 generations led to significant divergence in several temperature response traits, demonstrating local adaptation and the existence of tradeoffs associated with adaptation to different temperatures. In contrast, competitive ability for nitrogen (commonly limiting in marine systems), measured after 450 generations of temperature selection, did not diverge in a systematic way between temperatures. After 500 generations of temperature selection, T. pseudonana populations had diverged in their thermal reaction norms for fatty acid composition and in their carbon, nitrogen, phosphorus and chlorophyll a contents. Divergence in C:N:P stoichiometry was not apparent. 31°C-selected populations showed morphological evidence of selection for more efficient nutrient uptake at high temperatures. This study shows how rapid thermal adaptation affects important growth, nutrient uptake and utilization, and cellular physiology and morphology traits in a key marine phytoplankton species, and may thus play a role in long-term physiological, ecological and biogeographic responses to climate change. This dissertation is dedicated to Uncle Joe and Aunt Martha. As a young boy, I once pointed to the inside of a trout I was gutting and asked my dad, “Hey, what’s that thing?” “Beats me”, he said. “Ask your Uncle Joe and Aunt Martha. They’re scientists.” iii The rest is history. ACKNOWLEDGEMENTS Dr. Elena Litchman has been the perfect boss, adviser, teacher and friend. I have always come out of meetings with Elena feeling better than when I went in. Dr. Chris Klausmeier is among a rarified subset of my elders who have permanently rewired my brain. He is a uniquely gifted thinker and educator. Drs. Jen Lau and Gary Mittelbach exemplify the uncommon character of academics who take pride in caring about and nurturing those who look up to them. In the unlikely event I someday achieve at their level, I hope I can be as nice as they are. I hope that someday I can be as good at my job as Pam Woodruff and Ally Hutchens are at theirs. I wanted for nothing in the Litchman lab. Stacey Vander Wulp, Linda Danhof, and Dr. Zhi-yan “Rock” Du also went out of their way to help me, even though I’m not even in their labs. Dr. Kay Gross has had my back in more ways than even I will ever know. Andy Fogiel, Tyler Brownell, Jennifer Smith, Jennae Roseboom, and countless other KBS staff worked tirelessly to pick up after me and make sure I had everything I needed, even when I made that task as burdensome as humanly possible for them. Dr. Colin Kremer taught me more about statistics on a given day than I learned in all of my stats courses combined. My proudest and happiest moments in research were all thanks to my mentees, Carolyn Hamman, Evan Johnson, Connie High, Farhana Haq, Claire Harper, Olivia Porth, Sophie Beery and Ayley Shortridge. I hope I have left as lasting a mark on their lives as they have left on mine. No person survives a PhD without the support of their friends and loved ones. My wife, Abhijna, has stood by me for worse, for poorer, and in sickness. When I suffered, she suffered with me. When I lost my mind, she helped me find it again. There are many others to thank. They know who they are. Except Felix. He probably doesn’t get it. iv TABLE OF CONTENTS LIST OF TABLES………………………………………………...……………………………..vii LIST OF FIGURES…………………………………………………………………...………….ix KEY TO SYMBOLS……….…………………………………………………………………….xi INTRODUCTION………………………………………………………………………………...1 REFERENCES……………………………………………………………………………………3 CHAPTER 1: Rapid thermal adaptation in a marine diatom reveals constraints and tradeoffs…..6 Introduction……...…………………………………………………………………....…...6 Materials and Methods…………………………………………………………………...11 Temperature selection experiment ……………………………………………….11 Temperature-dependent growth assays…………………………………………..13 Nitrate-dependent growth assays…………………………………………………14 Model fitting, trait estimation and statistical analysis…………………………….16 Results…………………………………………………..………………………………..17 Evolutionary change in thermal reaction norms………………………………….17 Evolution of competitive ability for nitrate……………………………………….22 Discussion…………………………………………………………………..…………....24 APPENDICES ...……………………………………………………………………....................30 APPENDIX A: Chapter 1 supplemental figures, tables, and equation.….………………31 APPENDIX B: Chapter 1 supplemental derivations...…………………………………...42 REFERENCES …………………………………………………………………………..............46 CHAPTER 2: Experimental evolution of fatty acid thermal reaction norms in a marine diatom..52 Introduction……...……………………………………………………………………….52 Materials and Methods………………………………….……………………………..…55 500 generation temperature selection experiment………………………………..55 Cryopreservation and revival…………………………………………………….55 Algal growth and harvesting……………………………………………………..56 Fatty Acid Methyl Ester reaction and Gas Chromatrography……………………57 Calculations………………………………………………………………………58 Linear mixed-effects models……………………………………………………..58 Results…..………………………………………………………………………………..59 Discussion…..……………………………………………………………………………69 REFERENCES …………………………………………...…………………………………...…74 CHAPTER 3: Evolution of C:N:P elemental stoichiometry, cell size and cell morphology in a marine diatom in response to temperature………………………………………………………..81 Introduction……......…………………………………………...………………………...81 v Materials and Methods……………………………………………….………………......85 Temperature assay ……………………………………………………………….85 Particulate chemical analysis …………………………………………………….86 Thalassiosira pseudonana morphometry………………………………………...87 Calculations and statistical analysis ……………………………………………...88 Results…..………………………………………………………………………………..90 Particulate C, N, P and Chlorophyll a …………………………………………...90 Stoichiometric ratios ……………………………………………………………..95 Photosynthetic efficiency………………………………………………………...96 Thalassiosira pseudonana frustule size and morphology………………………..98 Discussion…..…………………………………………………………………………..105 APPENDICES ………………………………………………………………………………….110 APPENDIX C: Chapters 2 & 3 description of linear mixed-effects models….………....111 APPENDIX D: Chapter 3 Tukey HSD contrast tables………………………………......114 REFERENCES …………………………………………………………………..……………..132 vi LIST OF TABLES Table A1: ANOVA table for traits described in Figure 2………………………………………..34 Table A2: ANOVA tables comparing traits described in Figure 3………………………………36 Table A3: ANOVA comparing maximum growth rates (µmax) and nitrate affinities (aNO3)…….37 Table 1: RM-ANOVA testing effects of selection temperature and assay temperature on T. pseudonana mean chain length (MCL), unsaturation (WUnSat), total fatty acids (Total FA), percent saturated FA (% SFA), percent monounsaturated FA (% MUFA), percent polyunsaturated FA (% PUFA), and percentages of individual FA classes……………………..62 Table 2: RM-ANOVA testing effects of selection temperature and assay temperature on T. pseudonana particulate C, N, P and chlorophyll a per-cell (µmol µm-3) and all molar ratios thereof, and particulate chlorophyll a (µg µm-3)………………………………………………...92 Table 3: RM-ANOVA testing effects of selection temperature and assay temperature on T. pseudonana particulate C, N, P per biovolume (µmol µm-3) and all molar ratios thereof, and particulate chlorophyll a (µg µm-3)………………………………………………………………95 Table 4: RM-ANOVA testing effects of selection temperature and assay temperature on T. pseudonana carbon assimilation number (C/Chl a) (µg C µg Chl a-1) and assimilation rate (µg C µg Chl a-1 d-1)…………………………………………………………………………………….98 Table 5: Nonlinear regression fits of Equation (6) to carbon assimilation rate per unit Chl a (DC/Chl a) (µg C µg Chl a-1 d-1) as a function of specific growth rate (d-1)…………………….98 Table 6: RM-ANOVA testing effects of selection temperature and assay temperature on T. pseudonana frustule diameter, frustule length, frustule volume, valve surface area-to-frustule volume ratio, girdle surface area-to-frustule volume ratio, and the ratio of valve surface area to girdle surface area………………………………………………………………………………102 Table 7: RM-ANOVA testing effects of selection temperature, assay temperature and frustule volume on ratios of T. pseudonana total frustule surface area, valve surface area, girdle surface area to frustule volume, and the ratio of valve surface area to girdle surface area……………..104 Table D1: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate C per cell (µmol cell-1)………………………………………………...115 Table D2: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate N per cell (µmol cell-1)………………………………………………..116 vii Table D3: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate P per cell (µmol cell-1)………………………………………………...117 Table D4: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate chlorophyll a per cell (µg cell-1).……………………………………...118 Table D5: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate C per biovolume (µmol µm-3)…………………………………………119 Table D6: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate N per biovolume (µmol µm-3)………………………………………...120 Table D7: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate P per biovolume (µmol µm-3)…………………………………………121 Table D8: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate chlorophyll a per biovolume (µg µm-3)……………………………….122 Table D9: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana carbon assimilation number (µg C µg Chl a-1)……………………………………123 Table D10: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana carbon assimilation rate (µg C µg Chl a-1 d-1)…………………………………….124 Table D11: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana frustule diameter (µm)…………………………………………………………….125 Table D12: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana frustule length (µm)……………………………………………………………….126 Table D13: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana frustule volume (µm3)……………………………………………………………..127 Table D14: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana frustule total surface area (µm2) to volume (µm3) ratio…………………………..128 Table D15: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana valve only surface area (µm2) to total frustule volume (µm3) ratio……………….129 Table D16: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana girdle only surface area (µm2) to total frustule volume (µm3) ratio………………130 Table D17: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana girdle surface area (µm2) to valve surface area ratio……………………………...131 viii LIST OF FIGURES Figure 1: Thermal reaction norms for per-capita population growth of 10 T. pseudonana populations……………………………………………………………………………………….18 Figure 2: Thermal performance traits of 16°C- and 31°C-selected populations. I………………19 Figure 3: Thermal performance traits of 16°C- and 31°C-selected populations. II……………...21 Figure 4: Deming regression plots for nitrate growth affinity (aNO3) versus maximum growth rate (µmax)……………………………………………………………………………………………..23 Figure 5: The modified double-exponential model with allocation to thermal protection (p) ranging from 0.1 to 1…………………………………………………………………………….27 Figure A1: Residual plots for fits of the double-exponential model to data from temperature- dependent growth assays of all replicate T. pseudonana populations at 16 and 31°C…………..32 Figure A2: Regression coefficients from Model 2, describing µmax at the selection temperatures in a reciprocal transplant (n = 3)…………………………………………………………………35 Figure A3: Comparison of Fisher’s Information 95% confidence limits to bootstrap confidence limits for Topt parameter in double-exponential model fits to temperature-dependent growth data……………………………………………………………………………………………….36 Figure A4: Nitrate-dependent growth (Monod) assays at 16°C (top, blue boxes) and 31°C……38 Figure A5: Nitrate-dependent growth (Monod) kinetic parameters at ~450 generations………..39 Figure A6: Nonparametric plot of thermal reaction norms (n = 14 for each)……………………39 Figure A7: Relationships between thermal performance traits and the protection parameter (p) in Equations (3)……………………………………………………………………………………..40 Figure 6: Total content of 16 fatty acid classes in T. pseudonana selected at 16 °C (blue shades) and 31°C (red shades)……………………………………………………………………………60 Figure 7: (a) Total fatty acid (FA) content per biovolume in 16°C-selected (blue) and 31°C- selected (red) populations at 10, 16, 26, and 31°C; (b) Saturated FA as a fraction of total FA; (c) Monounsaturated FA as a fraction of total FA; (d) Polyunsaturated FA as a fraction of total FA………………………………………………………………………………………………..64 Figure 8: (a) Mean chain length (C atoms per molecule); (b) degree of unsaturation (double bonds per molecule)……………………………………………………………………………...65 ix Figure 9: Saturated fatty acids as percentages of total fatty acids in 16°C-selected (blue) versus 31°C-selected (red) T. pseudonana populations at 10, 16, 26, and 31°C………………………..66 Figure 10: Monounsaturated fatty acids as percentages of total fatty acids in 16°C-selected (blue) versus 31°C-selected (red) T. pseudonana populations at 10, 16, 26, and 31°C………………...67 Figure 11: Polyunsaturated fatty acids as percentages of total fatty acids in 16°C-selected (blue) versus 31°C-selected (red) T. pseudonana populations at 10, 16, 26, and 31°C………………...68 Figure 12: Particulate C, N, P (molar) and chlorophyll a (µg) per cell in 16°C-selected (blue) and 31°C-selected (red) T. pseudonana, assayed at 10, 16, 26 and 31°C……………………………91 Figure 13 Particulate C, N, P (molar) and chlorophyll a (µg) per biovolume in 16°C-selected (blue) and 31°C-selected (red) T. pseudonana, assayed at 10, 16, 26 and 31°C………………..94 Figure 14: Particulate molar C:N:P ratios in 16°C-selected (blue) and 31°C-selected (red) T. pseudonana, assayed at 10, 16, 26 and 31°C……………………………………………………95 Figure 15: (a) Carbon assimilation number, C/Chl a (µg C µg Chl a-1), of 16°C- (blue) and 31°C- selected (red) T. pseudonana populations grown at 10, 16, 26 and 31°C……………………….97 Figure 16: 16°C-selected (blue boxes) and 31°C-selected (red boxes) Thalassiosira pseudonana frustules photographed in lateral view under 1000x oil immersion……………………………..99 Figure 17: Frustule dimensions in 16°C (blue) and 31°C (red) selected T. pseudonana populations grown at 10, 16, 26 and 31°C……………………………………………………...100 Figure 18: T. pseudonana whole frustule (a), valve only (b), and girdle only (c) surface-area-to- volume ratios (SA:V), and (d) valve SA:girdle SA…………………………………………….101 Figure 19: Slopes of (a) total surface area-to-volume ratio (SAtot:V) of the frustule; (b) SAvalve:V; and (c) SAgirdle:V as functions of frustule volume at the 10, 16, 26 and 31°C assay temperatures…………………………………………………………………………………….103 x KEY TO SYMBOLS Thermal optimum for per-capita population growth Upper critical temperature for per-capita population growth Lower critical temperature for per-capita population growth Maximum per-capita growth rate Per-capita population growth rate at Topt Carbon Nitrogen Phosphorus Chlorophyll a Nitrate Nitrate growth affinity Confidence Interval degrees of freedom Mean squares Residual squared errors Sample size Standard error Fatty acid Saturated fatty acid Unsaturated fatty acid Topt CTmax CTmin µmax µopt C N P Chl a NO3 aNO3 CI df MS RSE n SE FA SFA UFA MUFA Monounsaturated fatty acid xi PUFA DMSO bv MCL Polyunsaturated fatty acid Dimethyl sulfoxide biovolume Mean chain length (C atoms per FA molecule) WUnSat Degree of unsaturation (double bonds per FA molecule) dAIC LRT Difference in Akaike’s Information Criterion Likelihood ratio test Tukey HSD Tukey’s Honestly Significant Difference SAtot SAvalve SAgirdle Total surface area Valve surface area Girdle surface area V Volume xii INTRODUCTION Temperature is a universal control on biological rates, including the physiological processes involved in the acquisition of resources and their conversion into biomass, and ultimately the vital rates of individuals and populations. Given the omnipresent control temperature exerts on biological systems, the current, unprecedented rate of change in global temperature regimes inevitably affects the biosphere at every level of organization, from ecosystems down to genes. The ecological effects of rising global temperatures on ecosystem function, community structure and composition, and population dynamics and range limits are already apparent (reviewed in Walther et al. 2002), and are likely to continue into the foreseeable future (Cramer et al. 2001; Sarmiento et al. 2004; Colwell et al. 2008; Thomas et al. 2012). However, the abilities of populations to adapt evolutionarily to a rapidly changing climate are only beginning to be understood (Parmesan 2006; Skelly et al. 2007; Listmann et al. 2016). The aim of this dissertation is to determine whether and how an ecologically indispensable group of organisms, the phytoplankton, may respond evolutionarily to global climate change. I examined this question through the lens of experimental evolution to determine the capacity of a common, ubiquitous marine diatom to undergo rapid evolution—that is, to evolve at a rate greater than or comparable to the current rate of change of global sea surface temperatures. Understanding the evolutionary capacity of phytoplankton is crucial to predicting the effects of global climate change on marine ecology, global biogeochemistry, and even human health (Kang 2011) and food security (Litzow et al. 2006; Barange et al. 2014). Phytoplankton are responsible for ~50% of global net primary production, and are fundamental drivers of global carbon, nitrogen, phosphorus, and other biogeochemical cycles (Redfield 1958; Field et al. 1998; 1 Falkowski et al. 1998). They are the foundation of most marine and freshwater food webs, and may be the single most important source of omega-3 fatty acids consumed by humans (Arts et al. 2001; Kang 2011). They are also indicator species, responding quickly and dramatically to environmental change (Nehring 1998; Hampton et al. 2008; Boyce et al. 2010). Given their vast population sizes and rapid rates of reproduction, a comprehensive understanding of the responses of phytoplankton to global climate change requires an integrative approach in which their physiology, their ecology, and their evolution are studied in concert. I subjected the ubiquitous model marine diatom Thalassiosira pseudonana to temperatures at the low and high extremes of its thermal tolerance range for 500 generations and quantified evolutionary divergence in a number of key physiological traits. I first examined the effects of directional temperature selection on T. pseudonana’s thermal reaction norm for per- capita population growth rate. I then tested the effects of temperature selection on the thermal reaction norms of a set of organismal traits—fatty acid content and composition, carbon- nitrogen-phosphorus stoichiometry, chlorophyll a content, photosynthetic carbon assimilation efficiency, and cell morphology—that are known to depend on temperature (Eppley 1972; James et al. 1989; Yvon-Durocher et al. 2017), and that are of great ecological importance. This dissertation contributes to a nascent but growing body of knowledge about rapid thermal adaptation in phytoplankton and about thermal adaptation in general. It expands on the known modes of variation of the thermal reaction norm (Izem & Kingsolver 2005), and offers a glimpse of some mechanisms that may underlie adaptation to temperature. Importantly, this work also illuminates a number of unanswered questions about the genetic basis of thermal adaptation, the relevance of the results to other species, and the ecological outlook for species that are able to evolve in step with rising global temperatures and for the communities they inhabit. 2 REFERENCES 3 REFERENCES Arts MT, Ackman RG, Holub BJ. 2001. “Essential fatty acids” in aquatic ecosystems: a crucial link between diet and human health and evolution. Canadian Journal of Fisheries and Aquatic Sciences 58:122-137. Barange M, Marino G, Blanchard JL, Scholtens J, Harle J, et al. 2014. Impacts of climate change on marine ecosystem production in societies dependent on fisheries. Nature Climate Change 4:211-216. Boyce DG, Lewis MR, Worm B. 2010. Global phytoplankton decline over the past century. Nature 466:591-596. Colwell RK, Brehm G, Cardelús CL, Gilman AC, Longino JT. 2008. Global warming, elevational range shifts, and lowland biotic attrition in the wet tropics. Science 322:258-261. Cramer W, Bondeau A, Woodward FI, Prentice IC, Betts RA, et al. 2001. Global response of terrestrial ecosystem structure and function to CO2 and climate change: results from six dynamic global vegetation models. Global Change Biology 7:357-373. Eppley RW. 1972. Temperature and phytoplankton growth in the sea. Fisheries Bulletin 70:1063-1085. Falkowski PG, Barber RT, Smetacek V. 1998. Biogeochemical controls and feedbacks on ocean primary production. Science 281:200-201. Field CB, Behrenfeld MJ, Randerson JT, Falkowski PG. Primary production of the biosphere: integrating terrestrial and oceanic components. Science 281:237-240. Hampton SE, Izmest’eva LR, Moore MV, Katz SL, Dennis B, et al. 2008. Sixty years of environmental change in the world’s largest freshwater lake – Lake Baikal, Siberia. Global Change Biology 14:1947-1958. Izem R, Kingsolver JG. 2005. Variation in continuous reaction norms: quantifying directions of biological interest. The American Naturalist 166:277-289. James CM, Al-Hinty S, Salman AE. 1989. Growth and w3 fatty acid and amino acid composition of microalgae under difference temperature regimes. Aquaculture 77:337-351. Kang JX. Omega-3: a link between global change and human health. Biotechnology Advances 29:388-390. 4 Listmann L, LeRoch M, Schlüter L, Thomas MK, Reusch TBH. 2016. Swift thermal reaction norm evolution in a key marine phytoplankton species. Evolutionary Applications 9:1156- 1164. Litzow MA, Bailey KM, Prahl FG, Heintz R. 2006. Climate regime shifts and reorganization of fish communities: the essential fatty acid limitation hypothesis. Marine Ecology Progress Series 315:1-11. Nehring S. 1998. Establishment of thermophilic phytoplankton species in the North Sea: biological indicators of climatic changes? ICES Journal of Marine Science 55:818-823. Parmesan C. 2006. Ecological and evolutionary responses to recent climate change. Annual Review of Ecology, Evolution and Systematics 37:637-669. Redfield AC. 1958. The biological control of chemical factors in the environment. American Scientist 46:205-221. Sarmiento JL, Slater R, Barber L, Bopp L, Doney SC, et al. 2004. Response of ocean ecosystems to climate warming. Global Biogeochemical Cycles 18:GB3003. Skelly DK, Liana NJ, Possingham HP, Freidburg LK, Farrugia TJ, et al. 2007. Evolutionary responses to climate change. Conservation Biology 21:1353-1355. Thomas MK, Kremer CT, Klausmeier CA, Litchman E. 2012. A global pattern of thermal adaptation in marine phytoplankton. Science 338:1085-1088. Walther G-R, Post E, Convey P, Menzel A, Parmesan C, et al. 2002. Ecological responses to recent climate change. Nature 416:389-395. Yvon-Durocher G, Dossena M, Trimmer M, Woodward G, Allen AP. 2015. Temperature and biogeography of algal stoichiometry. Global Ecology and Biogeography 24:562-570. 5 CHAPTER 1: Rapid thermal adaptation in a marine diatom reveals constraints and tradeoffs Introduction The dependence of physiological processes on temperature is perhaps the most important factor determining the fitness of organisms across latitudinal and altitudinal gradients, and thus their distributions and abundances on Earth. Due to climate warming, many studies have focused on the temperature dependence of physiological traits (e.g., photosynthesis, respiration, per- capita population growth) near the upper bounds of organisms’ thermal tolerance ranges (Rowan, 2004; Bradford, 2013; Listmann et al., 2016) and the ecological consequences of temperatures rising beyond those ranges (Rowan, 2004; Bradford, 2013; Listmann et al., 2016; Thomas et al., 2012). Studies have increasingly focused on the potential for rapid evolution of traits in response to temperature change, and how such evolution may mitigate the ecological impacts of climate warming (Hoffmann & Sgrò, 2011; Schlüter et al., 2014; Listmann et al., 2016). Evolution on ecological timescales may be commonplace (Schoener, 2011) and can ‘rescue’ populations from potentially catastrophic environmental change (Gomulkiewicz & Holt, 1995; Bell, 2013; Schiffers et al., 2013). In particular, the enormous population sizes and fast reproduction of microorganisms, on the order of hours to days, offer many opportunities for rapid evolution in response to environmental change. Phytoplankton are a diverse and often overlooked microbial group of global importance: they are the foundation of most marine and freshwater food webs and are major drivers of global biogeochemical cycles, carrying out ~50 % of global carbon fixation (Field et al., 1998) and linking nitrogen and phosphorus cycles (Redfield, 1958). Rising temperatures may negatively impact phytoplankton productivity, biomass and species diversity (Boyce et al., 2010; Thomas et al., 2012). However, rapid 6 adaptation to warming may mitigate some ecological impacts of climate change (Litchman et al., 2012; Listmann et al., 2016). For example, evolutionary change in the thermal optimum for per- capita population growth (Topt) and the maximum temperature at which positive growth is possible (CTmax) in response to ocean warming may reduce heat-induced mortality, allowing some species to persist at low latitudes where they might otherwise go regionally extinct (Thomas et al., 2012). A number of recent evolution experiments have investigated how marine phytoplankton may evolve to cope with environmental change (Reusch & Boyd, 2013), though most have focused only on increased atmospheric pCO2 and/or ocean acidification (Crawfurd et al., 2011; Lohbeck et al., 2012; Jin et al., 2013). Thermal adaptation has been studied experimentally in only a couple of phytoplankton species, only one of which, a coccolithophorid, was a marine species of global importance (Schlüter et al., 2014; Padfield et al., 2015; Listmann et al., 2016). It is therefore difficult to predict if other ecologically important groups, such as diatoms that contribute up to 25% of all global carbon fixation (Nelson et al., 1995), would respond similarly. Moreover, while we may expect Topt and CTmax to increase after evolving at high temperatures (Listmann et al., 2016; Schaum et al., 2017), knowledge of how other important traits may change in response to elevated temperatures is limited, though some significant work has been done on rapid temperature-dependent evolution of metabolic traits in freshwater chlorophytes (Padfield et al., 2015; Schaum et al., 2017). Schlüter et al., (2014) found that after one year of experimental adaptation to elevated temperature (26.3 °C), the marine coccolithophore Emiliania huxleyi evolved a higher population growth rate, smaller cell diameter and lower particulate organic and inorganic carbon content compared to cold (15 °C)-adapted populations. Hinners et al. (2017) found that 100-year-old resurrected strains of the marine dinoflagellate Apocalathium 7 malmogiense did not differ in cell size or in its thermal reaction norm for growth compared to current strains, but that it did form resting cysts at a higher rate. Large-scale evolutionary change in nutrient utilization traits in response to changing climate could significantly impact biogeochemical cycling (Field et al., 1998). A number of studies have examined the interactive effects of temperature and nutrient on per-capita population growth rates in phytoplankton (Tilman et al., 1981; Thomas et al., 2017), and some have tested the effects of temperature on elemental (C, N, P) stoichiometry. Martiny et al., (2016) studied temperature effects on growth and elemental stoichiometry in several Prochlorococcus species, but did not find consistent trends across closely related species, suggesting that effects of thermal adaptation on nutrient utilization traits are idiosyncratic and not necessarily consistent; Thrane et al., (2017) conducted a large, factorial experiment to elucidate the effect of temperature on optimal N:P supply ratio in the freshwater chlorophyte Chlamydomonas reinhardtii; and Yvon-Durocher et al., (2017) studied temperature effects on seston stoichiometry in a mesocosm community, emphasizing long-term evolutionary effects on cellular stoichiometry, also in C. reinhardtii. However, this study is novel in that it directly tests whether a key measure of competitive ability for nitrate, nitrate growth affinity, evolves in distinct temperature environments in the model marine diatom Thalassiosira pseudonana. Biological rates in ectotherms vary continuously across temperature and are, thus, often referred to as “function-valued traits” (Kingsolver et al., 2001). Our understanding of how diverse function-valued traits may evolve in different organisms is limited. Tradeoffs (e.g., generalist-specialist or resource-allocation tradeoffs) may be important in determining how the shapes of trait functions, including the thermal reaction norm, evolve (Angilletta et al., 2003). The trait values derived from thermal reaction norms and many other function-valued traits are 8 non-independent across temperature gradients (Angilletta et al., 2003); thus, selection at one temperature may alter rates (e.g., population growth rate) at every other temperature along the thermal reaction norm. The underlying mechanisms of thermal adaptation and its associated tradeoffs may lie in any of myriad of metabolic and physiological pathways within the organism. For example, allocation of resources to constitutive adaptations to cold environments may preclude those resources from being used to synthesize heat shock proteins when the environment unexpectedly warms; nitrate uptake machinery optimized for warm environments may not function well in cold environments; membrane lipids may be comprised of many highly unsaturated fatty acids in a cold-adapted organism, but such fatty acids lead to excessive membrane fluidity in a warm environment (Jiang & Gao, 2004); or, in any of these cases, plasticity in traits (i.e., in a generalist) may lower the maximum achievable functionality of each trait at every temperature (a generalist-specialist tradeoff). While recent studies showed that Topt and CTmax increase after evolving at higher temperatures (Listmann et al., 2016), we do not know how the whole thermal reaction norm may evolve in phytoplankton; for example, adaptation to high temperatures may lead to a decrease in fitness at low temperatures, as previously observed in bacteria (Bennett & Lenski, 1993) and bacteriophages (Knies et al., 2006). Selection may change the slope and curvature (first and second derivatives) of the thermal reaction norm as well (Kutcherov, 2016), especially if some cardinal temperatures are more evolutionarily labile than others (Araújo et al., 2013). For example, by comparing the peak sharpness at Topt and the width (°C) of the upper tail of the thermal reaction norm in cold- versus warm-selected populations, we may make inferences regarding the evolutionary lability of the maximum growth rate at Topt (µopt), and of Topt itself, relative to that of CTmax; the upward concavity below Topt and the location of the inflection point 9 on the thermal reaction norm contribute greatly to the magnitude of tradeoffs between growth at temperatures below Topt versus above, which can be indicative of generalist-specialist and/or resource allocation tradeoffs (Gilchrist, 1995; Angilletta et al., 2003). The thermal reaction norm for population growth depends on the temperature dependences of enzyme activities (Ratkowsky et al., 2005; Corkrey et al., 2014), and few processes or pathways within the cell should be immune to the effects of directional temperature selection (Nedwell, 1999; Padfield et al., 2015; Baker et al., 2016; Schaum et al., 2017). Depending on the genes affected, changes in growth rate may be accompanied by changes in traits not directly associated with the thermal reaction norm. For example, stoichiometry may respond to thermal adaptation due to changes in relative resource requirements of phytoplankton cells, which in turn may affect species’ competitive abilities (e.g. for N and P) and ultimately global biogeochemical cycles (Redfield, 1958; Litchman et al., 2007; Baker et al., 2016). An increase in maximum growth rate at the selection temperature may lead to a decrease in affinity for a given nutrient due to a tradeoff between allocation of cellular resources to reproduction versus nutrient uptake (Grover, 1991; Klausmeier et al., 2004). If the nutrient in question is never limiting, selection may favor a high maximum growth rate over affinity, and the tradeoff may ultimately result in a weakening of competitive ability for that nutrient in the selection environment. Moreover, adaptation to high temperature may require higher investment in repair machinery, such as heat shock proteins, potentially increasing the demand for nitrogen and other nutrients. Here, we performed a 350-generation selection experiment on a model marine diatom, Thalassiosira pseudonana, to determine whether and how its thermal performance curves evolve and diverge under selection at two different temperatures, above and below the growth optimum, 10 and whether other key traits, such as nitrate utilization traits, evolve as well. Post-selection, we conducted temperature and nutrient-dependent growth assays to assess evolutionary divergence in temperature and nutrient traits resulting from thermal adaptation. Materials and Methods Temperature selection experiment We obtained a monoculture of Thalassiosira pseudonana CCMP1335 (Hustedt) Hasle et Heimdal from the Provasoli-Guillard National Center for Culture of Marine Phytoplankton (CCMP), Maine, USA, and isolated a single cell by plating on agarose ESAW marine culture medium (Harrison et al., 1980). Nitrogen (as NaNO3), phosphorus (as NaH2PO4·H2O) and silicon (as NaSiO3·9H2O) concentrations were 549 µmol l-1, 22.4 µmol l-1, and 106 µmol l-1, respectively. We cultured the resulting colony in a flask of ESAW medium until it reached a high density. “Generation zero” was on April 14, 2014, when we propagated ten replicate populations from the progenitor culture (“selection lines” hereafter) into 20 ml ESAW medium in 50 ml Cellstar polystyrene tissue culture flasks with breathable caps (Greiner Bio-One GmbH, Frickenhausen, Germany), which we gently agitated daily to keep cells in suspension. We assigned five replicates to the 16°C treatment and the other five to 31°C. These two temperatures were below and above the previously-recorded thermal optimum (Topt) of T. pseudonana (between 24 and 26°C) (Boyd et al., 2013) under similar culture conditions. We chose these selection temperatures so that the maximum growth rates (µmax) at the two temperatures were roughly equal (~0.8 d-1 at the start of the experiment). The low temperature treatment was close to the ancestral maintenance temperature of 14°C (see technical specifications on CCMP1335 culture maintenance at https://ncma.bigelow.org/catalogsearch/result/?q=1335) and 16°C was 11 thus used for the main comparison of the evolution in response to different temperatures. Such a comparison allows isolation of the selection effect of temperature itself from the effect of the cultivation regime, as both temperature treatments experience the same maintenance for the duration of the experiment. In contrast, the common comparison in many experimental evolution studies of the treatment and the ancestral strain that did not experience the same selection regime may confound the effect of the focal selection force and the culturing conditions. Selection lines were maintained in temperature-controlled growth chambers at 110 µmol photons m-2 s-1 under a 14:10 light:dark cycle (note that this is 10 µmol photons m-2 s-1 higher than the light level in Boyd et al., (2013), but still in the saturating range). All lines were maintained in ESAW medium for ~50 generations at a dilution rate of 0.5 d-1 (diluted daily), after which (August 12, 2014) all cultures were transferred to L1 medium (Guillard & Hargraves, 1993) due to poor culture health in ESAW. L1 was made using 43.465 g l-1 artificial sea salt (Tropic Marin, Wartenberg, Germany), which yields a specific gravity of 1.025, similar to natural seawater, with NaNO3, NaH2PO4·H2O, and NaSiO3·9H2O concentrations of 882 µmol l-1, 36.3 µmol l-1, and 107 µmol l-1, respectively. Upon switching culture media, we also altered the dilution regime, such that 106 cells (usually ~0.5-1.0 ml) were transferred to fresh medium every 4 d (with occasional deviations of ±1 d). The selection lines required an additional ~50 generations to adjust to the new dilution regime before consistent growth rates and culture health were achieved. These changes led to a substantial reduction in the variability of growth rates over time. Cell densities of all replicate cultures were determined at each transfer for the remainder of the experiment using a CASY particle counter (Schärfe System GmbH, Reutlingen, Germany), and average population growth rates for each propagation period determined by taking log(Nt/N0)/t, where N is the population density and t is time in days. 12 Temperature-dependent growth assays To evaluate evolutionary change in thermal reaction norms after ~350 generations, we conducted temperature-dependent population growth assays on evolved selection lines. First, we sub-cultured all selection lines and acclimated sub-cultures to ten temperatures spanning the thermal niche for ~12 generations (days to acclimation ranged from ~10 d at near-optimal temperatures to more than a month at 3°C). Approximately 105 cells from each acclimated culture were then propagated into an identical flask containing fresh L1 medium, and the cell density was estimated daily at the acclimated temperature until stationary phase (5-10 d, depending on the temperature). Light conditions were as above. Culture density was estimated daily at all temperatures ≥16°C by placing the polystyrene culture flask in a Shimadzu UV- 2401PC spectrophotometer (Shimadzu Corporation, Kyoto, Japan) and measuring Abs436 (the wavelength corresponding to the absorbance maximum of chlorophyll a: Neori et al., 1986); At 3 and 10°C, due to fogging of culture flasks, daily density estimates were conducted using the CASY counter instead; to make sure density estimation method did not bias estimates, we checked model residuals at 3 and 10°C against those from other temperatures, and found no indication of method-induced bias (see Appendix A, Figure A1). We calculated exponential growth rates by fitting regression lines to log-transformed population densities over time; with a minimum of three points that best represented exponential-phase growth (those that formed a straight line when log-transformed). Single cultures of each selection line were assayed at each temperature, except those at 16 and 31°C, which were assayed in triplicate to facilitate direct statistical comparison of growth rates (n = 14 total points per thermal reaction norm). This comparison serves as a “reciprocal transplant” experiment, a common approach to test for local adaptation (Kawecki & Ebert, 2004; Blanquart et al., 2013). 13 We derived temperature-dependent trait values by fitting a model proposed by Logan et al., (1976) and recently used by Thomas et al., (2017), in which birth and death are both exponential functions of temperature. The resulting curve is a unimodal, left-skewed thermal reaction norm. In the following double-exponential equation, the first term is the exponential birth term, the second term is the exponential mortality term: !"#$=&'()*+ − .'(/*+−.0 (1), where !"#$ is the specific growth rate (d-1), T is temperature, &' and .' are the pre-exponential constants for birth and death, respectively, &1 and .1 are the exponential rates of increase in both terms, and .0 is the temperature-independent death rate. To facilitate fitting this model, we used an alternative parameterization of (1) that includes an explicit Topt parameter, thus making it possible to directly compute 95% confidence intervals for this trait (see Appendix B); the two formulations are mathematically equivalent and yield similar fits. We determined the “sharpness” of the thermal reaction norm peaks by taking the second derivative of the thermal reaction norm with respect to temperature at Topt (“Peak sharpness” [Figure 3b] is the negative of this number), and the upward-concavity below Topt by computing the maxima in the second derivative across 0 ≤ T ≤ Topt. We estimated the inflection point of each thermal reaction norm by setting 2*3456 2+* =0 and solving for T within the range 0 ≤ T ≤ Topt. Nitrate-dependent growth assays Per-capita population growth as a function of external nitrate concentration can be described using the Monod equation (Monod, 1949): 14 '8/8/9= 3456#: 3456 ; #: (2), where < is the population density, = is the resource (nitrate) concentration, !"#$ is the resource- saturated (maximum) population growth rate, and > (liters µmol-1 d-1) is the nitrate growth ?. Growth affinity is a more mechanistic parameter because it is more directly indicative of competitive ability (Healy, 1980). The two parameters are related by ?=!@AB/>. For a full the affinity a, rather than the traditional formulation with the half-saturation constant for growth, affinity (the initial slope of the Monod curve). Note that this formulation directly incorporates derivation of Equation 2, see Appendix B. To determine if the nutrient-related traits changed as a result of adaptation to different temperatures, we estimated Monod growth curves for nitrate at 16 and 31°C after ~450 generations. The difference in the number of generations elapsed for the temperature and nutrient trait assays was due to the logistical infeasibility of running those experiments simultaneously. We investigated whether temperature selection had caused evolutionary divergence in nitrate growth affinities between lines from each temperature treatment when assayed at both selection temperatures, and whether this divergence corresponded to the tradeoff between the maximum growth rate and nitrate growth affinity. Selection lines were assayed in triplicate at each selection temperature and at ten nitrate concentrations (0, 2.5, 5 10, 20, 40, 60, 90, 180 and 882 µmol l-1), the latter being the N concentration of undiluted L1 medium). Prior to the Monod assay, all lines were acclimated to assay temperatures as above, followed by a 5-day N-starvation period in which cultures were kept in modified N-free L1 medium. Population growth rates were estimated as for temperature-dependent growth assays. In some replicates, an opaque, white contaminant 15 was visible, which appeared under the microscope to be bacteria; these replicates were excluded from analyses. Model fitting, trait estimation and statistical analysis We conducted all statistical analyses using R statistical programming language, version 3.3.2. We fit Equation (1) to temperature-dependent growth data using the “mle2()” function from the “bbmle” package (Bolker et al., 2017). Due to high uncertainty in some growth rate estimates in Monod assays, we used weighted, nonlinear least squares regression to fit Equation (2) to nutrient-dependent growth data, weighting growth rate estimates by their inverse variances. We compared traits derived from thermal reaction norms by conducting ANOVA on weighted least squares linear models. To quantify uncertainty in all trait estimates (except Topt) for each selection line, we conducted 1000 residual bootstraps of the double-exponential fits using a procedure described in Listmann et al., (2016). We incorporated this uncertainty into linear models by weighting each trait estimate for each selection line by the inverse of the variance of its bootstrap distribution (see Appendix A, Model 1), except in the case of Topt, where we were able to directly employ its variance, though Topt variances estimated directly were nearly identical to those estimated via bootstrapping (Appendix A, Figure A3). We used a nested ANOVA to compare maximum growth rates (µmax) at the selection temperatures in a reciprocal transplant (all 16°C- and 31°C-selected lines assayed at both temperatures). Uncertainty in µmax could be estimated directly (n = 3, no bootstraps), and we fit a selection temperature × assay temperature interaction and nested replicate selection line within selection temperature to test for local adaptation and strain-level selection temperature effects, 16 respectively (Appendix A, Model A2). We compared µmax and nitrate growth affinity (aNO3) estimates derived from fits of the Monod equation to growth rates estimated across a nitrate gradient at the selection temperatures (reciprocal transplant scenario) using two-way ANOVA conducted on weighted, least squares linear models (Appendix A, Equation A3), as above. Uncertainties in these trait estimates were estimated directly when fitting the Monod equation, as its parameters are the traits being compared; uncertainties were incorporated into the linear model as weights (1/E1), as was done with bootstrap variances above. To test for a global tradeoff between µmax and aNO3 within selection groups (pooled across assay temperatures), we used Deming regression, a form of orthogonal regression that allows for uncertainty in individual measurements and assumes unequal variances in x and y. We conducted this test using the “mcreg()” function in the “mcr” package in R (Manuilova et al., 2014). Note in the results below that we did not obtain t-statistics or p-values from the Deming regression, but that we used a bootstrapping option built into the mcreg() function to calculate 95% confidence intervals for slopes and intercepts. All data files and R code used in this study are included in the supporting information. Results Evolutionary change in thermal reaction norms Growth for 350 generations at two different temperatures, below and above the temperature optimum for T. pseudonana (16 and 31°C), led to a significant divergence of the thermal reaction norms (Figure 1). Topt was 1.87°C higher, on average, in selection lines evolved at 31°C than in those evolved at 16°C (F1,8 = 17.02; P = 0.0033; Figure 2a; Appendix A, Table A1); the maximum growth rate at Topt (µopt) was also ~0.1 d-1 higher in 31°C-selected lines (F1,8 = 7.99; P 17 = 0.022; Figure 2b; Appendix A, Table A1). CTmax trended higher in 31°C-selected lines than in 16°C-selected lines, with one exception (replicate line 31_5), though the difference was not statistically significant (F1,8 = 2.09; P = 0.19; Figure 2c; Appendix A, Table A1). Figure 1: Thermal reaction norms for per-capita population growth of 10 T. pseudonana populations. Growth rates were measured at 10 temperatures for each replicate population (five selected at 16°C [red] and five selected at 31°C [blue]) after 350 generations of experimental selection. Growth rates were measured in triplicate at the selection temperatures (reciprocal transplant scenario). N = 14 for each curve. A “reciprocal transplant” comparison between temperature treatments revealed that selection lines from each treatment had higher maximum growth rates (µmax) in their “home” temperature environments than in “away” environments; “local” lines also had higher growth 18 rates than “foreign” lines (assay temp. × selection temp. interaction: F1,8 = 254.82; P < 0.0001; Figure 2d; Appendix A, Table A1; see Appendix A, Figure A2 for population-level effects). Taken together, these results are the signature of local adaptation to temperature (Kawecki & Ebert, 2004; Blanquart et al., 2013). While µmax estimates differed slightly between 350 and 450 generations (measured for thermal reaction norm and Monod assays, respectively), differences were not directionally consistent across selection lines, and the overall pattern of local adaptation persisted (see next section; Appendix A, Figure A5a; Appendix A, Table A3). Figure 2: Thermal performance traits of 16°C- and 31°C-selected populations. I. (A) Thermal optimum (Topt); (B) Maximum growth rates at Topt (µopt); (C) Upper critical temperature (CTmax); (D) µmax at the selection temperatures in the reciprocal transplant scenario (both selection groups assayed at both selection temperatures). All trait estimates in (B-C) were 19 derived numerically from fits of the double-exponential model. Bars in (A) are Fisher’s 95% CI estimated directly from the standard errors of double-exponential Topt parameter estimates. Bars in (B-C) are 95% CI estimated as 2.5%-97.5% quantiles of bootstrap distributions (n = 1000 bootstraps for each replicate population); bars in (D) are 95% CI directly from triplicate growth rate estimates (no bootstrapping). In (A-C), the horizontal line across the whole panel is the grand mean across trait estimates for all replicate populations; in (D), horizontal lines are grand means for assay temperature groups. The peaks of the thermal reaction norms of 31°C-selected lines appeared sharper, and the left tails more deeply concave-up than those of 16°C -selected lines, both for the parametric (Figure 1) and nonparametric (Appendix A, Figure A6) thermal reaction norms. These observations, combined with the statistically nonsignificant difference in CTmax between the two selection regimes, led us to hypothesize that CTmax is less evolutionarily labile than Topt, resulting in greater skew in thermal reaction norms of 31°C-selected lines, rather than a simple rightward shift of the thermal reaction norms along the temperature axis. The difficulty in reliably estimating CTmin prevented statistical comparisons of CTmin and of thermal niche width among selection lines, and also prevented estimation of skewness. However, we determined the distance between Topt and CTmax for all bootstrapped thermal reaction norms; ideally, this distance would be scaled by thermal niche width, but absence of reliable thermal niche width estimates prevented scaling. The distance between Topt and CTmax (“upper tail length”) was smaller in 31°C-selected lines than in 16°C-selected lines (F1,8 = 6.58; P = 0.033; Figure 3a; Appendix A, Table A2), suggesting that the change in Topt was greater than the change in CTmax at one or, possibly, both selection temperatures. 31°C-selected lines had sharper peaks at Topt (F1,8 = 26.77; 20 P = 0.00085; Figure 3b; Appendix A, Table A2), were more deeply concave-up below Topt (F1,8 = 91.41; P < 0.0001; Figure 3c; Appendix A, Table A2), and had inflection points at higher temperatures than 16°C-selected lines (F1,8 = 12.36; P = 0.0079; Figure 3d; Appendix A, Table A2), suggesting greater negative skew. Figure 3: Thermal performance traits of 16°C- and 31°C-selected populations. II. (A) Upper tail lengths (CTmax – Topt) derived numerically from double-exponential curves (n = 1000 bootstraps). (B) Peak sharpness of double-exponential curves. Sharpness is calculated as the negative of the second derivative of the double-exponential curve with respect to temperature at Topt (larger values represent more negative gradients). (C) Upward concavity of double- exponential curves below Topt, calculated as the maximum of the second derivative across 0 ≤ T 21 ≤ Topt. (D) The inflection point below Topt, where 2*3456 2+* =0. Symbols are as in Figure 2. Evolution of competitive ability for nitrate Temperature selection did not lead to consistent evolutionary divergence in competitive ability for nitrate; there was no significant effect of assay temperature (F1,16 = 0.27; P = 0.61) or selection temperature (F1,16 = 1.26; P = 0.28) on nitrate growth affinity (aNO3), nor was there an assay temperature × selection temperature interaction (F1,16 = 0.12; P = 0.74; Appendix A, Table A3; Appendix A, Figure A4; Appendix A, Figure A5). The differences in aNO3 among selection groups or assay temperatures appeared idiosyncratic, with estimates ranging over nearly an order of magnitude across replicate populations. However, as in the reciprocal transplant above, there was a significant selection temperature × assay temperature interaction effect on µmax estimates derived from Monod fits (F1,16 = 10.79, P = 0.0047; Appendix A, Table A3; Appendix A, Figure A4; Appendix A, Figure 5). A tradeoff between the maximum growth rate and nitrate growth affinity was apparent in the 16°C-selected lines (Deming regression ± 95% CI: slope = -0.098 ± 0.075; intercept = 0.11 ± 0.074) but was not statistically significant in the 31°C-selected lines (slope = -0.21 ± 0.35; intercept = 0.24 ± 0.61), with replicates pooled across assay temperatures (Figure 4). 22 Figure 4: Deming regression plots for nitrate growth affinity (aNO3) versus maximum growth rate (µmax). Traits are estimated from maximum likelihood fits of the Monod equation to nitrate-dependent growth assay data. (A) 16°C-selected lines; (B) 31°C-selected lines. Curves are Deming regression fits. Confidence bands are calculated using a residual bootstrap of the regression fit (n = 1000 bootstraps), and account for uncertainty in each point, taken from Monod equation fits. All points are displayed with ±1 SEM from Monod parameter estimates. Note that y-axis scales differ between panels, and that y-axes in both panels are on a log scale, resulting in apparent curvature of linear fits. 23 Discussion Evidence that phytoplankton can evolve rapidly in response to environmental change is mounting (Collins & Bell, 2004; Hutchins et al., 2015; Listmann et al., 2016), but the response even to a single, directional selection pressure can be multifaceted (Low-Décarie et al., 2013; Schlüter et al., 2014, Hutchins et al., 2015). We showed that adaptation to different temperatures not only changes the growth rate at the selection temperature, but also significantly alters the shape of the whole thermal reaction norm, a function-valued trait. The changes included shifts of a critical temperature trait, Topt, and changes in the slope and curvature of the thermal reaction norm. Adaptation to high temperature resulted in a growth rate decline at low temperatures (a performance tradeoff). In contrast, despite some evidence for a tradeoff between maximum growth rate (µmax) and competitive ability for nitrate (indicated by nitrate growth affinity, aNO3) in the 16°C-selected populations (Figure 4), evolution of high µmax did not consistently lead to a reduction in aNO3 under the selection conditions in this study. Pleiotropic effects (Elena & Lenski, 2003) or resource allocation or acquisition tradeoffs (Gilchrist, 1995; Angilletta et al., 2003) associated with nitrogen acquisition or utilization were apparently minimal or absent. While we cannot be certain that selection was stronger and the degree of evolution greater in warm- versus cold-adapted populations or vice versa (we lack a common, ancestral thermal reaction norm for our T. pseudonana populations, as assays on ancestral strains were conducted in ESAW medium preventing direct comparisons to contemporary strains), we can confidently infer evolutionary divergence between 16°C- and 31°C-adapted populations from their measured derived traits (as in Huey et al., 1991). Selection at 31°C, well above the previously recorded Topt for T. pseudonana (Boyd et al., 2013) led to a divergence in Topt of ~2 °C, on average. Furthermore, at the conclusion of the evolution experiment, Topt of 16°C-selected 24 were comparable to previously observed Topt for T. pseudonana (Boyd et al. 2013), suggesting that the divergence between 16°C- and 31°C-selected populations was primarily driven by evolution of higher Topt in 31°C-selected populations. Concurrent with this shift, the maximum growth rate at Topt (µopt) diverged by ~0.1 d-1 (“vertical shift” [Kingsolver et al. 2005]); notably, the vertical shift also coincided with a narrowing of the region of the thermal reaction norm immediately adjacent to Topt in 31°C-selected lines, such that less of the curve was in the near- optimal range (Figure 1). All strains but one (replicate line 16_1) met both the “local versus foreign” and “home versus away” criteria for demonstrating local adaptation at both 16°C and 31°C (Kawecki & Ebert, 2004; Blanquart et al., 2013). The results suggest that the variation in T. pseudonana’s thermal reaction norm caused by thermal adaptation led to (or was driven by) a number of tradeoffs. First, while we could not precisely estimate CTmin or the thermal niche width, the performance tradeoff resulting from selection at 16°C versus at 31°C, combined with change in Topt (and a suggestion of change in CTmax, though this was non-significant) suggest some horizontal shift in the thermal reaction norm (Izem & Kingsolver, 2005; Kingsolver, 2009), as previously observed in bacteriophages (Knies et al., 2006) and the marine coccolithophore Emiliana huxleyi (Listmann et al., 2016)— however, this tradeoff may also result from thermal reaction norms of 31°C-selected populations becoming more concave-up below Topt, and from changes in their higher inflection points (Figure 3). In addition, both temperature “generalist-specialist” and resource allocation tradeoffs were apparent (Gilchrist 1995; Angilletta et al., 2003), leading to a “diagonal stretch” in thermal reaction norms toward the upper right. The higher, sharper peaks and compressed upper tails in the thermal reaction norm of 31°C-selected lines, accompanied by a reduction in fitness at 16°C, suggest greater specialization for growth at high temperatures in 31°C-selected lines and less 25 specialization in general in 16°C-selected lines. We hypothesized that the evolutionary changes in the shape of the thermal reaction norm in response to selection at high temperature may be explained by increased allocation to cellular repair machinery. To test the plausibility of this hypothesis, we modified Equation (1) to include differential allocation to repair machinery, which provides the benefit of high-temperature tolerance while incurring the costs of increased nutrient requirements and reduced growth. In high-temperature adapted species, the thermodynamically predicted temperature of maximum enzyme stability falls well below the statistically fit Topt (Ratkowsky et al., 1995; Corkrey et al., 2014), suggesting that it is rapid repair, rather than enhanced stability of enzymes that allows for positive growth at very high temperatures. The tradeoff in the following model assumes lack of plasticity in resource allocation to enzyme repair at high temperatures versus reproduction across temperatures (Angilletta et al., 2003). G(I)=&'()*+ − .'(/*+−.0, .'=/JKL M=M0+M'O &'=)JKP , (3.1) (3.2) (3.3) Here, the exponential mortality term (.'(/*+ in [3.1]) is weighted inversely by p, which (3.4) represents investment in protection from heat-induced denaturation of enzymes (or the induction of repair enzyme activity) (3.2); q is the nutrient content of a cell (quota), which is linearly related to p (3.3). Thus, as investment in protection and repair increases, temperature-dependent mortality decreases. However, temperature-dependent birth is also negatively affected as resources are diverted from reproduction and allocated to protection (3.4). The modified double- exponential model can produce changes in the curvature of the thermal reaction norm similar to 26 those observed in our T. pseudonana selection experiment (Figure 5). The model predicts that peak sharpness at Topt, upward concavity below Topt, and the location of the inflection point are all positive, saturating functions of p (Appendix A, Figure A7). Figure 5: The modified double-exponential model with allocation to thermal protection (p) ranging from 0.1 to 1. Other parameter values are as follows: b1 = 0.124, b2 = 0.1, d0 = 0.1, d1 = 3.059×10-7, d2 = 0.5, q0 =1, q1 = 0.3. There are two observed results that this model does not predict, though the second follows from the first. First, the model predicts comparable evolutionary lability of Topt and CTmax (Appendix A, Figure A7a), while in experiments, the difference in CTmax between 31°C- and 16°C-selected populations was not statistically significant. Evolution of CTmax may thus be constrained by other physiological barriers not incorporated in the model. Second, the model predicts that upper tail length is a positive, saturating function of p, only changing appreciably where p < 0.2 (Appendix A, Figure A7b). In contrast, the upper tails of our experimentally derived thermal reaction norms were more compressed in 31°C-selected lines, compared to 27 16°C-selected lines (Figure 3a), and ranged in width from ~4°C to ~9°C. Although evolution of CTmax in response to temperature selection has been observed in the bacterium Escherichia coli (Mongold et al., 1996) and in the marine coccolithophore Emiliana huxleyi (Listmann et al., 2016), the lack of significant change in CTmax relative to Topt in this study suggest that, at least in T. pseudonana, Topt is more evolutionarily labile than CTmax (Araújo et al., 2013). However, the vast diversity in CTmax in nature (Thomas et al., 2012; Corkrey et al., 2014), and in diatoms in particular (Thomas et al., 2016), indicates that such constraints are not evolutionarily insurmountable. It is possible that on timescales relevant to the current rate of increase of global sea surface temperature, constraints on evolution of a higher CTmax may cause regional extinctions, contributing to diversity loss, especially at low latitudes (Thomas et al., 2012). The tradeoff between µmax and resource affinities may be a general trend in nature (Grover, 1991; Klausmeier et al., 2007), but the overall weak and idiosyncratic response of nitrate affinity to temperature selection in our experiment may imply that temperature selection acted most strongly on systems other than those involved in nitrate uptake and utilization. However, the apparent tradeoff in 16°C-selected lines but not in 31°C-selected lines (Figure 4) suggests that thermal adaptation to high temperature may make the tradeoff weaker. In the absence of persistent nutrient limitation (and thus any selection for enhanced nutrient competitive ability, as was likely the case here), selection at a constant temperature should favor high µmax at that temperature. If a µmax-aNO3 tradeoff were driven by temperature selection, we would predict a concomitant decrease in competitive ability for nitrate, but changes in aNO3 in response to temperature selection were not consistent and did not suggest a tradeoff driven by temperature selection. Here, µmax at both “home” and “away” temperatures changed predictably in response to temperature selection, but aNO3 did not. The absence of an assay temperature effect 28 on aNO3 suggests that any µmax-aNO3 tradeoff does not depend on the temperature environment in the short term, at least within the temperature range explored here, and that strong selection acted on µmax alone, without pleiotropic effects on nitrate competitive ability in most replicates. Thermal adaptation in the ocean would frequently occur under nitrogen limitation, at least in the temperate ocean where T. pseudonana is commonly found (Fong, 2008), and may thus produce changes in nitrogen affinities different from those we observed here. Therefore, we suggest that future temperature selection experiments should also account for nutrient limitation to reflect possible climate change scenarios in natural systems. Evolution experiments are essential to enhancing our understanding of the effects of climate change on marine and freshwater food webs. Arguably, phytoplankton are among the highest-payoff organisms upon which we can conduct evolution experiments: they offer a unique combination of potential for rapid evolutionary responses to environmental change, tractability as experimental subjects, and global ecological importance (Reusch & Boyd, 2013). Eco- evolutionary responses to changes in temperature, ocean acidity, nutrient limitation and other environmental factors are complex, and are made more complex by interactions among these factors. Single-stressor evolution experiments are a valuable first step, but future evolution experiments must account for interactive effects of multiple stressors (e.g., temperature and nitrate limitation). 29 APPENDICES 30 APPENDIX A: Chapter 1 supplemental figures, tables, and equations 31 A.1: Double-exponential residual plots s l a u d i s e r l e d o m l a i t n e n o p x e − e l b u o D . 0 2 16_1 1 . 0 0 . 0 1 . 0 − 2 . 0 − . 0 2 31_1 16_2 16_3 16_4 16_5 31_2 31_3 31_4 31_5 1 . 0 0 . 0 1 . 0 − 2 . 0 − 5 10 15 20 25 30 5 10 15 20 25 30 10 15 5 Assay temperature (°C) 25 20 30 5 10 15 20 25 30 5 10 15 20 25 30 Figure A1: Residual plots for fits of the double-exponential model to data from temperature-dependent growth assays of all replicate T. pseudonana populations at 16 and 31°C. In each panel, note the two points farthest to the left (assays at 3°C and 10°C). n = 14 in each panel. Replicate population ID is indicted in the upper left of each panel. A.2: Linear models and statistical output. Models compare traits shown in Figure 1 and Figure 2. Model A1. Thermal performance traits (Figures 2A-C) RS>TR~<(V0+V'× W(X(YRTZ[ R(\O., E1 _⁄ ), where _=1/E)1 ; the subscript b indicates that this is the variance of the bootstrap distribution, 32 not to be confused with E1 of the linear model above. Data fit to Model 1 are trait estimates numerically derived from the double-exponential model (Thomas et al. 2017). In Model 1, replicate selection lines are pooled within assay temperature, as there were insufficient degrees of freedom to test for strain-level effects. To quantify uncertainty in trait estimates compared in Model 1, we conducted 1000 residual bootstraps of the double-exponential fits using a procedure identical to that described in Listmann et al. (2016). We incorporated the uncertainty estimated from bootstrap distributions into linear models by weighting each trait estimate by the inverse of the variance of its bootstrap distribution. Model A2: abcd in a reciprocal transplant (Figure 2D) !"#$ ~ <(V0+V'× >ee>f R(\O.+ V1× W(X(YRTZ[ R(\O.+ Vg(>ee>f R(\O × W(X(YRTZ[ R(\O)+V'(h)× WRS>T[,E1) Note that this is not a weighted model. We used Model 2 to directly compare maximum growth rates (µmax) at the selection temperatures in a reciprocal transplant (all 16°C- and 31°C-selected lines assayed at both temperatures). Uncertainty in µmax could be estimated directly (n = 3, no bootstraps), and we fit a selection temperature × assay temperature interaction and nested replicate selection line within selection temperature to test for local adaptation and strain-level selection temperature effects, respectively. 33 Model A3: abcd and ijkl from Monod equation fits (Figure 5) RS>TR ~ <(V0+V'× >ee>f R(\O.+ V1× W(X(YRTZ[ R(\O.+ Vg(>ee>f R(\O × W(X(YRTZ[ R(\O),E1/_) Model 3 was applied to µmax and nitrate growth affinity (aNO3) estimates derived from fits of the Monod equation to growth rates estimated across a nitrate gradient (10 concentrations) at the selection temperatures, also in a reciprocal transplant scenario. Uncertainties in these estimates were estimated directly when fitting the Monod equation, as its parameters are the traits being compared; these uncertainties were incorporated into the linear model as weights (w = 1/E)1), as in Model 1. ANOVA tables show statistical outputs of ANOVA conducted on linear models (Table A1; Table A2; Table A3). Following Table A1, Figure A1 shows within-group, replicate population-level effects of Model 2. Table A1: ANOVA table for traits described in Figure 2: thermal optimum (Topt), maximum growth rate at Topt (µopt), upper critical temperature (CTmax) (all Model 1), and maximum growth rates at the selection temperatures (µmax) in a reciprocal transplant (Model 2). Linear model coefficient df MS F-value P-Value 17.02 0.0033 1 8 1 8 1 8 6.48 0.38 20.41 2.55 6.57 3.15 -- 7.99 -- 2.09 -- -- -- 0.022 0.19 -- Trait Source Topt (°C) Selection temperature Residual µopt (d-1) Selection temperature Residual CTmax (°C) Selection temperature Residual (95% CI) 1.87 (0.83, 2.92) 0.095 (0.018, 0.17) 0.30 (-0.18, 0.77) 34 -0.051 (-0.15, 0.048) -0.19 (-0.29, -0.094) 0.43 (0.29, 0.57) See Figure A2 -- Table A1 (cont’d) µmax (d-1) at selection temps Assay temperature Selection temperature Assay temp. × Selection temp Sel. temp. : Strain Residual Assay31:strain31_4 Assay31:strain31_3 Assay31:strain31_2 Assay31:strain31_1 Assay31:strain16_5 Assay31:strain16_4 Assay31:strain16_3 Assay31:strain16_2 Assay16:strain31_4 Assay16:strain31_3 Assay16:strain31_2 Assay16:strain31_1 Assay16:strain16_5 Assay16:strain16_4 Assay16:strain16_3 Assay16:strain16_2 Assay temp x Selection temp Selection temp Assay.temp Intercept −0.2 0.0 0.2 0.4 Regression estimate <0.0001 0.69 <0.0001 0.052 -- 1 1 1 8 40 0.092 0.00058 0.92 0.0069 0.0036 25.43 0.16 254.82 1.89 -- 0.6 0.8 Figure A2: Regression coefficients from Model 2, describing µmax at the selection temperatures in a reciprocal transplant (n = 3). For Assay temp : strain estimates, strains 16.1 and 31.5 are reference levels and thus do not appear on this plot. Bars are 95% CI. 35 16°C lines Upper Lower 31°C lines Upper Lower 9 2 ) 8 2 C ° ( t i i m L e c n e d i f n o C p a r t s t o o B 7 2 6 2 5 2 4 2 3 2 2 2 22 24 23 28 Fisher’s Information Confidence Limit (°C) 25 26 27 1:1 29 Figure A3: Comparison of Fisher’s Information 95% confidence limits to bootstrap confidence limits for Topt parameter in double-exponential model fits to temperature- dependent growth data. Table A2: ANOVA tables comparing traits described in Figure 3. Traits compared between selection temperature groups: Upper tail length (CTmax -Topt), Peak sharpness at Topt, upper concavity below Topt, and the inflection point of the thermal reaction norm (see Model A1). Trait Source Upper tail length (°C) Selection temperature Residual Peak sharpness Selection temperature Residual Upward concavity Selection temperature Residual Inflection point (°C) Selection temperature Residual Linear model coefficient (95% CI) df MS F-value P-Value 1 8 1 8 1 8 1 8 28.50 4.33 38.39 1.43 45.66 0.86 32.15 2.60 6.58 -- 26.77 -- -- -- 0.033 0.00085 -- -- -- -- 0.0023 (0.0014, 0.0029) 91.41 < 0.0001 5.07 (1.75, 8.39) 12.36 0.0079 -1.47 (-2.80, -0.1495) 0.011 (0.0060, 0.016) 36 Table A3: ANOVA comparing maximum growth rates (µmax) and nitrate affinities (aNO3). Traits are compared between temperature selection groups and between assay temperatures in the reciprocal transplant scenario (see Model A3). Strain-level effects could not be tested, due to insufficient degrees of freedom. Trait Source µmax (d-1) Assay temperature Selection temperature Assay temp. × Selection temp. Residual ln[aNO3 (l µmol-1 d-1)] Assay temperature Selection temperature Assay temp. × Selection temp. Residual Coefficient (95% CI) df MS F-value P-Value -0.12 (-0.21, -0.18) -0.088 (-0.18, -0.00085) 0.27 (0.11, 0.43) -0.13 (-0.66, -0.41) 0.31 (-0.18, 0.79) -0.045 (-0.88, 0.79) 0.00047 1 0.026 1 1 2.83 16 0.26 0.068 1 0.32 1 1 0.029 16 0.25 0.0018 0.10 10.79 -- 0.27 1.26 0.12 -- 0.97 0.76 0.0047 -- 0.61 0.28 0.74 -- 37 A.3: Figures A4-A6 Figure A4: Nitrate-dependent growth (Monod) assays at 16°C (top, blue boxes) and 31°C (bottom, red boxes). Curves were fit to T. pseudonana exponential growth rates at ambient NO3 concentrations ranging from 0 to 882 µmol l-1 (the NO3 concentration in unaltered L1 marine medium). For each panel, n ≤ 30 (see “Methods”). Confidence bands were generated from fits to 1000 residual bootstraps. 38 (a) (b) Figure A5: Nitrate-dependent growth (Monod) kinetic parameters at ~450 generations. (a) T. pseudonana maximum specific growth rates (µmax) for 16°C- and 31°C-selected lines (blue and red, respectively) assayed at 31°C and at 16°C. (b) Nitrate growth affinity (aNO3). Bars are 95% CI, based on uncertainties from bootstrap Monod parameter distributions (n = 1000). Note that the y-axis in panel B is on a log scale, hence the asymmetry in confidence intervals. Figure A6: Nonparametric plot of thermal reaction norms (n = 14 for each). Bars on points at the two selection temperatures are ±1 SE (n = 3 at each temperature). 39 A.4: Graphical analysis of the modified double-exponential model (Eq. 3 and 4 in main text) Panels show theoretical relationships between the traits estimated for T. pseudonana and the “protection” parameter, p, in the modified double-exponential model. (cid:9) (cid:8) (cid:7) (cid:1) (cid:6) (cid:5) (cid:4) (cid:3) (cid:2) (cid:1) (cid:8)(cid:4) (cid:8)(cid:3) (cid:8)(cid:1) (cid:3)(cid:6) (cid:3)(cid:5) (cid:3)(cid:4) (cid:3)(cid:3) (cid:3)(cid:1) (a) (b) (cid:5)(cid:1)(cid:6)(cid:7)(cid:8) (cid:1)(cid:2)(cid:3)(cid:4) (cid:1)(cid:2)(cid:3) (cid:1)(cid:2)(cid:4) (cid:3) (cid:1)(cid:2)(cid:5) (cid:1)(cid:2)(cid:6) (cid:7)(cid:2)(cid:1) Figure A7: Relationships between thermal performance traits and the protection parameter (p) in Equations (3). (a) CTmax and Topt, (b) upper tail length, (c) thermal niche width, (d) inflection point, and (e) peak sharpness, and (f) upward concavity as functions of protection, as defined in Equation (3). 40 Figure A7 (cont’d) (c) (d) (e) (f) 41 APPENDIX B: Chapter 1 supplemental derivations 42 B.1: Analytical determination of Topt First, we derive Topt for Equation (1) by taking its derivative with respect to temperature (T) and setting it equal to 0. Solving for T yields the solution for Topt. G(I)= &'()*+ − .'(/*+−.0 G′(I)= &'&1()*+ − .'.1(/*+=0 InL9=log(&'&1/.'.1)/(.1−&1) G(I)= (1/M)&'()*+ − (1/O).'(/*+−.0 G′(I)= &'&1()*+/M − .'.1(/*+/O=0 InL9=log(.'.1M/&'&1O)/(&'−.1) InL9 can likewise be derived from the modified double-exponential model. (Equation 1) (1’) (B1) (B2) (B2’) (B3) B.2: Re-parameterization of the double-exponential model with explicit Topt Equation (B3) can be solved for &': &'=/J/*)* ((/*r)*)+stu, (B4) which can then be substituted into Equation (1) and simplified, yielding: G(I)=/J/*)* ()*∙+;(/*r)*)+stu−(.0+.'(/*+). (B5) 43 Equation (B5) has the same number of parameters (5) as Equation (1), can be directly parameterized using Maximum Likelihood Estimation, and contains Topt as an explicit parameter. This formulation both facilitates fitting by allowing “eyeballing” of the starting value for Topt, and facilitates calculating confidence intervals for Topt estimates based on properties of likelihood surfaces rather than requiring bootstrapping approaches. For Equation (B5) to produce a canonical unimodal thermal performance curve, d1 > d2. Additionally, d0 > 0 for the function to cross zero below Topt, ensuring a proper CTmin value. B.3: Derivation of Equation 2 from Monod (1949) The Monod (1949) equation describing per-capita population growth as a function of resource concentration is '8/8/9= 3456: :;w , (B6) where < is the population density, !"#$ is the nutrient-saturated (maximum) Malthusian parameter, = is the resource concentration, and K is the half-saturation constant (the value of R at which growth is 0.5 × !"#$. According to Healy (1980), ?= 3456# , so it follows that Equation (B6) can be written as '8/8/9= 3456: :;(y4565 ) (B6’) 44 Multiplying the right side by ##, we see that :;(y4565 )×##= 3456#: '8/8/9= 3456: 3456 ; #: , which is Equation (2). 45 REFERENCES 46 REFERENCES Angilletta MJ, Wilson RS, Navas CA, James RS. 2003. Tradeoffs and the evolution of thermal Araújo M, Ferri-Yáñez F, Bozinovic F, Marquet PA, Valladares F, et al. 2013. Heat freezes reaction norms. Trends in Ecology and Evolution 18:234-240. niche evolution. Ecology Letters 16:1206–1219. Baker KG, Robinson CM, Radford DT, McInnes AS, Evenhuis C, et al. 2016. Thermal performance curves of functional traits aid understanding of thermally induced changes in diatom-mediated biogeochemical fluxes. Frontiers in Marine Science 3:1–14. Bell G. 2013. Evolutionary rescue and the limits of adaptation. Philosophical Transactions of the Royal Society of London B: Biological Sciences 368:20120080. Bennett AF, Lenski RE. 1993. Evolutionary adaptation to temperature II. Thermal niches of experimental lines of Escherichia coli. Evolution 47:1-12. Blanquart F, Kaltz O, Nuismer SL, Gandon S. 2013. A practical guide to measuring local adaptation. Ecology Letters 16:1195–1205. Bolker B, R Core Team. 2017. bbmle: tools for general maximum likelihood estimation. R package version 1.0.19. https://cran.r-project.org/web/packages/bbmle/index.html. Boyce DG, Lewis MR, Worm B. 2010. Global phytoplankton decline over the past century. Nature 466:591–596. Boyd PW, Rynearson TA, Armstrong EA, Fu F, Hayashi K, et al. 2013. Marine phytoplankton temperature versus growth responses from polar to tropical waters – outcome of a scientific community-wide study. PLoS One 8:e63091. Bradford MA. 2013. Thermal adaptation of decomposer communities in warming soils. Frontiers in Microbiology 4:1-16. Collins S, Bell G. 2004. Phenotypic consequences of 1,000 generations of selection at elevated CO2 in a green alga. Nature 431:566–569. Corkrey R, McMeekin TA, Bowman JP, Ratkowsky DA, Olley J, et al. 2014. Protein thermodynamics can be predicted directly from biological growth rates. PloS One 9:e96100. Crawfurd KJ, Raven JA, Wheeler GL, Baxter EJ, Joint I. 2011. The response of Thalassiosira pseudonana to long-term exposure to increased CO2 and decreased pH. PloS One 6:e26695. 47 Elena SF, Lenski RE. 2003. Evolution experiments with microorganisms: the dynamics and genetic bases of adaptation. Nature Reviews Genetics 4:457–469. Field CB, Behrenfeld MJ, Randerson JT, Falkowski PG. 1998. Primary production of the biosphere: integrating terrestrial and oceanic components. Science 281:237–240. Fong P. 2008. Macroalgal-dominated ecosystems. In: Nitrogen in the Marine Environment, 2nd ed. (eds Capone DG, Bronk DA, Mulholland MR, Carpenter EJ), pp. 917-947. Academic Press (Elsevier), Cambridge, MA, USA. Gilchrist GW. 1995. Specialists and generalists in changing environments. I. Fitness landscapes of thermal sensitivity. The American Naturalist 146:252–270. Gomulkiewicz R, Holt RD. 1995. When does evolution by natural selection prevent extinction? Evolution 49:201–207. Grover JP 1991. Resource competition in a variable environment: phytoplankton growing according to the variable-internal-stores model. The American Naturalist 138:811–835. Guillard R, Hargraves P. 1993. Stichochrysis immobilis is a diatom, not a chrysophyte. Phycologia 32:234-236. Harrison P, Waters R, Taylor F. 1980. A broad spectrum artificial seawater medium for coastal and open ocean phytoplankton. Journal of Phycology 16:28-35. Healy FP. 1980. Slope of the Monod equation as an indicator of advantage in nutrient competition. Microbial Ecology 5:281-286. Hinners J, Kremp A, Hense I. 2017. Evolution in temperature-dependent phytoplankton traits revealed from a sediment archive: do reaction norms tell the whole story? Proceedings of the Royal Society B 284:20171888. Hoffmann AA, Sgrò CM. 2011. Climate change and evolutionary adaptation. Nature 470:479- 485. Huey RB, Partridge L, Fowler K. 1991. Thermal sensitivity of Drosophila melanogaster responds rapidly to laboratory natural selection. Evolution 45:751-756. Hutchins DA, Walworth NG, Webb EA, Saito MA, Moran D, et al. 2015. Irreversibly increased nitrogen fixation in Trichodesmium experimentally adapted to elevated carbon dioxide. Nature Communications 6:8155. Izem R, Kingsolver JG. 2005. Variation in continuous reaction norms: quantifying directions of biological interest. The American Naturalist 166:277–289. 48 Jiang H, Gao K. 2004. Effects of lowering temperature during culture on the production of (cid:1) polyunsaturated fatty acids in the marine diatom Phaeodactylum tricornutum (Bacillariophyceae). Journal of Phycology 40:651–654. Jin P, Gao K, Beardall J. 2013. Evolutionary responses of a coccolithophorid Gephyrocapsa oceanica to ocean acidification. Evolution 67:1869–1878. Kawecki TJ, Ebert D. 2004. Conceptual issues in local adaptation. Ecology Letters 7:1225– 1241. Kingsolver JG, Gomulkiewicz R, Carter PA. 2001. Variation, selection and evolution of function-valued traits. Genetica 112:87–104. Kingsolver JG. 2009. The well-temperatured biologist. The American Naturalist 174:755–768. Klausmeier CA, Litchman E, Daufresne T, Levin SA. 2004. Optimal nitrogen-to-phosphorus stoichiometry of phytoplankton. Nature 429:171–174. Knies JL, Izem R, Supler KL, Kingsolver JG, Burch CL. 2006. The genetic basis of thermal reaction norm evolution in lab and natural phage populations. PLoS Biology 4:e201. Kutcherov D. 2016. Thermal reaction norms can surmount evolutionary constraints: comparative evidence across leaf beetle species. Ecology and Evolution 6:4670–4683. Listmann L, LeRoch M, Schlüter L, Thomas MK, Reusch TBH. 2016. Swift thermal reaction norm evolution in a key marine phytoplankton species. Evolutionary Applications 9:1156– 1164. Litchman E, Klausmeier CA, Schofield OM, Falkowski PG. 2007. The role of functional traits and tradeoffs in structuring phytoplankton communities: scaling from cellular to ecosystem level. Ecology Letters 10:1170–1181. Litchman E, Edwards K, Klausmeier CA, Thomas MK. 2012. Phytoplankton niches, traits and eco-evolutionary responses to global environmental change. Marine Ecology Progress Series 470:235–248. Logan JA, Wollkind DJ, Hoyt SC, Tanigoshi LK. 1976. An analytical model for description of temperature dependent rate phenomena in arthropods. Environmental Entomology 5:1133- 1140. Lohbeck KT, Riebesell U, Reusch TBH. 2012. Adaptive evolution of a key phytoplankton species to ocean acidification. Nature Geosciences 5:346–351. Low-Décarie E, Jewell MD, Fussmann GF, Bell G. 2013. Long-term culture at elevated atmospheric CO2 fails to evoke specific adaptation in seven freshwater phytoplankton species. Proceedings of the Royal Society B 280:20122598. 49 Manuilova E, Schuetzenmeister A, Model F. 2014. mcr: Method Comparison Regression. R package version 1.2.1. https://cran.r-project.org/package=mcr. Martiny AC, Ma L, Mouginot C, Chandler JW, Zinser ER. 2016. Interactions between thermal acclimation, growth rate, and phylogeny influence Prochlorococcus elemental stoichiometry. PLoS One 11:e0168291. Mongold JA, Bennett AF, Lenski RE. 1996. Evolutionary adaptation to temperature. IV. Adaptation of Escherichia coli at a niche boundary. Evolution 50:35-43. Monod J. 1949. The growth of bacterial cultures. Annual Reviews in Microbiology 3:371-394. Nedwell D. 1999. Effect of low temperature on microbial growth: lowered affinity for substrates limits growth at low temperature. FEMS Microbiology and Ecology 30:101–111. Nelson DM, Tréguer P, Brzezinski MA, Leynaert A, Quéguiner B. 1995. Production and dissolution of biogenic silica in the ocean: revised global estimates, comparison with regional data and relationship to biogenic sedimentation. Global Biogeochemical Cycles 9:359–372. 50 Neori A, Vernet M, Holm-Hansen O, Haxo FT. 1986. Relationship between action spectra for chlorophyll a fluorescence and photosynthetic O2 evolution in algae. Journal of Plankton Research 8:537-548. Padfield D, Yvon-Durocher G, Buckling A, Jennings S, Yvon-Durocher G. 2015. Rapid evolution of metabolic traits explains thermal adaptation in phytoplankton. Ecology Letters 19:133–142. Ratkowsky DA, Olley J, Ross T. 2005. Unifying temperature effects on the growth rate of bacteria and the stability of globular proteins. Journal of Theoretical Biology 233:351–362. Redfield AC. 1958. The biological control of chemical factors in the environment. American Reusch TBH, Boyd PW. 2013. Experimental evolution meets marine phytoplankton. Evolution Scientist 46:205–221. 67:1849–1859. Rowan R. 2004. Thermal adaptation in reef coral symbionts. Nature 430:742. Schaum C-E, Barton S, Bestion E, Buckling A, Garcia-Carreras B, et al. 2017. Adaptation of phytoplankton to a decade of experimental warming linked to increased photosynthesis. Nature Ecology Evolution 1:0094. Schiffers K, Bourne EC, Lavergne S, Thuiller W, Travis JMJ. 2013. Limited evolutionary rescue of locally adapted populations facing climate change. Philosophical Transactions of the Royal Society of London B: Biological Sciences 368:20120083. Schlüter L, Lohbeck KT, Gutowska MA, Gröger JP, Riebesell U, et al. 2014. Adaptation of a globally important coccolithophore to ocean warming and acidification. Nature Climate Change 4:1024–1030. Schoener TW. 2011. The newest synthesis: understanding the interplay of evolutionary and ecological dynamics. Science 331:426–429. Thomas MK, Kremer CT, Klausmeier CA, Litchman E. 2012. A global pattern of thermal adaptation in marine phytoplankton. Science 338:1085–1088. Thomas MK, Kremer CT, Litchman E. 2016. Environment and evolutionary history determine phytoplankton temperature traits globally. Global Ecology & Biogeography 25:71-86. Thomas MK, Aranguren-Gassis M, Kremer CT, Gould MR, Anderson K, et al. 2017. Temperature-nutrient interactions exacerbate sensitivity to warming in phytoplankton. Global Change Biology 23:3269-3280. Thrane J-E, Hessen DO, Anderson T. 2017. Plasticity in algal stoichiometry: experimental evidence of a temperature-induced shift in optimal supply N:P ratio. Limnology and Oceanography 62:1346-1354. Tilman DA, Mattson M, Langer S. 1981. Competition and nutrient kinetics along a temperature gradient: an experimental test of a mechanistic approach to niche theory. Limnology and Oceanography 26:1020-1033. Yvon-Durocher G, Schaum C-E, Trimmer M. 2017. The temperature dependence of phytoplankton stoichiometry: investigating the roles of species sorting and local adaptation. Frontiers in Microbiology 8:2003. 51 CHAPTER 2: Experimental evolution of fatty acid thermal reaction norms in a marine diatom Introduction Temperature affects practically every aspect of poikilotherms’ physiology, from rates of respiration (Beamish & Mookherjii 1964; Caron et al. 1986; Bradford 2013) and photosynthesis (Neori & Holm-Hansen 1982; Robarts & Zohary 1987; Pastenes & Horton 1996) to reproductive rates (Knoblauch & Jorgensen 1999; Lee & Ahn 2000; Thomas et al. 2012) and body size (Montagnes & Franklin 2001; Peter & Sommer 2013; Lindmark et al. 2018). As body temperature rises, the physical properties of cellular structures change: proteins change conformation or denature (Ratkowsky et al. 2005; Corkrey et al. 2014) and membranes become more fluid and permeable to solutes (reviewed in Hazel 1995). These changes necessitate the evolution of physiological strategies to maintain cellular function in a thermally variable environment. Rates of resource acquisition, efficiencies of resource use (Bradford 2013), and strategies for storing resources such as carbon, nitrogen and phosphorus can lead to changes in relative resource requirements (Thrane et al. 2017), internal stoichiometry (Sakamoto & Bryant 1998; Thrane et al. 2017), and the content and composition of fatty acid (FA) molecules (Canvin 1965; James et al. 1989; Thompson et al. 1992b; Teoh et al. 2004). One strategy for maintaining membrane function in a thermally variable environment is to change the FA composition of cellular membranes to maintain homeoviscosity of membranes across temperature within an organism’s thermal niche (Sinensky 1975). FA in cell walls and organelle membranes may become increasingly unsaturated at low temperatures (Patterson 1970; Joh et al. 1993; Jiang & Gao 2004) as unsaturated FA (UFA) pack together more loosely, conferring greater membrane fluidity and permeability; conversely, membrane FA tend to be 52 more saturated at high temperatures (Quinn 1981; Wada et al. 1990; Murata & Los 1997; Zhu et al. 1997), as saturated FA (SFA) pack together tightly and in a more ordered fashion, conferring greater membrane rigidity (Quinn 1981; Murata & Los 1997; Renaud et al. 2002; Converti et al. 2009). Plasticity in FA content and composition has been studied in a wide variety of taxa (Sinensky 1974; Bhakoo and Herbert 1979; reviewed in Neidleman 1987), however the FA content and composition of phytoplankton, in particular, has attracted extensive study, in part due to their potential utility as feedstock for production of biofuels (Hu et al. 2008; Converti et al. 2009; Williams & Laurens 2010) and to their global importance as the leading source of omega-3 FA (Litzow et al. 2006; Kang 2011; Hixson & Arts 2016). Published trends in algal FA profiles across temperature are extremely variable, even within a given functional group (e.g. diatoms); however, a number of generalities seem to exist. Specifically: 1) overall FA content is generally highest at the thermal niche boundaries (Patterson 1970; Thompson et al. 1992a; Van Wagenen et al. 2012), which results in an overall negative correlation between population growth rate and FA content. Notably, a negative correlation between growth rate and FA content in algae has also been observed when the former was controlled by factors other than temperature, such as irradiance (Shifrin & Chisolm 1991; Thompson et al. 1990; Williams & Laurens 2010) or nutrients (Yang et al. 2018). 2) As has been observed in other organisms, % SFA tends to increase with temperature, especially below the thermal optimum for population growth (Topt). 3) Below Topt, % polyunsaturated FA (PUFA) tends to decrease with temperature. 4) Trends in % monounsaturated FA (MUFA) across temperature are variable and often non- significant (Patterson 1970; Sato et al. 1979; Thompson et al. 1992b; Renaud et al. 2002; Rousch et al. 2003; Pasquet et al. 2014). Exceptions to these trends can be found in most of the studies 53 cited here. While plasticity in FA profiles has been characterized for many phytoplankton species (see previous paragraph), the aim of this study is to investigate evolutionary change in temperature-dependent reaction norms for total FA content, FA composition (representation of FA classes as a percentages of total FA), and overall saturation/unsaturation of FA in response to long-term experimental selection at low (16°C) and high (31°C) temperatures. Evolutionary adaptation to an extreme temperature may lead to a temperature-specialized FA profile (a hot- cold tradeoff). While an increase in temperature can lead to greater FA saturation over the short term via physiological plasticity (Wada et al. 1990; Murata & Los 1997; Zhu et al. 1997), prolonged exposure (over many generations) to high temperature may cause an evolutionary reduction in the ability to maintain homeoviscosity through production of UFA should the temperature drop (a tradeoff). A similar tradeoff may be predicted in the opposite direction for cold-adapted populations. Alternatively, thermal adaptation may simply lead to increased plasticity in the FA profile, contributing to (or resulting from) the broadening of the thermal niche (see Izem & Kingsolver 2005; Kingsolver 2009). Thermal reaction norms for per-capita population growth have been shown to evolve rapidly in response to experimental selection in phytoplankton (Listmann et al. 2016; Padfield et al. 2016; O’Donnell et al. 2018) and other organisms (Mongold et al. 1996; Knies et al. 2006; Geerts et al. 2013). However, only a few studies have examined the metabolic, stoichiometric or biochemical causes and consequences of these changes (Schlüter et al. 2014; Padfield et al. 2016; O’Donnell et al. 2018). Knowledge of evolutionary changes in the temperature dependence of cellular chemistry in poikilotherms is crucial to achieving a comprehensive understanding of the effects of temperature selection at the level of the organism. This study is 54 among the first to examine the effects of experimental thermal adaptation on the fatty acid profile of a phytoplankton species. Materials and Methods 500 generation temperature selection experiment The temperature-dependent growth assays described here were conducted after ~500 generations of thermal adaptation at 16°C and 31°C, a detailed account of which can be found in O’Donnell et al. (2018). Briefly, we maintained five replicate populations of Thalassiosira pseudonana CCMP1335 (Hustedt) Hasle et Heimdal (obtained from the Provasoli-Guillard National Center for Culture of Marine Phytoplankton (CCMP), Maine, USA) in exponential growth under cool, white, fluorescent bulbs at a photon flux density of 110 µmol photons m-2 s-1 on a 14:10 light:dark cycle, transferring 106 cells into fresh L1 marine culture medium (Guillard & Hargraves 1993; modified by O’Donnell et al. 2018) every four days, with occasional deviations of ±1 day. After the selection experiment, all populations were cryopreserved in liquid nitrogen until they were needed, at which point they were revived for assays (see below). Cryopreservation and revival To cryopreserve T. pseudonana populations, we first grew them to late log-phase in their selection (“home”) temperature environments before placing 1 ml of culture in a 2 ml cryogenic vial. We then added 1 ml of 24% dimethyl sulfoxide (mixed with 76% L1 marine medium), for a final DMSO concentration of 12%. After 20 minutes, cryogenic vials were placed in a passive freezer unit (Mr. Frosty, Thermo Scientific, USA) containing 250 ml isopropyl alcohol, and cooled to -80°C at ~1 °C min-1, after which vials were plunged and stored in liquid nitrogen (- 55 196 °C). We revived cultures two weeks prior to beginning acclimation for temperature-dependent assays. We thawed cryogenic vials by floating them in 500 ml beakers full of 20°C tap water. We preformed this operation in the dark. Once thawed, each vial was immediately emptied into a 20 ml scintillation vial containing 10 ml of L1 marine medium, also at 20°C. Scintillation vials were loosely capped and placed in a light-proofed cardboard box at 20°C for 36 h (as in Hipkin et al. 2013), after which they were placed under cool, white fluorescent lights at 110 µmol photons m-2 s-1 on a 14:10 light:dark cycle. Each vial was gently inverted daily to keep cells in suspension. As soon as a revived culture grew to a visible density, it was transferred to a 50 ml tissue culture flask containing 20 ml of L1 marine medium. Starting at this point, all cultures were maintained at 20 °C for 10 generations prior to acclimation at assay temperatures. Algal growth and harvesting We assayed all five replicate T. pseudonana populations from each experimental selection temperature group (16°C and 31°C) in triplicate at 10, 16, 26 and 31°C (2 selection temperatures x 4 assay temperatures x 5 populations x 3 replicates = 120 total cultures). We chose these temperatures to achieve broad coverage of T. pseudonana’s physiology at temperatures spanning the majority of its thermal tolerance range, and to include the two selection temperatures and one temperature (26°C) approximating T. pseudonana’s optimal temperature for population growth (Boyd et al. 2013; O’Donnell et al. 2018). We conducted acclimations and assays in temperature-controlled growth chambers, under cool, white fluorescent grow lights at 110 µmol photons m-2 s-1 on a 14:10 light:dark cycle. Each replicate population was maintained in exponential growth for a 10-generation acclimation period at each 56 assay temperature prior to the assay. From each replicate population, 105 cells were transferred to each of three 50 ml polystyrene tissue culture flasks containing 40 ml of sterile-filtered L1 marine culture medium. We monitored relative change in cell densities daily by placing each tissue culture flask in a spectrophotometer (Shimadzu UV-2401PC; Shimadzu Corporation, Kyoto, Japan) and measuring Abs436, corresponding to the absorbance peak of chlorophyll-a (Neori et al. 1986). We harvested all 120 cultures in late log-phase; time to harvest varied by temperature. We filtered 15 ml of culture through a Whatman GF/B glass fiber filter for fatty acid analysis and froze these samples at -20°C. We retained 25 µl for population density and mean cell size (µm3) measurements using a CASY particle counter (Schärfe System GmbH, Reutlingen, Germany); the remaining ~25 ml of each culture was filtered for analyses not addressed here. Fatty Acid Methyl Ester reaction and Gas Chromatography To analyze the fatty acid profiles of replicate T. pseudonana populations, we extracted neutral lipids from the frozen, filtered samples described above and performed fatty acid methyl ester (FAME) reactions on each according to the protocol by Wang and Benning (2011) and adapted for algae by Boyle et al. (2012). We first extracted lipids from Whatman GF/B glass fiber filters using a solution of chloroform, methanol and formic acid (10:20:1). After washing with a buffer (0.2M H3PO4 + 1M KCl), we spun samples down for three minutes at 2000 x g, transferring the remaining chloroform layer to a fresh tube and dried the samples with a stream of N2 gas. After drying, we added 100 µl of internal 15:0 standard (50 µg ml-1 pentadecanoic acid), then proceeded with the FAME protocol (Wang et al. 2011; Boyle et al. 2012) and quantified fatty acids using an Agilent Technologies 7890A Gas Chromatography system 57 (Agilent Technologies, Santa Clara, CA, USA). Calculations All percentages presented here are calculated based on molar masses of fatty acid molecules on a per-biovolume basis (µm-3). Individual fatty acids are presented as percentages of total fatty acid weight (molar). Mean chain length (MCL) is calculated as {|}= ∑ , where \ZX ÜáÉ is the weight in moles of a given fatty acid, \ZX Üá9n9 is "n ÄÅusu× XYÉ "n ÄÅÇ 'ÑÉÖ' the total weight of all fatty acids, and XYÉ is the length, (number of carbon atoms) of ÜáÉ (Renaud et al. 2002). Degree of unsaturation (“unsaturation”, hereafter) is calculated as àâ[W>R= ∑ , where the first term is as above, and [.&É is the number of double bonds "n ÄÅusu× [.&É "n ÄÅÇ 'ÑÉÖ' in ÜáÉ (Thompson et al. 1992b). Linear mixed-effects models All linear mixed models describing unsaturation, chain length, etc. are nested versions of the same model, fit using the “lme” function from R package “nlme” (Pinheiro & Bates 2018). The full model includes the main effects selection temperature and assay temperature as factors, plus an interaction between these two factors. For each replicate selection line, we included random intercepts and slopes to account for repeated measures across assay temperatures. Finally, the full model includes two identity variance functions (which allow for different variances by factor level: Pinheiro & Bates 2018) to account for unequal variances between selection temperature levels and among assay temperature levels. These variance functions are included as weights in the model using the “varIdent” function from nlme (Pinheiro and Bates 2018). To include two varIdent functions simultaneously, the “varComb” function (also from 58 nlme) was used. In some cases, weights and/or selection temperature ´ assay temperature interaction terms did not improve model fits and were dropped from the model. We used likelihood ratio tests to select models; formal models can be found in Appendix C. Results As observed in other marine diatoms (see, e.g., Harwood and Jones 1989), T. pseudonana contains FA ranging in length from C14 to C22, with substantial representation of longer (C20 and C22) FA molecules. The dominant fatty acid in both 16°C- and 31°C-selected populations was 16:1Δ3 at all four assay temperatures (Figure 11). Total fatty acid content per biovolume (µm3) trended higher on average in 31°C-selected populations than in 16°C-selected populations, though not significantly (Table 1; Figure 7A), and total fatty acid content was lowest in both groups at 10°C and highest at 31°C (Table 1; Figure 7A). The percentage of total FA represented by saturated fatty acids (SFA) was higher in 16°C-selected populations than in 31°C-selected populations, though not at the 10°C assay temperature; the difference in % SFA was especially pronounced at the 31°C assay temperature, where % SFA in 31°C-selected populations was ~20% lower than in 16°C-selected populations (Table 1; Figure 7B). The percentage of total FA represented by monounsaturated FA (MUFA) was ~10% higher in 31°C-selected populations than in 16°C-selected populations, but did not vary across assay temperatures (Table 1; Figure 7C). Percent polyunsaturated FA (% PUFA) was not different between the two selection groups at 10, 16, and 26°C, however 16°C-selected populations experienced a ~14% decrease in % PUFA between 26°C and 31°C, leading to a significant difference in % PUFA between the two groups at the 31°C assay temperature (Table 1; Figure 7D). 59 (a) (b) Figure 6: Total content of 16 fatty acid classes in T. pseudonana selected at 16 °C (blue shades) and 31°C (red shades) and grown at (a) 10°C; (b) 16°C; (c) 26°C; and (d) 31°C. error bars are ±1 SE. Y-axis values are multiplied by 1010 for clarity. 60 Figure 6 (cont’d) (c) (d) 61 Table 1: RM-ANOVA testing effects of selection temperature and assay temperature on T. pseudonana mean chain length (MCL), unsaturation (WUnSat), total fatty acids (Total FA), percent saturated FA (% SFA), percent monounsaturated FA (% MUFA), percent polyunsaturated FA (% PUFA), and percentages of individual FA classes. Effect Assay temp. Selection temp. Error Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error Assay temp. Selection temp. Error Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error % 14:0 Assay temp. Selection temp. Error % 16:0 Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error Trait MCL WUnSat Total FA (mol) % SFA % MUFA % 18:0 Assay temp. Selection temp. Error % 20:0 Assay temp. Selection temp. Error % 16:1D7 Assay temp. % 16:1D3 Assay temp. Assay temp. Selection temp. Error Selection temp. Error Selection temp. Error % 18:1D9 Assay temp. Selection temp. Error Table 1 continued on next page 62 df 3,12 1,96 112 3,12 1,93 3,93 112 3,12 1,93 109 3,12 1,94 3,94 110 3,12 1,97 113 3,12 1,94 3,94 110 3,12 1,97 113 3,12 1,97 113 3,12 1,97 113 3,12 1,97 113 3,12 1,97 113 3,12 1,97 113 MS 2.98 0.32 0.089 0.87 0.24 0.24 0.035 0.027 0.012 0.0053 329.3 467.7 244.8 33.9 343.6 276.6 14.30 267.47 38.55 146.04 11.68 1017.0 8.30 9.90 28.29 2.77 1.32 24.90 163.00 22.06 20.56 0.004 1.58 132.85 1.55 10.92 0.25 0.0002 0.26 F-statistic 33.31 3.61 -- 24.97 7.01 6.83 -- 5.04 2.21 -- 9.71 13.79 7.22 -- 24.03 19.35 -- 22.89 3.30 12.50 -- 102.46 0.83 -- 21.43 2.10 -- 1.13 15.22 -- 13.03 0.002 -- 12.97 0.14 -- 0.97 0.001 -- P-value <0.0001 0.060 -- <0.0001 0.0093 0.0003 -- 0.0026 0.14 -- <0.0001 0.0003 0.0002 -- <0.0001 <0.0001 -- <0.0001 0.072 <0.0001 -- <0.0001 0.36 -- <0.0001 0.15 -- 0.34 0.0076 -- <0.0001 0.96 -- <0.0001 0.71 -- 0.41 0.98 -- Table 1 (cont’d) % 18:1D11 Assay temp. % PUFA * No model fit, due to zero inflation. Selection temp. Assay temp. ´ Sel. Temp. Error Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error Selection temp. Assay temp. ´ Sel. Temp. Error Selection temp. Assay temp. ´ Sel. Temp. Error Selection temp. Assay temp. ´ Sel. Temp. Error Selection temp. Assay temp. ´ Sel. Temp. Error -- -- Selection temp. Assay temp. ´ Sel. Temp. Error Selection temp. Assay temp. ´ Sel. Temp. Error % 16:2D7,10 Assay temp. % 16:2D9,12 Assay temp. % 16:3D7,10,13 Assay temp. % 16:4D4,7,10,13 Assay temp. * % 18:2D9,12 * % 18:3D9,12,15 % 20:5D5,8,11,14,17 Assay temp. % 22:6D4,7,10,13,16,19 Assay temp. 11.04 25.36 2.84 -- 6.34 1.47 5.74 -- 9.81 4.41 16.78 -- 57.81 7.05 13.43 -- 41.51 15.09 11.39 -- 110.08 6.26 1.85 -- -- -- 27.37 0.002 4.13 -- 7.32 0.96 0.74 -- <0.0001 <0.0001 0.041 -- 0.0005 0.23 0.0011 -- <0.0001 0.038 <0.0001 -- <0.0001 0.0091 <0.0001 -- <0.0001 0.0002 <0.0001 -- <0.0001 0.014 0.14 -- -- -- <0.0001 0.97 0.0081 -- 0.0002 0.33 0.53 -- 58.29 133.88 15.01 5.28 337.7 78.50 305.8 51.30 3.72 1.67 6.36 0.38 55.64 6.78 12.93 0.96 125.08 45.45 34.32 3.01 59.92 3.41 1.01 0.54 -- -- 219.89 0.010 33.17 8.03 131.55 17.29 13.20 17.96 3,12 1,94 3,94 110 3,12 1,94 3,94 110 3,12 1,94 3,94 110 3,12 1,94 3,94 110 3,12 1,94 3,94 110 3,12 1,94 3,94 110 -- -- 3,12 1,94 3,94 110 3,12 1,94 3,94 110 63 (a) (c) (b) (d) Figure 7: (a) Total fatty acid (FA) content per biovolume in 16°C-selected (blue) and 31°C- selected (red) populations at 10, 16, 26, and 31°C; (b) Saturated FA as a fraction of total FA; (c) Monounsaturated FA as a fraction of total FA; (d) Polyunsaturated FA as a fraction of total FA. n = 3 for each replicate population. Lines across boxes indicate the median of each group, with the lower and upper bounds indicate the 25th and 75th quartiles, respectively. Points beyond the lower and upper whiskers are outliers, falling outside the 0th and 100th quartiles. The composite metric of FA unsaturation (WUnSat) was largely driven by % PUFA. Unsaturation was higher overall in 31°C-selected populations (Table 1; Figure 8A), with the greatest difference in unsaturation at the 31°C assay temperature (Table 1 Figure 8A). Mean 64 chain length (C atoms per FA molecule) was also higher in 31°C-selected populations overall (Table 1; Figure 8B). (a) (b) Figure 8: (a) Mean chain length (C atoms per molecule); (b) degree of unsaturation (double bonds per molecule) in 16°C-selected (blue) versus 31°C-selected (red) T. pseudonana populations at 10, 16, 26, and 31°C. n = 3 for each replicate population. Symbols are as above. While composite metrics of FA content and unsaturation differed between selection groups primarily at the 31°C assay temperature, these metrics mask some underlying divergence (or lack thereof) in thermal reaction norms between selection groups for key FA classes. We will highlight a few notable FA classes in the text; for all others, refer to Table 1 and Figures 4-6. Among the SFA, 16:0 FA were notable in that, at the 10°C assay temperature, 31°- selected populations had higher % 16:0, while at 16°C there was little difference between selection groups and at 26 and 31°C, 31°C-selected populations had the highest % 16:0 (Table 1; Figure 9b). Percent 18:0 FA decreased ~60% between the 16°C and 26°C assay temperatures in both selection groups, dropping an additional ~25% between 26°C and 31°C (Table 1; Figure 65 9c); % 18:0 FA was the only SFA that was consistently higher in 31°C selected populations across assay temperatures, with the greatest difference observed at 10°C (Table 1; Figure 9c). (a) (b) (c) (d) Figure 9: Saturated fatty acids as percentages of total fatty acids in 16°C-selected (blue) versus 31°C-selected (red) T. pseudonana populations at 10, 16, 26, and 31°C. n = 3 for each replicate population. Numbers in the lower (a,b) and upper right corners indicate the fatty acid classes shown in each panel; for example, “14:0” is a FA with 14 C atoms and no double bonds. Other symbols are as above. Among the MUFA, 16:1D7 and 16:1D3 FA represented between 20 and 30% of total FA in both selection groups and at all assay temperatures. However, these two FA classes differed the least between the two temperature selection groups. Thermal reaction norms for % 16:1D7 FA classes were parallel and differed by << 1% at every assay temperature (Table 1; Figure 10a). Percent 16:1D3 reaction norms were effectively identical (Table 1; Figure 10b). Assay temperature ´ selection temperature interaction effects were non-significant, did not improve model fits, and were dropped from both % 16:1 FA models altogether, as was also the case for % 18:1D9. Percent 18:1D11 was ~3x higher in 31°C-selected populations than in 16°C-selected 66 populations across assay temperatures (Table 1; Figure 10d). Percent 18:1D11 was lowest for both selection groups at the 31°C assay temperature (assay temperature ´ selection temperature interaction Table 1; Figure 10d). (a) (b) (c) (d) Figure 10: Monounsaturated fatty acids as percentages of total fatty acids in 16°C-selected (blue) versus 31°C-selected (red) T. pseudonana populations at 10, 16, 26, and 31°C. n = 3 for each replicate population. Numbers in the lower (a,b) and upper right corners indicate the fatty acid classes shown in each panel. “Δ” indicates the position, from the carboxyl end, of the double bond; for example, “Δ3” indicates a double bond in the third position form the carboxyl end. Other symbols are as above. Percent 16:3D7,10,13 FA was higher in 16°C-selected populations than in 31°C-selected populations at 10°C, while the reverse was true at 16, 26, and 31°C (Table 1; Figure 11c). Percent 16:4D4,7,10,13 FA was consistently higher in 31°C-selected populations than in 16°C- selected populations (Table 1; Figure 11d), but decreased by roughly an order of magnitude between 10°C and 31°C assay temperatures in both selection groups (Table 1; Figure 11d). 18:2D9,12 and 18:3D9,12,15 FA were notable in their near-absence. Combined, they rarely exceeded 67 1% of total FA, and were often undetectable, preventing statistical analysis. 20:5D5,8,11,14,17 and 22:6D4,7,10,13,16,19 FA were the two longest (most C atoms) FA classes, and each was notable in its relative abundance, representing between 1% and 10% of total FA. Percent 20:5D5,8,11,14,17 and % 22:6D4,7,10,13,16,19 did not differ between selection groups overall, but was higher in 31°C-selected populations at the 31°C assay temperature (assay temperature ´ selection temperature interaction: Table 1; Figure 11g,h). Percent 20:5D5,8,11,14,17 was highest overall for both selection groups at 26°C (Table 1; Figure 11g), while % 22:6D4,7,10,13,16,19 peaked for both groups at 16°C, then declined by nearly an order of magnitude between 16°C and 31°C (Table 1; Figure 11h). (a) (b) (c) (d) (e) (f) (g) (h) Figure 11: Polyunsaturated fatty acids as percentages of total fatty acids in 16°C-selected (blue) versus 31°C-selected (red) T. pseudonana populations at 10, 16, 26, and 31°C. n = 3 for each replicate population. Symbols are as above. 68 Discussion Previous studies have suggested that many poikilotherms increase levels of FA saturation with increasing temperature to maintain homeoviscosity of membranes (Patterson 1970; Sinensky 1974; Joh et al. 1993; reviewed in Neidleman et al. 1987; Jiang & Gao 2004). Indeed, Thompson et al. (1992b) observed a strong negative trend in the ratio of unsaturated to saturated FA in T. pseudonana between 10 and 25°C. We therefore predicted that 1) T. pseudonana in our study would exhibit a similar negative trend in unsaturation across temperature, and 2) selection of T. pseudonana populations for ~500 generations at 16°C would lead to a greater predisposition to FA unsaturation at low temperatures, and possibly across all assay temperatures if plasticity in FA unsaturation was low. Conversely, 31°C-selected populations should evolve enhanced production of saturated FA to maintain membrane rigidity at high temperatures, or across all temperatures if the same tradeoff applied. Many of our results were nearly exactly opposite these predictions. We showed that the temperature dependence of fatty acid profiles in Thalassiosira pseudonana can evolve rapidly in response to experimental temperature selection. Replicate T. pseudonana populations that underwent ~500 generations of experimental natural selection at low (16°C) and high (31°C) temperatures exhibited marked divergence in FA saturation at the four assay temperatures (10, 16, 26, and 31°C), driven by major divergence in thermal reaction norms of a few key FA classes. However, the temperature-dependent responses of fatty acid content and composition in both T. pseudonana temperature selection groups were complex, idiosyncratic, and often opposite what we predicted based on prior research (Patterson 1970; Sinensky 1975; Jiang & Gao 2004; Wada et al. 1990); for example, SFA were higher at low temperatures and UFA were higher at high temperatures in many cases. We note that unexpected 69 and idiosyncratic responses of phytoplankton FA to temperature seem to be common among previous studies of the topic (Thompson et al. 1992a,b; Rousch et al. 2003; Jiang & Gao 2004). Average total FA per biovolume varied little across assay temperatures, but was lowest at 10°C and highest at 31°C for both selection groups, on average. In contrast, Patterson (1970) showed that FA as a percentage of dry weight in the chlorophyte Chlorella sorokiniana increased toward the thermal niche boundaries, and was lowest near Topt. Van Wagenen et al. (2012) showed a similar pattern for the ochrophyte Nannochloropsis salina. In eight species of phytoplankton including T. pseudonana, Thompson et al. (1992a) found thermal reaction norms for lipid cell-1 to be highly variable, ranging from concave up to concave down between 10 and 25°C; of particular note, however, the reaction norm for T. pseudonana was distinctly concave- up. While it is possible that our assay temperature range simply failed to adequately capture the temperature regions closest to the niche boundaries, we would point out that assay temperatures in Thompson et al. (1992a) ranged from 10°C to 25°C, and that ours matched (on the low end) or exceeded (on the high end) this range. Percent SFA decreased slightly between 10 and 31°C in 16°C-selected populations, with a dramatic drop between 16 and 26°C; % SFA decreased dramatically in 31°C-selected populations between 10°C and 31°C; and % SFA was significantly higher in 16°C-selected populations than in 31°C-selected populations at all assay temperatures except 10°C. These unexpected trends are not unprecedented. Teoh et al. (1994) observed increased unsaturation at low temperatures in only one of six phytoplankton species studied. There were some indications, however, that T. pseudonana adjusted percentages of certain PUFA as a means of maintaining homeoviscosity across temperature, and that thermal reaction norms for these PUFA evolved in response to temperature selection in the direction predicted. First, % PUFA, and, consequently, 70 WUnSat (average double bonds per FA molecule) were slightly higher in 16°C-selected populations than in 31°C-selected populations at 10°C. Second, the most highly unsaturated fatty acids, 20:5D5,8,11,14,17 and 22:6D4,7,10,13,16,19, which ranged roughly from 1% to 10% of total FA in both 16 and 31°C-selected populations, were significantly higher in 16°C-selected populations than in 31°C-selected populations at 10, 16 and 26°C, but were significantly lower at 31°C. It would appear, therefore, that in addition to evolving enhanced production of these highly- unsaturated FA (HUFA), 16°C-selected populations were also more plastic in their ability to regulate the amount of HUFA produced, depending on temperature. Percent 22:6D4,7,10,13,16,19 decreased overall in both groups between 10 and 31°C, but 20:5D5,8,11,14,17 did not, in contrast to previous studies (Jiang & Gao 2004; Pasquet et al. 2014). Besides unsaturation, FA chain length (C atoms per molecule) has been suggested to play a role in regulating membrane fluidity (Sinensky 1974; Hochachka & Somero 1984). These authors suggest that, as temperature increases, microbes tend to incorporate longer-chain FA into their membranes, as shorter FA molecules (especially SFA) “melt” at lower temperatures (Sinensky 1974; Hochachka & Somero 1984). Incorporation of short-chain SFA (14:0 and 16:0 FA, in this case) into membranes when cold and long-chain SFA (18:0 and 20:0 here) when warm may thus be considered a strategy for maintaining homeoviscosity across temperature. However, 14:0 and 16:0 FA have melting points of 53.9 and 63.1°C, respectively (Hochachka & Somero 1984), both of which lie well outside the thermal niche of T. pseudonana. No SFA detected in our study were thus ever in liquid phase. The longest-chain SFA in our study were 18:0 and 20:0, both of which decreased their % abundances dramatically between 10 and 31°C. Mean chain length (MCL), which incorporates both SFA and UFA, was fairly constant between 10 and 26°C in both T. pseudonana temperature selection groups and dropped significantly from 71 26 to 31°C. MCL was also significantly higher in 31°C-selected populations than in 16°C- selected populations at the 31°C assay temperature. While % 14:0 FA was higher in 16°C- selected populations than in 31°C-selected populations across assay temperatures, both 14:0 and 16:0 FA increased by ~50% in 16°C-selected populations between 10 and 31°C, indicating a tendency for cold-adapted T. pseudonana to increase production of short, saturated FA at high, rather than at low temperatures. This is consistent with some studies more recent than Sinensky (1974) and Hochachka & Somero (1984), which have shown non-significant (Thompson et al. 1992b; Renaud et al 1995; Renaud et al. 2002) or negative (Renaud et al 1995; Renaud et al. 2002) trends in MCL with increasing temperature. In sum, favoring shorter or longer MCL does not appear to be a strategy for maintaining homeoviscosity across temperature in T. pseudonana populations in either of our temperature selection groups, nor does this strategy appear to be the rule in other phytoplankton species. Some well-targeted follow-up research could shed a great deal of light on temperature dependence of FA saturation and desaturation in temperature-selected populations of T. pseudonana, and in phytoplankton in general. For example, a study of mechanisms of FA desaturation at the cellular level in temperature-selected T. pseudonana populations combined with information on underlying molecular evolution would be a powerful and illuminating follow-up to this study. Seven desaturases were identified in T. pseudonana (Tonon et al. 2005) using the draft genome sequenced by Armbrust et al. (2004). Assays of the activities of these desaturases in cold- and warm-adapted T. pseudonana populations across a broad temperature range could be combined with the amino acid sequences of the desaturases themselves to better understand the role of FA profile evolution in T. pseudonana’s thermal adaptation. 72 Phytoplankton, especially diatoms, are a vital source of fatty acids for marine organisms at higher trophic levels, and are the primary source of long-chain HUFA in the world’s marine (Kang 2011; Hixson & Arts 2016) and freshwater (Kainz et al. 2004) systems. Marine-derived HUFA are also a direct and essential link between marine phytoplankton and human health (Kang 2011), and global production of HUFA by marine phytoplankton may be declining as a result of rising sea surface temperatures (Hixson & Arts 2016); climate effects on HUFA production in phytoplankton must therefore be considered of global human interest. Studies of the effect of warming on the content and composition of FA in marine and freshwater phytoplankton communities have primarily focused on multi-species responses (Hixson & Arts 2016) or shifts in phytoplankton assemblages leading to changes in the availability of vital HUFA to higher trophic levels (Litzow et al. 2006; Kang 2011). Here, we showed that FA content and composition within a species of marine diatom, Thalassiosira pseudonana, is sensitive to temperature on both short (physiological) and long (evolutionary) timescales, and that thermal reaction norms of key FA are evolutionarily labile. Importantly, T. pseudonana populations experimentally adapted to 16°C and 31°C for ~500 generations exhibited different responses to medium temperature by three highly unsaturated omega-3 FA, 16:4D4,7,10,13, 20:5D5,8,11,14,17, and 22:6D4,7,10,13,16,19. 16:4 and 22:6 FA declined sharply between 10 and 31°C in all experimental populations, but 31°C-selected populations were significantly higher in all three omega-3 HUFA at the highest assay temperature, 31°C. Thus, while warming may lead to a decrease in production of some key HUFA, T. pseudonana may be able to recover some HUFA production through rapid evolutionary responses to warming seas. Long-term implications of evolutionary adaptation to warming in phytoplankton for availability of HUFA to higher trophic levels, including humans, are worthy of extensive future research. 73 REFERENCES 74 REFERENCES Armbrust EV, Berges JA, Bowler C, Green BR, Martinez D, et al. 2004. The genome of the diatom Thalassiosira pseudonana: ecology, evolution, and metabolism. Science 306:79–86. Beamish FWH, Mookherjii PS. 1964. Respiration of fishes with special emphasis on standard oxygen consumption. I. Influence of weight and temperature on respiration of goldfish, Carassius auratus L. Canadian Journal of Zoology 50:161-175. Bhakoo M, Herbert RA. 1979. The effects of temperature on fatty acid and phospholipid composition of four obligately psychrophilic Vibrio spp. Archives of Microbiology 121:121- 127. Boyle NR, Page MD, Liu B, Blaby IK, Casero D, et al. 2012. Three acyltransferases and nitrogen-responsive regulator are implicated in nitrogen starvation-induced triacylglycerol accumulation in Chlamydomonas. Journal of Biological Chemistry 287:15811–15825. Boyd PW, Rynearson TA, Armstrong EA, Fu F, Hayashi K, et al. 2013. Marine phytoplankton temperature versus growth responses from polar to tropical waters—outcome of a scientific community-wide study. PLoS One 8:e63091. Bradford M. 2013. Thermal adaptation of decomposer communities in warming soils. Frontiers in Microbiology 4:1-16. Canvin DT. 1965. The effect of temperature on the oil content and fatty acid composition of the oils from several oil seed crops. Canadian Journal of Botany 43:63-69. Caron DA, Goldman JC, Dennett MR. 1986. Effect of temperature on growth, respiration, and nutrient regeneration by an omnivorous microflagellate. Applied Environmental Microbiology 52:1340-1347. Chen M, Tang H, Ma H, Holland TC, Ng KYS, et al. 2011. Effect of nutrients on growth and lipid accumulation in the green algae Dunaliella tertiolecta. Bioresource Technology 102:1649–1655. Converti A, Casazza AA, Ortiz EY, Perego P, Del Borghi M. 2009. Effect of temperature and nitrogen concentration on the growth and lipid content of Nannochloropsis oculata and Chlorella vulgaris for biodiesel production. Chemical Engineering and Processing: Process Intensification 48:1146–1151.(cid:1) Corkrey R, McMeekin TA, Bowman JP, Ratkowsky DA, Olley J, et al. 2014. Protein thermodynamics can be predicted directly from biological growth rates. PLoS One 9:e96100. Dean AP, Sigee DC, Estrada B, Pittman JK. 2010. Using FTIR spectroscopy for rapid 75 determination of lipid accumulation in response to nitrogen limitation in freshwater microalgae. Bioresource Technology 101:4499–4507. Gardner R, Peters P, Peyton B, Cooksey KE. 2011. Medium pH and nitrate concentration effects on accumulation of triacylglycerol in two members of the chlorophyta. Journal of Applied Phycology 23:1005–1016. Geerts AN, Vanoverbeke J, Vanschoenwinkel B, Van Doorslaer W, Feuchtmayr H, et al. 2015. Rapid evolution of thermal tolerance in the water flea Daphnia. Nature Climate Change 5:665-669. Guillard R, Hargraves P. 1993 Stichochrysis immobilis is a diatom, not a chrysophyte. Guschina IA, Harwood JL. 2006. Lipids and lipid metabolism in eukaryotic algae. Progress in Phycologia 32:234–236. Lipid Research 45:160-186. 53. Harwood JL, Jones AL. 1989. Lipid metabolism in algae. Advances in Botanical Research 16:1- Hazel JR. 1995. Thermal adaptation in biological membranes: is homeoviscous adaptation the explanation? Annual Review of Physiology 57:19-42. Hipkin R, Day JG, Rad-Menéndez C, Mock T. 2013. The first evidence for genotypic stability in a cryopreserved transgenic diatom. Journal of Applied Phycology 26:65-71. Hixson SM, Arts MT. 2016. Climate warming is predicted to reduce omega-3, long-chain, polyunsaturated fatty acid production in phytoplankton. Global Change Biology 22:2744- 2755. Hochachka PW, Somero GN. 1984. Temperature Adaptation. In: Biochemical Adaptation, Hochachka PW, Somero GN [Authors]. Princeton University Press, Princeton, NJ, USA. pp. 355-449. Hu Q, Sommerfeld M, Jarvis E, Ghirardi M, Posewitz M, et al. 2008. Microalgal triacylglycerols as feedstocks for biofuel production: Perspectives and advances. The Plant Journal 54:621- 639. Jiang H, Gao K. 2004. Effects of lowering temperature during culture on the production of polyunsaturated fatty acids in the marine diatom Phaeodactylum tricornutum (Bacillariophyceae). Journal of Phycology 40:651–654. Joh T, Yoshida T, Yoshimoto M, Miyamoto T, Hatano S. 1993. Composition and positional distribution of fatty acids in polar lipids from Chlorella ellipsoidea differing in chilling susceptibility and frost hardiness. Physiologia Plantarum 89:285–290. 76 Kainz M, Arts MT, Mazumder A. 2004. Essential fatty acids in the planktonic food web and their ecological role for higher trophic levels. Limnology and Oceanography 49:1784-1793. Kang JX. 2011. Omega-3: A link between global climate change and human health. Biotechnology Advances 29:388-390. Knies JL, Izem R, Supler KL, Kingsolver JG, Burch CL. 2006. The genetic basis of thermal reaction norm evolution in lab and natural phage populations. PLoS Biology 4:e201. Knoblauch C, Jorgensen B. 1999. Effect of temperature on sulphate reduction, growth rate and growth yield in five psychrophilic sulphate-reducing bacteria from Arctic sediments. Environmental Microbiology 1:457-467. Lee J-H, Ahn JJ. 2000. Temperature effects on development, fecundity, and life table parameters of Amblyseius womersleyi (Acari: Phytoseiidae). Environmental Entomology 29:265-271. Lindmark M, Huss M, Ohlberger J, Gårdmark A. 2018. Temperature-dependent body size effects determine population responses to climate warming. Ecology Letters 21:181-189. Listmann L, LeRoch M, Schlüter L, Thomas MK, Reusch TBH. 2016. Swift thermal reaction norm evolution in a key marine phytoplankton species. Evolutionary Applications 9:1156- 1164. Litzow MA, Bailey KM, Prahl FG, Heintz R. 2006. Climate regime shifts and reorganization of fish communities: The essential fatty acid limitation hypothesis. Marine Ecology Progress Series 315:1-11. Lynn SG, Kilham SS, Kreeger DA, Interlandi SJ. 2000. Effect of nutrient availability on the biochemical and elemental stoichiometry in freshwater diatom Stephanodiscus minutulus (Bacillariophyceae). Journal of Phycology 36:510–522. McLachlan JL, Curtis JM, Boutilier K, Keusgen M, Seguel MR. 1999. Tetreutreptia pomquetensis (Euglenophyta), a psychrophilic species: growth and fatty acid composition. Journal of Phycology 35:280-286. Mongold JA, Bennett AF, Lenski RE. 1996. Evolutionary adaptation to temperature. IV. Adaptation of Escherichia coli at a niche boundary. Evolution 50:35-43. Montagnes DJS, Franklin DJ. 2001. Effect of temperature on diatom volume, growth rate, and carbon and nitrogen content: reconsidering some paradigms. Limnology and Oceanography 46:2008-2018. Mortensen SH, Børsheim KY, Rainuzzo JR, Knutsen G. 1988. Fatty acid and elemental composition of the marine diatom Chaetoceros gracilis Schütt. Effects of silicate deprivation, temperature and light intensity. Journal of Experimental Marine Biology and Ecology 122:173-185. 77 Murata N, Los D. 1997. Membrane fluidity and temperature perception. Plant Physiology 115:875-859. Neidleman SL. 1987. Effects of temperature on lipid unsaturation. Biotechnology and Genetic Engineering Reviews 5:245-268. Neori A, Holm-Hansen O. 1982. Effect of temperature on rate of photosynthesis in Antarctic phytoplankton. Polar Biology 1:33-38. Neori A, Vernet M, Holm-Hansen O, Haxo FT. 1986 Relationship between action spectra for chlorophyll a fluorescence and photosynthetic O2 evolution in algae. Journal of Plankton Research 8:537–548. O’Donnell DR, Hamman CR, Johnson EC, Klausmeier CA, Litchman E. 2018. Rapid thermal adaptation in a marine diatom reveals constraints and trade-offs. Global Change Biology doi: 10.1111/gcb.14360. [Epub ahead of print] Padfield D, Yvon-Durocher G, Buckling A, Jennings S, Yvon-Durocher G. 2015. Rapid evolution of metabolic traits explains thermal adaptation in phytoplankton. Ecology Letters 19:133-142. Pasquet V, Ulmann L, Mimouni V, Guihéneuf F, Jacquette B, et al. 2014. Fatty acids profile and temperature in the cultured marine diatom Odontella aurita. Journal of Applied Phycology 26:2265–2271. Pastenes C, Horton P. 1996. Effect of high temperature on photosynthesis in beans. I. Oxygen evolution and chlorophyll fluorescence. Plant Physiology 112:1245-1251. Peter KH, Sommer U. 2013. Phytoplankton cell size reduction in response to warming mediated by nutrient limitation. PLoS One 8:e71528. Pinheiro J, Bates D, R Core Team. 2017. nlme: linear and nonlinear mixed effects models. R package version 3.1-131. https://cran.r-project.org/web/packages/nlme/nlme.pdf. Quinn PJ. 1981. The fluidity of cell membranes and its regulation. Progress in Biophysics and Molecular Biology 38:1-104. Ratkowsky DA, Olley J, Ross T. 2005. Unifying temperature effects on the growth rate of bacteria and the stability of globular proteins. Journal of Theoretical Biology 233:351-362. Reitan KI, Rainuzzo JR, Olsen Y. 1994. Effect of nutrient limitation on fatty acid and lipid content of marine microalgae. Journal of Phycology 30:972–979. Renaud SM, Zhou HC, Parry DL, Thinh L-V, Woo KC. 1995. Effect of temperature on the growth, total lipid content and fatty acid composition of recently isolated tropical microalgae 78 Isochrysis sp., Nitzschia closterium, Nitzschia paleacea, and commercial species Isochrysis sp. (clone T.ISO). Journal of Applied Phycology 7:595-602. Renaud SM, Thinh L-V, Lambrinidis G, Parry DL. 2002. Effect of temperature on growth, chemical composition and fatty acid composition of tropical Australian microalgae grown in batch cultures. Aquaculture 211:195–214. growth rates of bloom-forming cyanobacteria. New Zealand Journal of Marine and Robarts R, Zohary T. 1987. Temperature effects on photosynthetic capacity, respiration, and Freshwater Research 21:391-399. Sakamoto T, Bryant D. 1998. Growth at low temperature causes nitrogen limitation in the cyanobacterium Synechococcus sp . PCC 7002. Archives of Microbiology 162:10-19. Sato N, Murata N, Yoshiro M, Ueta N. 1979. Effect of growth temperature on lipid and fatty acid compositions in the blue-green algae, Anabaena variabilis and Anacystis nidulans. Biochimica et Biophysica Acta 572:19-28. Schlüter L, Lohbeck KT, Gutowska MA, Gröger JP, Riebesell U, et al. 2014. Adaptation of a globally important coccolithophore to ocean warming and acidification. Nature Climate Change. 4:1024-1030. Sharma KK, Schuhmann H, Schenk PM. 2012. High lipid induction in microalgae for biodiesel production. Energies 5:1532–1553. Shifrin N, Chisolm SW. 1981. Phytoplankton lipids-Interspecific differences and effects of nitrate, silicate and light-dark cycles. Journal of Phycology 17:374-384. Sinensky M. 1974. Homeoviscous adaptation—A homeostatic process that regulates the viscosity of membrane lipids in Escherichia coli. Proceedings of the National Academy of Sciences of the United States of America 571:522-525. Teoh M, Chu W, Marchant H, Phang S. 2004. Influence of culture temperature on the growth, biochemical composition and fatty acid profiles of six Antarctic microalgae. Journal of Applied Phycology 16:421-430. Thomas MR, Kremer CT, Klausmeier CA, Litchman E. 2012. A global pattern of thermal adaptation in marine phytoplankton. Science 338:1085-1088. Thompson PA, Harrison PJ, Whyte JNC. 1990. Influence of irradiance on the fatty acid composition of phytoplankton. Journal of Phycology 26:278–288. Thompson PA, Guo M, Harrison PJ. 1992a. Effects of variation in temperature. I. On the biochemical composition of eight species of marine phytoplankton. Journal of Phycology 8:481-488. 79 Thompson PA, Guo M, Harrison PJ, Whyte JNC. 1992b. Effects of variation in temperature. II. On the fatty acid composition of eight species of marine phytoplankton. Journal of Phycology 8:488-497. Thrane J-E, Hessen DO, Andersen T. 2017. Plasticity in algal stoichiometry: experimental evidence of a temperature-induced shift in optimal supply N:P ratio. Limnology and Oceanography 62:1346-1354.(cid:1) Tonon T, Sayanova O, Michaelson LV, Qing R, Harvey D, et al. (2005). Fatty acid desaturases from the microalga Thalassiosira pseudonana. FEBS Journal 272:3401-3412. Van Wagenen J, Miller TW, Hobbs S, Hook P, Crowe B, et al. 2012. Effects of light and temperature on fatty acid production in Nannochloropsis salina. Energies 5:731-740. Wada H, Gombos Z, Murata N. 1990. Enhancement of chilling tolerance of a cyanobacterium by genetic manipulation of fatty acid desaturation. Nature 307:200-203. Wada H, Murata N. 1990. Temperature-Induced Changes in the Fatty Acid Composition of the Cyanobacterium, Synechocystis PCC6803. Plant Physiology 92:1062-1069. Wang Z, Benning C. 2011. Arabidopsis thaliana polar glycerolipid profiling by thin layer chromatography (TLC) coupled with gas-liquid chromatography (GLC). Journal of Visualized Experiments. 49:e2518. Williams PJB, Laurens LML. 2010. Microalgae as biodiesel & biomass feedstocks: review & analysis of the biochemistry, energetics & economics. Energy and Environmental Science 3:554-590. Yang L, Chen J, Qin S, Zeng M, Jiang Y, et al. 2018. Growth and lipid accumulation by different nutrients in the microalga Chlamydomonas reinhardtii. Biotechnology for Biofuels 11, 1-12. Zhu CJ, Lee YK, Chao TM. 1997. Effects of temperature and growth phase on lipid and biochemical composition of Isochrysis galbana TK1. Journal of Applied Phycology 9:451– 457. (cid:1)(cid:1) 80 CHAPTER 3: Evolution of C:N:P elemental stoichiometry, cell size and cell morphology in a marine diatom in response to temperature Introduction Temperature is a well-known control on carbon-nitrogen-phosphorus (C:N:P) stoichiometry and cell size in phytoplankton (Montagnes & Franklin 2001; Edwards et al. 2011; Toseland et al. 2013). Competitive abilities for N and P often correlate negatively with cell volume (Edwards et al. 2011) due to biophysical constraints (reduced surface area-to-volume ratio). The same biophysical constraints apply to the uptake of C, the demand for which is higher in larger cells due to the size-dependence of metabolic rate (reviewed in Finkel et al. 2010). Cell size also depends indirectly on temperature, as all biological processes involve the expenditure of nutrients, and all biological rates, including those involved in the uptake of nutrients, depend directly on temperature according to a unimodal, left-skewed function (Ratkowsky et al. 2005; Corkrey et al. 2014). Nutrient-driven selection on cell size and temperature-driven selection on nutrient (C, N, P) uptake and utilization traits may act synergistically to increase or decrease cell size, or they may be act antagonistically, depending on the nutrient and temperature environments a population experiences. As global temperatures rise and cause greater stratification of oceans and lakes, a combination of nutrient depletion and warm temperatures (Sarmiento et al. 2004; Polovina et al. 2008; Weber & Deutch 2012) are predicted to drive a decrease in phytoplankton cell size, through a combination of species turnover and intraspecific morphological change (Hare et al. 2007; Winder et al. 2009; Finkel et al. 2010; Craig et al. 2013). The extent to which the latter will be attributable to rapid evolution in response to warming, rather than simple phenotypic plasticity, is not known. 81 The relationship between temperature and C:N:P stoichiometry of phytoplankton cells is far from settled, though a great deal of research has been done on the subject. Thompson et al. (1992a) estimated C:N ratios in 8 diverse species of phytoplankton at 4 temperatures between 10°C and 25°C and found 8 different relationships, from positive and linear to unimodal to slightly negative. Montagnes & Franklin (2001) found no clear relationship between temperature and either cellular C or N in 8 species of marine diatoms and 2 species of flagellates grown in the lab. On a global scale, however, evidence is mounting that phytoplankton C:P and N:P ratios increase with temperature (Martiny et al. 2013; Toseland et al. 2013; Yvon-Durocher et al. 2015b), which may also be true of terrestrial plants (Reich & Oleksyn 2004). Martiny et al. (2013) Yvon-Durocher et al. (2015b) both showed clear latitudinal patterns in N:P and C:P ratios, which were largely reflective of sea surface temperature. Toseland et al. (2013) showed that translation efficiency in the marine diatom Thalassiosira pseudonana is nearly 3x higher at 20°C than at 11°C, the implication being that at higher temperatures, fewer P-rich ribosomes are needed per unit protein (N-rich) produced in the cell. Toseland and colleagues then used a highly integrative ecosystems model to suggest that warming oceans may lead to a reduced demand for P relative to N and thus higher cellular N:P ratios in phytoplankton, leading to greater N limitation in the surface ocean. In contrast, in a long-term mesocosm warming experiment, Yvon-Durocher et al. (2017) found that warming led to an increase in seston N:P and C:P ratios due to species turnover, though C:N ratios were unaffected; notably, at least one chlorophyte species, Chlamydomonas reinhardtii, showed evidence of local adaptation resulting in N:P and C:P ratios that closely resembled those of the seston. To our knowledge, Yvon-Durocher et al. 82 (2017) is the first study to directly examine rapid, temperature-dependent evolution of N:P stoichiometry in a phytoplankton species, albeit a freshwater one. In phytoplankton and other aquatic ectotherms, one remarkably repeatable trend exists in body size as it relates to temperature (“body size” includes the sizes of single-celled organisms in this context): as temperature increases, body size decreases (Atkinson 1994; Atkinson 1995; Montagnes & Franklin 2001; Atkinson et al. 2003). This relationship has come to be known as the “temperature-size rule” (TSR) (Atkinson 1994; Atkinson et al. 2003). A few notable exceptions to this rule have been reported (Fawley 1984; reviewed in Atkinson 1995), but such cases are rare in the literature. In a meta-analysis of temperature-size data from an eclectic collection of protists, Atkinson et al. (2003) found evidence to support two mechanistic hypotheses for the TSR. 1) Resource limitation: large cells have small surface-are-to-volume ratios (SA:V) compared to smaller cells, due to the 2/3 exponent of SA:V scaling (see Aksnes & Egge 1991). Resource limitation is thus a challenge for large cells, as they must transfer resources across a smaller relative membrane area to support greater metabolic demands (Gilooly et al. 2001), and they may be limited by the diffusion rate CO2; compounding this challenge is the decreased solubility of CO2 in water with increasing temperature. 2) Reproductive rate: as temperature increases, so too does the rate of cell division. At a higher rate of cell division, cells may divide earlier, before they reach their full size (see Jørgensen 1966,1968), thus reducing the mean cell size of the population. A possible mechanism linking the temperature dependence of cellular C:N and cell size is the temperature dependence of photosynthetic efficiency. Chlorophyll a (Chl a) contains 4 N atoms, but rates of C assimilation per Chl a molecule (C/Chl a) can vary greatly in response to temperature—though not always in the same direction (Eppley & Sloan 1966; Jørgensen 1968; 83 Rhee & Gotham 1981; Thompson et al. 1992a). C/Chl a has been shown to be negatively correlated with cell volume (Eppley & Sloan 1966; Thompson 1992a), possibly as a result of reduced surface area for light absorption and C uptake per unit volume. Thus, temperature-driven selection for smaller cell size may increase photosynthetic efficiency, boosting cellular C and therefore C:N and C:P ratios. Diatoms are unique among the phytoplankton in that they are incased in a glass (SiO2) frustule which poses a challenge to exchanging C, N, P and other materials with their environments. Diatoms can only take up nutrients through their “valves”, areas of the frustule consisting mostly of holes (“punctae”). Each diatom frustule has two valves pointing in opposite directions, joined in the middle by the “girdle band”, an impermeable belt of glass. In centric diatoms, the valve faces are circular, giving the frustule a nearly perfectly cylindrical shape in many cases. For physiological purposes, surface area-to-volume ratio in diatoms must be calculated based only on the combined area of the valve faces, excluding the area of the girdle. Evolution and plasticity of the length-width relationship (length-diameter in centric diatoms) thus becomes a problem of geometry. While high temperature overwhelmingly leads to smaller cells in diatoms and most other phytoplankton, diatoms may also combat challenges of resource acquisition—caused by either nutrient depletion or by temperature—by simply adjusting the ratio of valve area to girdle area (Harrison et al. 1977; Joint et al. 1987; Cortese & Gersonde 2007). The aim of this study is to elucidate the effects of 500 generations of experimental evolution at low (16°C) and high (31°C) temperatures on the C:N:P stoichiometry, C assimilation rate per unit chlorophyll a, cell size, and cell morphology of a marine species, the model centric diatom Thalassiosira pseudonana. The replicate T. pseudonana populations used 84 in this study—5 experimentally adapted to 16°C, 5 adapted to 31°C—were shown by O’Donnell et al. (2018) to have diverged markedly in their temperature-dependent population growth functions (thermal reaction norms) after 350 generations of temperature selection. After an additional 150 generations, we explored some of the physiological causes and consequences of thermal adaptation, with an eye toward implications of evolutionary adaptation to warming for persistence of phytoplankton species in the face of global change, and for phytoplankton- mediated biogeochemical change in the sea (Field et al. 1998; Toseland et al. 2013). Materials and Methods We tested the effects of 500 generations of experimental natural selection at 16°C and 31°C on Thalassiosira pseudonana’s thermal reaction norms for carbon (C), nitrogen (N), and phosphorus (P) elemental stoichiometry, chlorophyll a content and frustule morphology. A detailed account of the methods of the long-term selection experiment can be found in O’Donnell et al. (2018). After 500 generations of selection, we grew replicate 16°C-selected populations and 31°C-selected populations (five each) at 10°C, 16°C, 26°C and 31°C—these temperatures represent, respectively, a temperature at the lower extreme of T. pseudonana’s thermal niche, the low-temperature selection environment, a temperature near T. pseudonana’s thermal optimum for per-capita population growth, and the high-temperature selection environment. Temperature assay To acclimate the replicate selection lines (n = 5 for each selection temperature) to the four assay temperatures pre-assay, we transferred 105 cells from each (determined by a CASY particle counter [Schärfe System GmbH, Reutlingen, Germany]) into 4 polystyrene tissue culture 85 flasks (Greiner Bio- One GmbH, Frickenhausen, Germany) containing 40 ml of L1 marine culture medium (Guillard & Hargraves 1993; modified by O’Donnell et al. 2018) and allowed each to grow for 5 generations at its respective assay temperature under cool, white fluorescent grow lights (110 µmol photons m-2 s-1) on a 14:10 light:dark cycle, agitating daily to maintain cells in suspension. We then transferred 105 cells and allowed an additional 5 generations of growth, for a total of 10 generations of temperature acclimation. After acclimation, we transferred 105 cells from each acclimated culture into triplicate tissue culture flasks (2 selection temperatures ´ 5 selection lines ´ 4 assay temperatures ´ 3 replicates = 120 total cultures), each containing 40 ml of L1 medium. We grew experimental cultures until late log-phase, estimating relative change in population daily by optical density, using a spectrophotometer (Shimadzu UV-2401PC; Shimadzu Corporation, Kyoto, Japan) set to 436 nm (the absorbance peak for chlorophyll a [Neori et al., 1986]), with filtered L1 medium serving as the blank. Time to late log-phase ranged from 4 days at 26°C to nearly two weeks at 10°C. Particulate chemical analysis From each experimental culture, we filtered 10 ml through a pre-combusted (4 hours at 480°C) Whatman GF/B glass fiber filter for C and N analysis, 5ml through a second GF/B for P analysis, and 10 ml through a Whatman GF/F glass fiber filter for chlorophyll analysis. GF/B filters containing samples were frozen at -20°C until needed for analysis. We retained the remaining 15 ml for final cell counts (CASY counter) and other analyses not addressed here. We analyzed samples for C and N via combustion gas chromatography using a Costech Elemental 86 Combustion System 4010 (Costech Analytical Technologies, Valencia, CA, USA). A complete protocol for this CN analysis can be found at https://lter.kbs.msu.edu/protocols/148. To analyze filtered cultures for particulate phosphorus, we used the persulfate oxidation method (Menzel & Corwin 1965). To analyze chlorophyll a, we extracted the GF/F filters overnight in 95% ethanol in the dark at 4°C, then read samples at 436 nm in the spectrophotometer described above, using 95% ethanol as a blank. Thalassiosira pseudonana morphometry After GF/B filters were digested for P analysis, many clean T. pseudonana frustules in each sample remained at the bottoms of the glass test tubes, affording an opportunity for morphometric analysis. We transferred each sample from its glass test tube into a 15 ml conical tube, simultaneously filtering the sample through a 62 µ mesh to remove large GF/B fibers. We allowed samples to settle for at least 24h. We then achieved blind sampling by haphazardly drawing each tube from the box, obscuring the label in our fist, and quickly placing the tube upright in an opaque bottle, with only the conical tube’s cap exposed. Keeping the conical tube in the bottle, we drew a ~0.1 ml aliquot from the bottom of the tube using a long, glass Pasteur pipette and placed it on a standard glass microscope slide with a glass coverslip. Only after counting was complete for a given sample did we reveal the tube’s label for data collection. We measured the lengths and widths of 10 T. pseudonana frustules from each sample under 1000x oil immersion on a Nikon Eclipse 80i compound microscope with a Photometrics CoolSNAP EZ camera, and used Image-Pro Plus v4 image processing software (Media Cybernetics, Rockville, MD, USA) to photograph and measure each frustule. T. pseudonana is nearly perfectly cylindrical in shape; viewed laterally, a frustule appears rectangular, with a 87 distinct thinning of its outline indicating the location of the girdle. We measured each outside diameter at the girdle, and each outside length bisecting the frustule up the middle, perpendicular to the diameter measurement. Each slide contained very few frustules, some of which were ill- positioned for measurement; it was thus usually necessary to view multiple aliquots per sample to achieve the 10-cell quota. Sampling was quasi-random, as we simply measured the first 10 measurable frustules encountered. Calculations and statistical analysis Estimates of surface area and volume of T. pseudonana frustules were based on length and diameter measurements (microns) using basic geometry. We first calculated the circumference of the valve face by multiplying the measured diameter by p, and the valve surface area as pr2, where r is the radius (diameter/2). Girdle surface area was calculated as measured frustule length ´ valve circumference. Volume of the cell was calculated as the volume of a cylinder, pr2h, where h (height) is the frustule length. Data used in statistical analyses were averages of the 10 cells measured from each experimental replicate. Models were fit to pooled replicate selection lines by temperature. Eppley (1972) showed that photosynthetic efficiency in phytoplankton can be expressed in terms of particulate C/Chl a, i.e. the amount of carbon (by mass) per unit of chlorophyll a (referred to henceforth as the carbon assimilation number). Furthermore, population growth rate, µ, (doublings per day) can be expressed in terms of the assimilation number per day (DC/Chl a), i.e. the carbon assimilation rate (Eppley 1972, Equation 5): != ' DãXZå1ç/çéè #;Dç/çéè # ç/çéè # , (4) 88 Where Dt is change in time. Rearranging algebraically, if Dt = 1 day, we find that DC/Chl >=C/Chl >×23−C/Chl >. (5) A significant negative correlation between µ and C/Chl a (see Results) led us to substitute a linear equation for C/Chl a as a function of temperature, such that DC/Chl >=(>µ+&)(23−1), (6) Where a is the slope and b is the intercept of C/Chl a as a function of µ. Thus, Equation (6) can be fit to DC/Chl a as a function of specific growth rate, µ (d-1). We did this for 16°C- and 31°C- selected populations separately (pooled assay temperature groups) using nonlinear least squares. We used the function “gnls()” from the R package “nlme” version 3.1-137 (Pinheiro et al. 2018). A preliminary fit to 16°C-selected data revealed severely funneled residuals (increasing variance with increasing µ), which led us to weight the nonlinear regression for 16°C-selected populations using the “varExp()” function to allow for the variance to increase exponentially as a function of µ. The exponential variance function significantly improved the model (dAIC = 24.2; LRT P < 0.0001). All data analyses were conducted using R statistical programming language, version 3.5.0. All stoichiometric ratio data were log-transformed to meet the normality assumptions of analyses. We analyzed stoichiometry and morphometric data using nested versions of a weighted linear mixed-effects model using the “lme()” function from the “nlme” R package. In the full model, selection temperature, assay temperature and selection temperature ´ assay temperature were the fixed effects. We fit a random intercept and slope for each replicate population 89 (selection replicate) across assay temperatures to account for repeated measures (each replicate population was assayed at four different temperatures). We used the “varIdent()” and “varComb()” functions in the nlme package to weight each level of selection temperature and assay temperature by its inverse variance, as unequal variances by factor level violated the assumptions of the linear models in some cases. For each data set, we performed model selection on nested models using likelihood ratio tests; in some cases, temperature ´ assay temperature interaction effects and/or weights did not improve model fits (based on likelihood ratio tests), and were dropped. We derived ANOVA tables from each mixed model, which were in effect repeated-measures ANOVA (RM-ANOVA) because of the random effects. We performed pairwise comparisons of selection temperature and assay temperature groups with Tukey’s Honestly Significant Difference (Tukey HSD), using the “lsmeans()” function from the R package “lsmeans” version 2.27-62 (Lenth 2016). Results Particulate C, N, P and Chlorophyll a Thalassiosira pseudonana populations experimentally evolved at 16°C and 31°C for 500 generations diverged significantly in their thermal reaction norms for chemistry. On a per-cell basis, particulate C in 16°C-selected populations was higher than in 31°C-selected populations, and increased across assay temperatures; In 31°C-selected populations, particulate C per cell increased between the 10°C and 26°C assay temperatures, but was lowest at 31°C (selection temp ´ assay temp interaction; Figure 12; Table 2). An increase in 16°C-selected populations and a decrease in 31°C-selected populations at the 31°C assay temperature resulted in a significant selection temperature ´ assay temperature interaction (Figure 12; Table 2), however there was 90 no main effect of selection temperature or assay temperature on particulate N per cell, but. As with C, particulate P per cell generally increased across assay temperatures, but was lowest in 31°C-selected populations at the 31°C assay temperature (selection temp. ´ assay temp. interaction; Figure 12; Table 2). Trends across assay temperature in chlorophyll a per cell generally mirrored those in Chl a per biovolume, but unlike Chl a per biovolume, Chl a on a per- cell basis was significantly lower in 31°C-selected populations than in 16°C-selected populations (Figure 12; Table 2). (a) (c) (b) (d) Figure 12: Particulate C, N, P (molar) and chlorophyll a (µg) per cell in 16°C-selected (blue) and 31°C-selected (red) T. pseudonana, assayed at 10, 16, 26 and 31°C. (a) Particulate carbon; (b) particulate nitrogen; (c) particulate phosphorus; and (d) particulate chlorophyll. 91 Letters at the top of each panel indicate statistical significance: if a letter is not shared between two groups, those groups are significantly different (Tukey HSD; a = 0.95). Lines across boxes indicate the median of each group, with the lower and upper bounds indicate the 25th and 75th quartiles, respectively. Points beyond the lower and upper whiskers are outliers, falling outside the 0th and 100th quartiles. All data are multiplied by 105. Table 2: RM-ANOVA testing effects of selection temperature and assay temperature on T. pseudonana particulate C, N, P and chlorophyll a per-cell (µmol µm-3) and all molar ratios thereof, and particulate chlorophyll a (µg µm-3). Ratio data were log-transformed to meet the normality assumption of RM-ANOVA. Post-hoc, pairwise comparisons (Tukey HSD) are shown in Appendix D, Tables D1-D4. Trait Particulate C cell-1 Particulate N cell-1 Particulate P cell-1 Partic. Chl a cell-1 Effect Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error Assay temp. Selection temp. Error df 3 1 3 106 3 1 3 106 3 1 3 106 3 1 109 MS 1.42 4.50 2.36 0.30 0.016 0.034 0.078 0.014 5.6´10-5 4.1´10-5 2.8´10-5 7.7´10-6 0.012 0.0021 0.0002 4.72 14.95 7.85 -- 1.11 2.38 5.50 -- 7.21 5.29 3.60 -- F-statistic P-value 0.0039 0.0002 <0.0001 -- 0.35 0.13 0.0015 -- 0.00019 0.023 0.016 -- <0.0001 0.0019 -- 60.09 10.15 -- Trends in C, N, P and chlorophyll a on a per-biovolume basis where markedly different from those estimated per cell. 31°C-selected populations had higher particulate carbon (Figure 13a; Table 2), nitrogen (Figure 13b; Table 3) and phosphorus (Figure 13c; Table 3) per 92 biovolume than 16°C-selected populations at all four assay temperatures (10, 16, 26, and 31°C); however, only the reaction norm for particulate P varied significantly across assay temperatures, with the highest particulate P µm-2 at 26°C in 16°C-selected populations and at 31°C in 31°C- selected populations (Figure 13c; Table 3; see Appendix D, Table D7 for Tukey HSD pairwise comparisons). Chlorophyll a per biovolume was also higher in 31°C-selected populations than in 16°C-selected populations, significantly so at 3 out of 4 assay temperatures (Figure 13d; Table 3). Chl a was lowest at the 10 and 16°C assay temperatures in both selection groups, and increased two- (in 16°C-selected populations) or threefold (31°C-selected populations) between 16°C and 26°C, then decreased by ~50% between 26°C and 31°C in both groups (Figure 13d; Table 3; Tukey pairwise comparisons in Appendix D, Table D8). The resulting thermal reaction norms for Chl a appeared left-skewed and unimodal, resembling the generic thermal reaction norm for any biological rate (see, e.g., Corkrey et al. 2014). 93 (a) (c) (b) (d) Figure 13: Particulate C, N, P (molar) and chlorophyll a (µg) per biovolume in 16°C- selected (blue) and 31°C-selected (red) T. pseudonana, assayed at 10, 16, 26 and 31°C. (a) Particulate carbon; (b) particulate nitrogen; (c) particulate phosphorus; and (d) particulate chlorophyll. Letters at the top of each panel indicate statistical significance: if a letter is not shared between two groups, those groups are significantly different (Tukey HSD; a = 0.95). Lines across boxes indicate the median of each group, with the lower and upper bounds indicate the 25th and 75th quartiles, respectively. Points beyond the lower and upper whiskers are outliers, falling outside the 0th and 100th quartiles. All data are multiplied by 107. 94 Stoichiometric ratios The particulate C:P molar ratio was ~200 in both selection groups; C:N was between 5 and 10; and N:P was between 20 and 50 in both selection groups (Figure 14). Due to several large outliers, there were no detectable effects of selection temperature or assay temperature on stoichiometric ratios (Figure 14; Table 3). a a a a a a a a 0 5 ) r a l o m 0 2 a a a a a a a a a a a ab a abc ac ab ) r a l o m 0 5 ) r a l o m 0 0 5 ( o i t a r P C : 0 0 2 0 0 1 0 5 0 1 5 ( o i t a r N C : 2 (a) a Selection temp.: 10 16 16°C 26 31°C 31 b (b) 10 16 31 Assay temperature (°C) 26 ( o i t a r P N : 0 2 0 1 5 c (c) 10 16 26 31 Figure 14: Particulate molar C:N:P ratios in 16°C-selected (blue) and 31°C-selected (red) T. pseudonana, assayed at 10, 16, 26 and 31°C. All ratios are calculated on a per biovolume (µm3) basis. (a) C:P ratio; (b) C:N ratio; (c) N:P ratio. Symbols are as above. Note that y-axes are on a log scale. Table 3: RM-ANOVA testing effects of selection temperature and assay temperature on T. pseudonana particulate C, N, P per biovolume (µmol µm-3) and all molar ratios thereof, and particulate chlorophyll a (µg µm-3). Ratio data were log-transformed to meet the normality assumption of RM-ANOVA. Post-hoc, pairwise comparisons (Tukey HSD) for particulate C, N, P and Chl a are shown in Appendix D, Tables D5-D7. Trait Particulate C bv-1 Effect Assay temp. Selection temp. Error df 3,12 1,93 109 MS 0.24 28.63 2.00 95 F-statistic P-value 0.95 0.0003 -- 0.121 14.344 -- Table 3 (cont’d) Particulate N bv-1 Particulate P bv-1 Partic. C:P ratio Partic. C:N ratio Partic. N:P ratio Partic. Chl a bv-1 Assay temp. Selection temp. Error Assay temp. Selection temp. Error Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error Assay temp. Selection temp. Error Assay temp. Selection temp. Error Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error 3,12 1,93 109 3,12 1,93 109 3,12 1,90 3,90 106 3,12 1,93 109 3,12 1,93 109 3,12 1,90 3,90 106 0.027 1.33 0.055 0.0003 0.0009 3.1´10-5 0.16 0.12 0.15 0.078 0.12 0.20 0.16 0.27 0.099 0.25 0.022 0.018 0.0016 0.0002 0.49 24.29 -- 3.26 28.52 -- 2.03 1.51 1.98 -- 0.74 1.24 -- 1.08 0.039 -- 115.38 92.45 8.44 -- 0.69 <0.0001 -- 0.024 <0.0001 -- 0.11 0.22 0.12 -- 0.53 0.27 -- 0.36 0.84 -- <0.0001 <0.0001 <0.0001 -- Photosynthetic efficiency Carbon assimilation number (C/Chl a) (g C g Chl a-1) generally decreased by as much as 50% between the 10°C and 31°C assay temperatures, but did not differ between 16°C- and 31°C- selected populations (Figure 15a; Table 4). C/Chl a was negatively correlated with specific growth rate, µ (d-1), in both 16°C-selected (Pearson’s correlation: R = -0.38; t58 = -3.11; P = 0.0029) and 31°C-selected (Pearson’s correlation: R = -0.37; t58 = -3.00; P = 0.0040) populations (Figure 15b). Carbon assimilation rate, that is the rate of photosynthetic C assimilation per unit Chl a per day (DC/Chl a) (g C g Chl a-1 d-1) varied across assay temperature, but did not differ between selection temperature groups (Figure 15c; Table 4). As with particulate Chl a, DC/Chl a was highest at intermediate temperatures and lowest at 10°C and 31°C (Figure 15c; Table 3). Equation 3 explained a significant amount of the variation in DC/Chl a as a function of specific growth rate (d-1) for both temperature selection groups (Figure 15d; Table 5), but 95% 96 confidence intervals for both a and b overlapped between selection groups; the nonlinear regression fits for the two selection groups therefore did not differ significantly (Table 5). (a) (c) (b) (d) Figure 15: (a) Carbon assimilation number, C/Chl a (µg C µg Chl a-1), of 16°C- (blue) and 31°C-selected (red) T. pseudonana populations grown at 10, 16, 26 and 31°C, determined from final C and Chl a samples. Symbols are as above. (b) Carbon assimilation number plotted against specific growth rate (d-1); different shades of blue (16°C-selected) and red (31°C- selected) dots denote assay temperatures. (c) Carbon assimilation rate (µg C µg Chl a-1 d-1) across assay temperatures at the time of harvest. Symbols are as in (a). (d) Carbon assimilation rate as a function of specific growth rate. Curves are Equation (6) fit separately to pooled selection group data using nonlinear least squares. Symbols are as in (b). 97 Table 4: RM-ANOVA testing effects of selection temperature and assay temperature on T. pseudonana carbon assimilation number (C/Chl a) (µg C µg Chl a-1) and assimilation rate (µg C µg Chl a-1 d-1). Pairwise contrasts (Tukey HSD) can be found in Appendix D, Tables D9 and D10. Trait C/Chl a DC/Chl a Effect Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error df 3,12 1,90 3,90 106 3,12 1,90 3,90 106 MS 11.2´105 8.4´104 2.2´105 8.7´104 1.7´105 7894 3.3´104 2.1´104 F-statistic 12.94 0.97 2.53 -- 8.51 0.40 1.69 -- P-value <0.0001 0.33 0.061 -- <0.0001 0.53 0.18 -- Table 5: Nonlinear regression fits of Equation (6) to carbon assimilation rate per unit Chl a (DC/Chl a) (µg C µg Chl a-1 d-1) as a function of specific growth rate (d-1). Equation (6) was fit separately to data from each selection group, pooled across assay temperatures. The regression fit to data from 16°C-selected was weighted using an exponential variance function, to account for heteroscedasticity. Model 16°C-selected* df Model P-val. 58 <0.0001 31°C-selected <0.0001 *Variance function exponent = 3.43 58 R2 0.29 0.26 RSE 11.96 176.6 Parameter a b a b Value 95% CI -374.14 744.11 (-1055.92— -274.05) (662.3223—1154.4621) -244.9 671.5 (-562.13—72.27) (386.36—956.61) Thalassiosira pseudonana frustule size and morphology T. pseudonana frustule size and morphology were strongly affected by both selection temperature (16°C vs 31°C) and assay temperature (10, 16, 26, 31°C). Both frustule diameter 98 (measured across the circular valve face) and frustule length were ~25% larger in 16°C-selected populations than in 31°C-selected populations (Figure 16; Figure 17a,b; Table 6); frustule volumes differed between selection groups by a factor of 2-3 (Figure 16; Figure 17c; Table 6). Frustule diameter and frustule length also increased across assay temperatures in both selection groups (Figure 16; Figure 17a,b; Table 6), though frustule volume was only marginally affected by assay temperature (F3,115 = 2.26; P = 0.085; Figure 16; Figure 17c; Table 6). a f 10°C 10°C 16°C 16°C 26°C 26°C 31°C 31°C b g c h d i Figure 16: 16°C-selected (blue boxes) and 31°C-selected (red boxes) Thalassiosira pseudonana frustules photographed in lateral view under 1000x oil immersion. Frustules are from (a,f) 10°C; (b,g) 16°C; (c,h) 26°C and (d,i) 31°C assay temperatures. These images were chosen as exceptional representations of the statistical trends observed in length and diameter, and do not necessarily represent averages. Images have been sharpened and brightened slightly to improve quality but are otherwise unchanged. 99 (a) (c) (b) Figure 17: Frustule dimensions in 16°C (blue) and 31°C (red) selected T. pseudonana populations grown at 10, 16, 26 and 31°C. (a) Frustule diameter (µm); (b) frustule length (µm); (c) frustule volume (µm3). Symbols are as above. Total surface area-to-volume ratio (SAtot:V) was ~20% higher in 31°C-selected populations than in 16°C-selected populations but was unaffected by assay temperature (Figure 16; Figure 18a; Table 6). The ratio of valve surface area to frustule volume (SAvalve:V) was >25% greater in 31°C-selected lines than in 16°C-selected lines, with the greatest difference at the 10°C and 31°C assay temperatures (Figure 16; Figure 18b; Table 6; see Appendix D, Table D15 for Tukey HSD pairwise comparisons). The ratio of girdle surface area to frustule volume (SAgirdle:V) was also as much as 25% greater in 31°C-selected populations than in 16°C-selected populations, and decreased by ~15% between the 16°C and 31°C assay temperatures in both groups (Figure 16; Figure 18c; Table 5; Tukey pairwise comparisons in Appendix D, Table D16). SAvalve:SAgirdle did not differ between selection groups, but was as much as 40% higher at the 26°C and 31°C assay temperatures than at 10°C and 16°C (Figure 16; Figure 18d; Table 6). 100 (a) (c) (b) (d) Figure 18: T. pseudonana whole frustule (a), valve only (b), and girdle only (c) surface- area-to-volume ratios (SA:V), and (d) valve SA:girdle SA in populations selected at 16°C (blue) and 31°C (red) and grown at 10, 16, 26 and 31°C. Symbols are as above. Surface area is in units of µm2, and volume is in µm3. 101 Table 6: RM-ANOVA testing effects of selection temperature and assay temperature on T. pseudonana frustule diameter, frustule length, frustule volume, valve surface area-to- frustule volume ratio, girdle surface area-to-frustule volume ratio, and the ratio of valve surface area to girdle surface area. Linear measurements are in units of µm, areas are µm2, volumes are µm3. Post-hoc, pairwise comparisons (Tukey HSD) are shown in Appendix D, Tables D11-D17. Effect Assay temp. Selection temp. Error Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error Assay temp. Selection temp. Error Trait Frustule diameter Frustule length Frustule volume SAtotal : Volume ratio Assay temp. SAvalve : Volume ratio Assay temp. SAgirdle : Volume ratio Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error Selection temp. Error Selection temp. Error SAvalve : SAgirdle ratio Assay temp. Selection temp. Assay temp. ´ Sel. Temp. Error df 3,12 1,99 115 3,96 1,96 3,96 112 3,12 1,99 115 3,12 1,99 115 3,12 1,96 3,96 112 3,12 1,99 115 3,12 1,96 3,96 112 MS 3.47 19.76 0.23 3.71 14.32 0.35 0.141 582 40871 257 0.0038 1.70 0.0086 0.0038 1.70 0.015 0.0085 0.0.096 0.61 0.0073 0.35 0.0026 0.011 0.0090 15.03 85.58 -- 0.439 196.67 26.37 101.81 2.51 -- 2.26 158.72 F-statistic P-value <0.0001 <0.0001 -- <0.0001 <0.0001 0.062 -- 0.085 <0.0001 -- 0.73 <0.0001 -- 0.72 <0.0001 0.16 -- <0.0001 <0.0001 -- <0.0001 0.60 0.32 -- -- 0.45 200.49 1.75 -- 38.44 0.28 1.19 -- 13.11 84.13 -- -- When regressed against frustule volume, the slopes of total surface are-to-volume ratio (SAtotal:V), SAvalve:V and SAgirdle:V differed between selection groups and among assay temperatures (Figure 19; Table 7). While similar at 10°C and 16°C, the slopes of SA:Vtotal in 102 16°C-selected populations were more strongly negative than those of 31°C-selected populations at 26°C and 31°C (Figure 19a; Table 7). Slopes of SAvalva:V as a function of frustule volume were also smaller (by value, not magnitude) at all temperatures in 16°C-selected populations at all assay temperatures, but generally increased across assay temperatures in both groups (Figure 19b; Table 7). The effects of selection temperature, assay temperature, and frustule volume on SA:Vgirdle were exactly opposite those of SAvalve:V (Figure 19c; Table 7)—these effects offset one another to produce the trends in SAtotal:V. ) e m u l o V g o l ~ V : t o t A S g o l ( e p o l S 4 2 . 0 − 8 2 . 0 − 2 3 . 0 − 6 3 . 0 − (a) 10 16 26 ) e m u l o V g o l ~ V : e v l a v A S g o l ( e p o l 2 . 0 0 . 0 2 . 0 − 4 . 0 31 − S ) e m u l o V g o l ~ V 3 . 0 − 4 . 0 − : e l d r i g 5 . 0 − A S g o l ( e p o l 6 . 0 − 7 . 0 31 − S (b) 10 26 Assay temperature (°C) 16 (c) 10 16 26 31 Figure 19: Slopes of (a) total surface area-to-volume ratio (SAtot:V) of the frustule; (b) SAvalve:V; and (c) SAgirdle:V as functions of frustule volume at the 10, 16, 26 and 31°C assay temperatures. Blue shades are 16°C-selected populations, red shades are 31°C-selected populations. Points are slopes estimated from the linear mixed-effects model fits. Error bars are 95% confidence intervals, also estimated from the model fits. Non-overlapping confidence intervals indicate significant differences. Areas are in units of µm2, volumes are in units of µm3. The units of SA:V are µm2/3. 103 Table 7: RM-ANOVA testing effects of selection temperature, assay temperature and frustule volume on ratios of T. pseudonana total frustule surface area, valve surface area, girdle surface area to frustule volume, and the ratio of valve surface area to girdle surface area. Linear measurements are in units of µm, areas are µm2, volumes are µm3. Trait SAtotal:V SAvalve:V SAgirdle:V* df MS Effect 1,104 1.43 log Volume 3,104 0.0029 Assay temp. 1,104 0.0009 Selection temp. 3,104 <0.0001 log V ´ Assay temp. 0.001 1,104 log V ´ Sel temp. 3,104 0.0003 Assay temp. ´ Sel. temp. 0.0002 Log V ´ A. temp. ´ S. temp. 3,104 0.0001 104 Error 0.80 1,104 log Volume Assay temp. 3,104 0.34 0.11 1,104 Selection temp. 0.016 3,104 log V ´ Assay temp. 1,104 0.073 log V ´ Sel temp. 0.026 3,104 Assay temp. ´ Sel. temp. 0.0048 Log V ´ A. temp. ´ S. temp. 3,104 0.0071 104 Error log Volume 1,104 1.90 Assay temp. 0.084 3,104 Selection temp. 0.027 1,104 3,104 0.0040 0.018 1,104 Assay temp. ´ Sel. temp. 0.0066 3,104 Log V ´ A. temp. ´ S. temp. 3,104 0.0012 Error 0.0018 104 F-statistic P-value <0.0001 <0.0001 0.0005 0.91 0.0002 0.0079 0.039 -- <0.0001 <0.0001 0.0002 0.091 0.0018 0.014 0.57 -- <0.0001 <0.0001 0.0002 0.091 0.0018 0.014 0.57 -- *F-statistics and P-values are the same because SAvalve:V and SAgirdle:V have ofsetting effects on SAtot:V. log V ´ Assay temp. log V ´ Sel temp. 21529 43.87 12.97 0.175 14.85 4.17 2.88 -- 112.129 47.21 15.15 2.21 10.19 3.70 0.68 -- 1066.01 47.21 15.15 2.21 10.19 3.70 0.68 -- 104 Discussion We maintained replicate populations of the marine diatom Thalassiosira pseudonana at 16°C and 31°C for 500 generations, which led to major evolutionary divergence in cellular elemental stoichiometry, photosynthetic efficiency and frustule morphology. We showed that, in addition to being smaller in size, 31°C-selected populations were denser in C, N, P and chlorophyll at all assay temperatures, though we were unable to detect differences between selection groups or trends across assay temperatures in C:N:P elemental ratios. Finally, T. pseudonana evolved at 31°C exhibited thermal reaction norms for morphological traits distinct from those of populations evolved at 16°C. Much attention has been given to warming-driven changes in global phytoplankton community assemblages in recent years (Hare et al. 2007; Winder et al. 2009; Finkel et al. 2010; Thomas et al. 2012), which may in turn lead to phytoplankton-mediated biogeochemical change on a global scale (Litchman et al. 2006; Weber et al. 2012; Craig et al. 2013; Toseland et al. 2013). However, we are only beginning to understand the role rapid thermal adaptation will play in this scenario. Our study adds a number of valuable insights to the growing body of research on evolutionary changes in phytoplankton metabolic, stoichiometric, morphological and other physiological traits (Schlüter et al. 2014; Padfield et al. 2016; Yvon-Durocher et al. 2017). For example, we show that the temperature dependence of cellular C, N and P contents and elemental ratios in phytoplankton may depend to a great extent on their evolutionary history, i.e. what temperature regimes a population has experienced, even, or perhaps especially in very recent evolutionary history. Schlüter et al. (2014) found that after ~460 generations of evolution at an elevated temperature (26.3°C), populations of the coccolithophore Emiliana huxleyi had lower particulate organic carbon (POC) per cell than those evolved at a lower temperature 105 (15°C), especially when assayed at the elevated temperature. Likewise, we observed lower particulate C per cell in T. pseudonana populations evolved at 31°C than in those evolved at 16°C, with the greatest difference observed at the 31°C assay temperature. However, in both Schlüter et al. (2014) and our study, lower C per cell was likely simply a result of smaller cell size. When controlled for cell size, particulate C per biovolume was higher in 31°C-selected T. pseudonana populations, indicating a higher C density per cell. Padfield et al. (2016) showed strong evidence that evolution of carbon use efficiency (CUE) (respiration/photosynthesis) could explain evolutionary adaptation to temperature in phytoplankton. They found evidence to support the hypothesis that respiration increases with temperature faster than photosynthesis, but that the difference between these two slopes could be reduced through thermal adaptation. While we did not test CUE directly, we showed that C assimilation rates per unit chlorophyll a (µg C µg chl a-1 d-1) were similar between 16°C- and 31°C-selected populations (pooled across assay temperatures), despite differences in C density. This discrepancy can likely be explained by higher chl a per biovolume in 31°C-selected populations, especially at the higher assay temperatures. Warm-adapted T. pseudonana may have increased chl a production per unit of biovolume produced, boosting photosynthesis relative to respiration and leading to an accumulation of C in the cell. This rapid accumulation of C may have facilitated earlier, more rapid cell division, contributing to selection for smaller cell sizes in the 31°C-selected populations (Jørgensen 1966,1968), which in turn would facilitate C uptake, as well as uptake of N and P (Atkinson et al. 2003). Denser N and P in 31°C-selected populations may simply be a result of a total increase in resource requirements driven by higher rates of photosynthesis and reproduction in 31°C-selected populations. High-temperature selection may have resulted in constitutive early cell division, as 31°C-selected cells were smaller at all four 106 assay temperatures. It is unclear, however, how 31°C-selected cells achieved higher nutrient densities at 10°C and 16°C, rather than at only the two highest assay temperatures (26°C and 31°C). To date, most research on global patterns in phytoplankton stoichiometry have indicated that both N:P and C:P increase with temperature (Martiny et al. 2013; Toseland et al. 2013; Yvon-Durocher et al. 2015b), and at least one mesocosm experiment has produced similar results (Yvon-Durocher et al. 2017). As in some other laboratory-based studies (Thompson et al. 1992b; Montagnes & Franklin 2001), we did not detect any influence of selection temperature or assay temperature on N:P molar ratios in our cultures, and large outliers prevented us from detecting any trends in C:P or C:N ratios. Notably, C:N ratio did not vary with temperature in either the global (Yvon-Durocher et al. 2015b) or mesocosm (Yvon-Durocher et al. 2017) studies that measured it. Given our inability to detect any influence of temperature on N:P ratios, we are unable to offer evidence for or against the mechanistic hypothesis proposed by Toseland et al. (2013) (higher translation efficiency leading to N limitation). T. pseudonana showed not only remarkable plasticity in its ability to respond to ambient temperature by altering the ratio of permeable valve surface area to nonpermeable girdle surface area via phenotypic plasticity, but also a marked divergence in the degree of this plasticity (that is, in the slopes of its thermal reaction norms for morphology) between 16°C- and 31°C-selected populations. Both selection groups allocated more surface area to the valve faces when grown at 26°C and 31°C without appreciably changing overall frustule volume or total surface area-to- volume ratio. This ability is strong evidence that, rather than change its cell size in response to rapid changes in temperature, T. pseudonana simply alters its geometry to compensate for the biophysical challenges of existing at high temperature (Harrison et al. 1977; Joint et al. 1987; 107 Cortese & Gersonde 2007; see also Atkinson et al. 2003). Despite the maintenance of frustule volume across assay temperatures, the effect of volume on SAvalve:V was relatively greater, and that on SAgirdle:V was relatively smaller in 16°C-selected populations. The difference in the magnitudes of these effects can be attributed to the relative scaling of the area of the circular ) versus the rectangular girdle ([ó∙.ï#ïñ]∙XôÉö/ñ). Whether maintenance valve face (p∙Sï#ïñ 1 of frustule volume is hard-coded in the genome or whether it is incidental is not clear. Support for both hypotheses for the TSR proposed by Atkinson et al. (2003) can be found here. First, the smaller overall size of 31°C-selected cells compared to 16°C-selected cells, combined with the higher nutrient density in the former, supports the hypothesis that rapid growth selects for earlier cell division, which decreases the average cell size of the population. This can be viewed as an evolutionary effect. In the short term, T. pseudonana is plastic with respect to its allocation of surface area to valves versus girdles, suggesting resource limitation at 26°C and 31°C, most probably by CO2, given the N- and P-saturated culture medium. This study is a valuable addition to two areas of knowledge about the ecology, evolution, and physiology of phytoplankton. First, we have shown that elemental stoichiometry, cell size, and thermal reaction norms for a set of critical physiological traits (cell morphological traits) in a marine diatom can evolve rapidly, over a few hundred generations, in response to a sustained temperature environment. Studies of this kind are few, and to our knowledge none to date has been done on a marine diatom (but see Schlüter et al. 2014; Padfield et al. 2016; Yvon-Durocher et al. 2017). Second, we have shown that a diatom can cope with the biophysical challenges of rapid temperature change by plastically altering its geometry without changing its size. To our knowledge, this has not been shown before in this context. Follow-up studies of the genetic 108 controls of cell size and morphology in T. pseudonana would speak volumes to this system, and may be well within reach (Armbrust et al. 2004). 109 APPENDICES 110 APPENDIX C: Chapters 2 & 3 description of linear mixed-effects models 111 To test effects of selection temperature and assay temperature on fatty acid content and composition, we fit weighted linear mixed-effects models to data for total fatty acid content (TFA), percent saturated FA (pSFA), percent monounsaturated FA (pMUFA), percent polyunsaturated FA (pPUFA), mean chain length (MCL), and degree of unsaturation (WUnSat), and for each individual FA class, with selection temperature and assay temperature as fixed effects, and a random intercept and slope (across assay temperatures) for each replicate population to account for repeated measures across assay temperatures. Unequal variances between selection temperature groups and among assay temperature groups were common, so we included two identity variance functions in the full model using the “varIdent()” and “varComb()” functions from the R package “nlme” (Pinheiro & Bates 2018). Some models were not improved by the inclusion of these variance functions, and some were equally good with and without a selection temperature ´ assay temperature interaction effect. Thus, we compared log likelihoods of four nested models to determine whether variance functions and/or interaction terms could be dropped in each case. The four nested models used are as follows: Model C1: full model ¢'≠0 ¢1≠0 V0Éú~<ûV0ú,E ü†1°, V'Éú~<ûV'ú,E üJ1°, ù'~<û0,G(á)°, ù1~<û0,G(W)°, õ ~ V0Éú + V0Éú+V'á+ V1W+ VgáW+ù, G(á)=¢0J+¢'á, G(W)=¢0*+¢1W, õ ~ V0Éú + V0Éú+V'á+ V1W+ VgáW+ù, V0Éú~<ûV0ú,E ü†1°, V'Éú~<ûV'ú,E üJ1° Model C2: interaction included, weights dropped 112 Model C3: weights included, interaction dropped V0Éú~<ûV0ú,E ü†1°, V'Éú~<ûV'ú,E üJ1°, ù'~<û0,G(á)°, ù1~<û0,G(W)°, õ ~ V0Éú + V0Éú+V'á+ V1W+ù, G(á)=¢0J+¢'á, ¢'≠0 G(W)=¢0*+¢1W, ¢1≠0 õ ~ V0Éú + V0Éú+V'á+ V1W+ù, V0Éú~<ûV0ú,E ü†1°, V'Éú~<ûV'ú,E üJ1° Model C4: weights dropped, interaction dropped 113 APPENDIX D: Chapter 3 Tukey HSD contrast tables 114 Tukey HSD post-hoc tests The following tables show results of post-hoc, pairwise comparisons (Tukey HSD) of stoichiometric and morphometric traits among all factors/factor levels for T. pseudonana in a factorial (selection temperature ´ assay temperature) assay first analyzed with RM-ANOVA (see Tables 2 and 3 in main text). Tables correspond, in order, to individual panels in main text Figures 12-15 and 17-19. Table D1: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate C per cell (µmol cell-1) between temperature selection groups and among assay temperature groups. Table corresponds to Figure 12a. t-ratio P-value Trait compared Particulate C Contrast Selection temp.@ Assay temp. 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Estimate -0.06 -0.06 0.12 -0.36 -0.11 -1.05 0.16 0.00 0.18 -0.29 -0.05 -0.99 0.22 0.18 -0.30 -0.05 -1.00 0.22 -0.47 -0.23 -1.17 0.04 0.24 -0.70 0.51 -0.94 0.27 1.21 Std. Error 0.23 0.21 0.24 0.20 0.23 0.21 0.27 0.28 0.31 0.28 0.30 0.29 0.33 0.15 0.19 0.22 0.20 0.27 0.23 0.25 0.24 0.29 0.13 0.20 0.26 0.23 0.29 0.21 df 90 12 12 12 12 12 12 12 12 12 12 12 12 90 12 12 12 12 12 12 12 12 90 12 12 12 12 90 115 -0.28 -0.29 0.49 -1.74 -0.48 -4.93 0.58 0.02 0.59 -1.05 -0.16 -3.47 0.67 1.19 -1.53 -0.23 -4.91 0.82 -2.08 -0.91 -4.96 0.14 1.84 -3.47 1.94 -4.13 0.94 5.82 1.00 1.00 1.00 0.67 1.00 0.006 1.00 1.00 1.00 0.96 1.00 0.06 1.00 0.93 0.78 1.00 0.006 0.99 0.48 0.98 0.01 1.00 0.59 0.063 0.55 0.022 0.98 <0.0001 Table D2: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate N per cell (µmol cell-1) between temperature selection groups and among assay temperature groups. Table corresponds to Figure 13b. t-ratio P-value -1.77 -1.16 -0.76 -0.86 -0.84 -3.21 0.12 0.10 0.51 0.38 0.40 -1.92 1.36 0.58 0.28 0.30 -2.05 1.27 -0.12 -0.10 -2.45 0.87 0.03 -2.28 0.97 -2.31 0.95 4.64 0.64 0.93 0.99 0.98 0.99 0.10 1.00 1.00 1.00 1.00 1.00 0.56 0.86 1.00 1.00 1.00 0.49 0.89 1.00 1.00 0.30 0.98 1.00 0.37 0.97 0.36 0.97 0.0003 Contrast Selection temp.@ Assay temp. 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Trait compared Particulate N Estimate -0.068 -0.062 -0.040 -0.047 -0.046 -0.171 0.006 0.006 0.027 0.021 0.022 -0.104 0.074 0.022 0.015 0.016 -0.109 0.068 -0.007 -0.005 -0.131 0.047 0.001 -0.124 0.053 -0.126 0.052 0.178 Std. Error 0.038 0.053 0.053 0.054 0.054 0.053 0.054 0.054 0.054 0.055 0.055 0.054 0.054 0.038 0.054 0.054 0.053 0.054 0.054 0.054 0.053 0.054 0.040 0.054 0.055 0.054 0.055 0.038 df 90 12 12 12 12 12 12 12 12 12 12 12 12 90 12 12 12 12 12 12 12 12 90 12 12 12 12 90 116 Table D3: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate P per cell (µmol cell-1) between temperature selection groups and among assay temperature groups. Table corresponds to Figure 12c. Estimate Std. Error t-ratio P-value 1E-04 -5E-04 2E-04 -2E-03 -2E-03 -4E-03 5E-04 -6E-04 6E-05 -3E-03 -2E-03 -4E-03 4E-04 7E-04 -2E-03 -2E-03 -3E-03 1E-03 -3E-03 -3E-03 -4E-03 3E-04 1E-04 -1E-03 3E-03 -2E-03 3E-03 4E-03 117 0.0006 0.0009 0.0010 0.0011 0.0012 0.0011 0.0012 0.0010 0.0010 0.0011 0.0013 0.0011 0.0012 0.0007 0.0011 0.0012 0.0011 0.0012 0.0012 0.0013 0.0011 0.0012 0.0012 0.0012 0.0013 0.0013 0.0014 0.0011 df 96 12 12 12 12 12 12 12 12 12 12 12 12 96 12 12 12 12 12 12 12 12 96 12 12 12 12 96 0.16 -0.53 0.17 -2.26 -1.92 -3.73 0.41 -0.62 0.06 -2.28 -1.96 -3.70 0.31 0.98 -1.80 -1.51 -3.24 0.83 -2.31 -1.98 -3.70 0.26 0.12 -1.21 2.25 -1.22 1.99 4.19 1.00 1.00 1.00 0.38 0.56 0.042 1.00 1.00 1.00 0.37 0.54 0.044 1.00 0.98 0.63 0.79 0.09 0.99 0.36 0.53 0.044 1.00 1.00 0.92 0.39 0.91 0.52 0.0015 Contrast Selection temp.@ Assay temp. 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Trait compared Particulate P Table D4: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate chlorophyll a per cell (µg cell-1) between temperature selection groups and among assay temperature groups. Table corresponds to Figure 12d. Trait compared Chlorophyll a Estimate Std. Error t-ratio P-value Contrast Selection temp.@ Assay temp. 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 0.006 0.000 0.006 -0.041 -0.035 -0.029 -0.022 -0.006 0.000 -0.047 -0.041 -0.035 -0.029 0.006 -0.041 -0.034 -0.028 -0.022 -0.047 -0.041 -0.035 -0.028 0.006 0.012 0.019 0.006 0.012 0.006 0.0019 0.0027 0.0033 0.0034 0.0039 0.0043 0.0047 0.0033 0.0027 0.0039 0.0034 0.0047 0.0043 0.0019 0.0032 0.0037 0.0042 0.0046 0.0037 0.0032 0.0046 0.0042 0.0019 0.0046 0.0050 0.0050 0.0046 0.0019 df 93 12 12 12 12 12 12 12 12 12 12 12 12 93 12 12 12 12 12 12 12 12 93 12 12 12 12 93 3.30 -0.05 1.91 -12.11 -8.93 -6.60 -4.70 -1.96 -0.05 -12.11 -12.11 -7.38 -6.60 3.30 -12.94 -9.33 -6.85 -4.82 -12.77 -12.94 -7.63 -6.85 3.30 2.66 3.71 1.20 2.66 3.30 0.029 1.00 0.57 <0.0001 <0.0001 0.0005 0.009 0.54 1.00 <0.0001 <0.0001 0.0002 0.0005 0.029 <0.0001 <0.0001 0.0003 0.0071 <0.0001 <0.0001 0.0001 0.0003 0.029 0.22 0.043 0.92 0.22 0.029 118 Table D5: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate C per biovolume (µmol µm-3) between temperature selection groups and among assay temperature groups. Table corresponds to Figure 13a. t-ratio P-value -4.51 -0.47 -2.70 -1.36 -3.46 -1.83 -3.81 1.88 -0.47 1.11 -1.36 0.63 -1.83 -4.51 -0.93 -3.16 -1.44 -3.53 1.57 -0.93 1.05 -1.44 -4.51 -0.54 -2.75 1.82 -0.54 -4.51 0.0005 1.00 0.21 0.86 0.065 0.61 0.0365 0.58 1.00 0.94 0.86 1.00 0.61 0.0005 0.98 0.11 0.82 0.058 0.76 0.98 0.96 0.82 0.0005 1.00 0.20 0.62 1.00 0.0005 Contrast Selection temp.@ Assay temp. 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Trait compared Particulate C Estimate -1.04 -0.18 -1.22 -0.53 -1.57 -0.74 -1.78 0.85 -0.18 0.50 -0.53 0.29 -0.74 -1.04 -0.35 -1.38 -0.56 -1.60 0.69 -0.35 0.48 -0.56 -1.04 -0.21 -1.25 0.82 -0.21 -1.04 Std. Error 0.23 0.39 0.45 0.39 0.45 0.41 0.47 0.45 0.39 0.45 0.39 0.47 0.41 0.23 0.37 0.44 0.39 0.45 0.44 0.37 0.45 0.39 0.23 0.39 0.45 0.45 0.39 0.23 df 93 12 12 12 12 12 12 12 12 12 12 12 12 93 12 12 12 12 12 12 12 12 93 12 12 12 12 93 119 Table D6: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate N per biovolume (µmol µm-3) between temperature selection groups and among assay temperature groups. Table corresponds to Figure 13b. t-ratio P-value <0.0001 <0.0001 0.88 0.024 0.94 0.031 0.70 0.015 0.67 0.88 0.55 0.94 0.85 0.70 1.00 0.22 1.00 0.11 0.13 1.00 0.30 1.00 1.00 0.08 0.36 1.00 -6.17 -1.32 -4.07 -1.12 -3.92 -1.67 -4.34 1.74 -1.32 1.94 -1.12 1.37 -1.67 -6.17 0.20 -2.68 -0.37 -3.12 3.04 0.20 2.46 -0.37 -6.17 -0.58 -3.32 2.31 -0.58 -6.17 <0.0001 <0.0001 Contrast Selection temp.@ Assay temp. 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Trait compared Particulate N Std. Error 0.041 0.077 0.087 0.076 0.086 0.079 0.089 0.087 0.077 0.087 0.076 0.089 0.079 0.041 0.078 0.088 0.081 0.091 0.089 0.078 0.091 0.081 0.041 0.080 0.090 0.090 0.080 0.041 df 93 12 12 12 12 12 12 12 12 12 12 12 12 93 12 12 12 12 12 12 12 12 93 12 12 12 12 93 Estimate -0.25 -0.10 -0.35 -0.09 -0.34 -0.13 -0.38 0.15 -0.10 0.17 -0.09 0.12 -0.13 -0.25 0.02 -0.24 -0.03 -0.28 0.27 0.02 0.22 -0.03 -0.25 -0.05 -0.30 0.21 -0.05 -0.25 120 Table D7: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate P per biovolume (µmol µm-3) between temperature selection groups and among assay temperature groups. Table corresponds to Figure 13c. Estimate Std. Error t-ratio P-value <0.0001 -0.0055 -0.0012 -0.0067 -0.0040 -0.0095 -0.0025 -0.0079 0.0043 -0.0012 0.0015 -0.0040 0.0030 -0.0025 -0.0055 -0.0028 -0.0083 -0.0013 -0.0067 0.0027 -0.0028 0.0042 -0.0013 -0.0055 0.0015 -0.0040 0.0070 0.0015 -0.0055 0.0008 0.0018 0.0020 0.0020 0.0021 0.0019 0.0021 0.0020 0.0018 0.0021 0.0020 0.0021 0.0019 0.0008 0.0020 0.0021 0.0019 0.0021 0.0021 0.0020 0.0021 0.0019 0.0008 0.0021 0.0022 0.0022 0.0021 0.0008 df 93 12 12 12 12 12 12 12 12 12 12 12 12 93 12 12 12 12 12 12 12 12 93 12 12 12 12 93 -7.05 -0.65 -3.34 -2.02 -4.47 -1.27 -3.79 2.13 -0.65 0.70 -2.02 1.44 -1.27 -7.05 -1.42 -3.90 -0.65 -3.22 1.27 -1.42 2.01 -0.65 -7.05 0.74 -1.79 3.18 0.74 -7.05 <0.0001 1.00 0.08 0.51 0.012 0.90 0.038 0.45 1.00 1.00 0.51 0.82 0.90 0.83 0.031 1.00 0.10 0.89 0.83 0.51 1.00 0.99 0.64 0.10 0.99 <0.0001 <0.0001 Contrast Selection temp.@ Assay temp. 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Trait compared Particulate P 121 Table D8: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana particulate chlorophyll a per biovolume (µg µm-3) between temperature selection groups and among assay temperature groups. Table corresponds to Figure 13d. Trait compared Chlorophyll a Selection temp.@ Assay temp. Contrast Estimate Std. Error t-ratio P-value 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 0.0047 0.0036 0.0042 0.0042 0.0052 0.0043 0.0056 0.0046 0.0051 0.0051 0.0060 0.0052 0.0064 0.0036 0.0039 0.0050 0.0040 0.0054 0.0044 0.0055 0.0046 0.0058 0.0053 0.0045 0.0058 0.0056 0.0066 0.0057 df 90 12 12 12 12 12 12 12 12 12 12 12 12 90 12 12 12 12 12 12 12 12 90 12 12 12 12 90 -2.18 -0.26 -4.01 -10.27 -16.26 -6.33 -10.86 1.99 -1.29 -6.39 -12.49 -3.27 -8.02 -4.40 -10.84 -16.84 -6.56 -11.13 -5.86 -12.54 -2.28 -7.60 -8.10 3.43 -3.15 10.46 3.66 -5.91 0.37 1.00 0.026 <0.0001 <0.0001 0.0007 <0.0001 0.52 0.89 0.0006 <0.0001 0.09 0.0001 0.0008 <0.0001 <0.0001 0.0005 <0.0001 0.0014 <0.0001 0.37 0.0001 <0.0001 0.07 0.11 <0.0001 0.047 <0.0001 -0.010 -0.001 -0.017 -0.043 -0.085 -0.027 -0.061 0.009 -0.007 -0.033 -0.075 -0.017 -0.051 -0.016 -0.042 -0.084 -0.026 -0.060 -0.026 -0.069 -0.010 -0.044 -0.043 0.016 -0.018 0.058 0.024 -0.034 122 Table D9: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana carbon assimilation number (µg C µg Chl a-1). Table corresponds to Figure 15a. Trait compared Contrast C/Chl a Selection temp.@ Assay temp. t-ratio P-value -2.05 -0.30 -1.16 2.96 3.09 1.42 2.67 1.66 0.60 4.09 4.18 3.04 3.89 -1.01 3.32 3.42 1.78 3.00 3.28 3.40 2.23 3.11 0.65 -2.07 -0.08 -2.25 -0.44 2.58 0.45 1.00 0.93 0.14 0.12 0.83 0.22 0.71 1.00 0.023 0.020 0.13 0.032 0.97 0.081 0.069 0.64 0.14 0.087 0.072 0.40 0.11 1.00 0.48 1.00 0.39 1.00 0.18 Estimate -256.6 -30.9 -157.8 248.4 277.3 119.9 243.0 225.7 98.8 505.0 533.9 376.5 499.6 -126.9 279.3 308.3 150.8 273.9 406.2 435.1 277.7 400.8 28.9 -128.5 -5.4 -157.5 -34.3 123.1 Std. Error 125.0 101.6 136.5 83.9 89.9 84.4 91.2 136.2 163.8 123.5 127.6 123.8 128.6 125.7 84.2 90.1 84.6 91.4 124.0 128.1 124.3 129.1 44.3 62.2 71.2 70.0 78.1 47.8 df 96 12 12 12 12 12 12 12 12 12 12 12 12 96 12 12 12 12 12 12 12 12 96 12 12 12 12 96 123 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Table D10: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana carbon assimilation rate (µg C µg Chl a-1 d-1). Table corresponds to Figure 15c. ∆|/|ℎX > Selection temp.@ Assay temp. Trait compared Contrast t-ratio P-value 0.23 -3.40 -2.30 -2.61 -3.93 -3.38 -1.17 -3.30 -2.31 -2.51 -3.86 -3.19 -1.19 -0.50 1.15 -1.07 0.78 1.97 1.12 -0.27 0.91 1.63 -2.52 -0.50 1.04 1.83 2.75 2.29 1.00 0.072 0.37 0.24 0.030 0.073 0.93 0.084 0.36 0.28 0.034 0.10 0.92 1.00 0.93 0.95 0.99 0.53 0.94 1.00 0.98 0.73 0.20 1.00 0.96 0.62 0.20 0.31 Estimate 4.2 -150.2 -191.1 -96.1 -216.8 -115.1 -49.2 -154.4 -195.3 -100.3 -221.0 -119.3 -53.4 -40.9 54.1 -66.6 35.1 101.0 95.1 -25.7 76.1 142.0 -120.7 -19.0 46.9 101.7 167.6 65.9 Std. Error 18.13 44.18 83.29 36.78 55.13 34.01 41.99 46.83 84.72 39.92 57.28 37.38 44.77 82.44 47.06 62.47 44.93 51.24 84.85 94.27 83.68 87.23 47.96 37.68 45.01 55.73 60.93 28.76 df 96 12 12 12 12 12 12 12 12 12 12 12 12 96 12 12 12 12 12 12 12 12 96 12 12 12 12 96 124 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Table D11: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana frustule diameter (µm) between temperature selection groups and among assay temperature groups. Table corresponds to Figure 17a. Trait compared Frustule diameter Selection temp.@ Assay temp. Contrast t-ratio P-value 11.64 2.17 8.37 -4.07 3.26 -3.80 2.26 -4.80 2.17 -9.97 -4.07 -8.93 -3.80 11.64 -6.27 1.47 -5.55 0.74 -11.79 -6.27 -10.46 -5.55 11.64 -0.57 5.12 -6.12 -0.57 11.64 <.0001 0.43 <.0001 0.024 0.090 0.037 0.38 0.007 0.43 <.0001 0.024 <.0001 0.037 <.0001 0.0008 0.81 0.002 0.99 <.0001 0.0008 <.0001 0.002 <.0001 1.00 0.004 0.0009 1.00 <.0001 Std. Error 0.07 0.10 0.12 0.10 0.12 0.12 0.14 0.12 0.10 0.12 0.10 0.14 0.12 0.07 0.09 0.12 0.12 0.14 0.12 0.09 0.14 0.12 0.07 0.12 0.14 0.14 0.12 0.07 df 99 12 12 12 12 12 12 12 12 12 12 12 12 99 12 12 12 12 12 12 12 12 99 12 12 12 12 99 Estimate 0.77 0.21 0.97 -0.39 0.38 -0.46 0.31 -0.56 0.21 -1.15 -0.39 -1.22 -0.46 0.77 -0.60 0.17 -0.66 0.10 -1.36 -0.60 -1.43 -0.66 0.77 -0.07 0.70 -0.83 -0.07 0.77 125 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Table D12: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana frustule length (µm) between temperature selection groups and among assay temperature groups. Table corresponds to Figure 17b. Trait compared Frustule length Selection temp.@ Assay temp. Contrast t-ratio P-value 6.03 1.04 4.52 3.74 9.55 3.10 10.86 -5.09 -2.30 -0.99 5.38 -1.50 7.27 3.64 2.93 9.03 2.29 10.43 0.42 7.82 -0.19 9.67 4.55 -0.46 5.80 -4.79 2.02 6.12 <.0001 0.96 0.012 0.041 <.0001 0.11 <.0001 0.005 0.36 0.97 0.003 0.79 0.0002 0.010 0.15 <.0001 0.37 <.0001 1.00 0.0001 1.00 <.0001 0.0004 1.00 0.002 0.007 0.51 <0.0001 Std. Error 0.13 0.15 0.13 0.17 0.13 0.18 0.14 0.12 0.09 0.14 0.09 0.15 0.10 0.11 0.16 0.12 0.17 0.13 0.14 0.09 0.15 0.09 0.14 0.18 0.15 0.15 0.10 0.15 df 96 12 12 12 12 12 12 12 12 12 12 12 12 96 12 12 12 12 12 12 12 12 96 12 12 12 12 96 Estimate 0.77 0.16 0.58 0.64 1.28 0.55 1.48 -0.61 -0.20 -0.14 0.50 -0.22 0.71 0.42 0.47 1.12 0.39 1.32 0.06 0.70 -0.03 0.90 0.64 -0.09 0.84 -0.73 0.20 0.93 126 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Table D13: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana frustule volume (µm3) between temperature selection groups and among assay temperature groups. Table corresponds to Figure 17c. Trait compared Frustule volume Selection temp.@ Assay temp. Contrast t-ratio P-value 13.72 1.51 10.35 -0.32 9.06 0.40 8.98 -8.12 1.51 -9.53 -0.32 -8.36 0.40 13.72 -1.91 8.11 -0.98 8.08 -10.87 -1.91 -9.58 -0.98 13.72 0.71 9.42 -8.34 0.71 13.72 <.0001 0.79 <.0001 1.00 <.0001 1.00 <.0001 0.0001 0.79 <.0001 1.00 <.0001 1.00 <.0001 0.57 0.0001 0.97 0.0001 <.0001 0.57 <.0001 0.97 <.0001 1.00 <.0001 <.0001 1.00 <.0001 Estimate 35.24 4.25 39.49 -0.88 34.36 1.26 36.50 -30.99 4.25 -36.13 -0.88 -33.99 1.26 35.24 -5.13 30.11 -2.99 32.25 -40.37 -5.13 -38.24 -2.99 35.24 2.14 37.38 -33.11 2.14 35.24 Std. Error 2.57 2.82 3.82 2.79 3.79 3.15 4.06 3.82 2.82 3.79 2.79 4.06 3.15 2.57 2.68 3.71 3.06 3.99 3.71 2.68 3.99 3.06 2.57 3.02 3.97 3.97 3.02 2.57 df 99 12 12 12 12 12 12 12 12 12 12 12 12 99 12 12 12 12 12 12 12 12 99 12 12 12 12 99 127 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Table D14: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana frustule total surface area (µm2) to volume (µm3) ratio between temperature selection groups and among assay temperature groups. Table corresponds to Figure 18a. Trait compared Fr. SAtotal : V ratio Selection temp.@ Assay temp. Contrast t-ratio P-value -16.99 -1.14 -8.99 -0.39 -8.55 -0.97 -8.41 6.99 -1.14 7.88 -0.39 6.67 -0.97 -16.99 0.85 -7.93 0.09 -7.79 9.38 0.85 7.96 0.09 -16.99 -0.68 -8.68 7.49 -0.68 <.0001 0.94 <.0001 1.00 <.0001 0.97 <.0001 0.0003 0.94 0.0001 1.00 0.0004 0.97 <.0001 0.99 0.0001 1.00 0.0001 <.0001 0.99 0.0001 1.00 <.0001 1.00 <.0001 0.0001 1.00 Estimate -0.23 -0.029 -0.26 -0.009 -0.24 -0.026 -0.26 0.20 -0.029 0.22 -0.009 0.20 -0.026 -0.23 0.019 -0.21 0.002 -0.23 0.25 0.019 0.23 0.002 -0.23 -0.017 -0.25 0.21 -0.017 Std. Error 0.014 0.025 0.029 0.024 0.028 0.027 0.030 0.029 0.025 0.028 0.024 0.030 0.027 0.014 0.023 0.027 0.026 0.029 0.027 0.023 0.029 0.026 0.014 0.025 0.028 0.028 0.025 df 99 12 12 12 12 12 12 12 12 12 12 12 12 99 12 12 12 12 12 12 12 12 99 12 12 12 12 128 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 Table D15: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana valve only surface area (µm2) to total frustule volume (µm3) ratio between temperature selection groups and among assay temperature groups. Table corresponds to Figure 18b. Trait compared SAvalve : V ratio Selection temp.@ Assay temp. Contrast t-ratio P-value -6.30 -0.98 -4.71 -4.11 -10.38 -3.09 -11.86 5.85 2.30 0.62 -5.45 1.17 -7.49 -4.08 -3.55 -10.22 -2.52 -11.74 -0.90 -7.60 -0.14 -9.45 -4.72 0.53 -6.50 4.83 -2.34 -6.47 <.0001 0.97 0.008 0.0222 <.0001 0.12 <.0001 0.001 0.36 1.00 0.003 0.93 0.0001 0.002 0.056 <.0001 0.27 <.0001 0.98 0.0001 1.00 <.0001 0.0002 1.00 0.0005 0.007 0.35 <.0001 Std. Error 0.014 0.014 0.013 0.019 0.016 0.021 0.018 0.013 0.011 0.018 0.015 0.020 0.017 0.012 0.018 0.015 0.020 0.017 0.017 0.014 0.020 0.016 0.020 0.024 0.021 0.022 0.019 0.023 df 96 12 12 12 12 12 12 12 12 12 12 12 12 96 12 12 12 12 12 12 12 12 96 12 12 12 12 96 Estimate -0.09 -0.01 -0.06 -0.08 -0.17 -0.07 -0.21 0.07 0.03 0.01 -0.08 0.02 -0.13 -0.05 -0.06 -0.16 -0.05 -0.20 -0.02 -0.11 0.00 -0.15 -0.09 0.01 -0.14 0.11 -0.04 -0.15 129 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Table D16: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana girdle only surface area (µm2) to total frustule volume (µm3) ratio between temperature selection groups and among assay temperature groups. Table corresponds to Figure 18c. Trait compared Frustule SA:Vgirdle Selection temp.@ Assay temp. Contrast t-ratio P-value -11.54 -2.10 -7.83 4.15 -3.15 3.85 -2.25 4.24 -2.10 9.98 4.15 8.97 3.85 -11.54 6.14 -1.15 5.53 -0.56 11.45 6.14 10.32 5.53 -11.54 0.53 -5.36 6.27 0.53 -11.54 <.0001 0.46 0.0001 0.021 0.11 0.035 0.39 0.018 0.46 <.0001 0.021 <.0001 0.035 <.0001 0.0009 0.93 0.002 1.00 <.0001 0.0009 <.0001 0.002 <.0001 1.00 0.003 0.0008 1.00 <.0001 Std. Error 0.012 0.019 0.023 0.017 0.021 0.021 0.024 0.023 0.019 0.021 0.017 0.024 0.021 0.012 0.018 0.022 0.022 0.025 0.022 0.018 0.025 0.022 0.012 0.020 0.023 0.023 0.020 0.012 df 99 12 12 12 12 12 12 12 12 12 12 12 12 99 12 12 12 12 12 12 12 12 99 12 12 12 12 99 Estimate -0.14 -0.04 -0.18 0.07 -0.07 0.08 -0.05 0.10 -0.04 0.21 0.07 0.22 0.08 -0.14 0.11 -0.02 0.12 -0.01 0.25 0.11 0.26 0.12 -0.14 0.01 -0.13 0.15 0.01 -0.14 130 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 16@31 - 31@31 Table D17: All post-hoc, pairwise comparisons (Tukey HSD) for RM-ANOVA comparing T. pseudonana girdle surface area (µm2) to valve surface area ratio between temperature selection groups and among assay temperature groups. Table corresponds to Figure 18d. Trait compared Fr. SAvalve:SAgirdle Selection temp.@ Assay temp. Contrast t-ratio P-value -1.51 -0.66 1.88 -5.54 -6.65 -4.82 -6.46 0.72 3.47 -4.74 -5.80 -4.21 -5.81 2.42 -4.98 -6.00 -4.44 -6.01 -6.71 -7.91 -5.68 -7.41 -0.48 -0.64 -1.68 -0.28 -1.33 0.80 1.00 0.59 0.002 0.0004 0.007 0.0006 0.99 0.064 0.008 0.0015 0.019 0.002 0.24 0.005 0.001 0.013 0.001 0.0004 0.0001 0.002 0.0002 1.00 1.00 0.70 1.00 0.87 Std. Error 0.017 0.019 0.018 0.028 0.026 0.038 0.035 0.018 0.017 0.028 0.025 0.038 0.034 0.019 0.029 0.027 0.039 0.035 0.028 0.026 0.038 0.035 0.034 0.044 0.041 0.043 0.040 df 96 12 12 12 12 12 12 12 12 12 12 12 12 96 12 12 12 12 12 12 12 12 96 12 12 12 12 Estimate -0.03 -0.01 0.03 -0.16 -0.17 -0.18 -0.22 0.01 0.06 -0.13 -0.15 -0.16 -0.20 0.05 -0.14 -0.16 -0.17 -0.21 -0.19 -0.21 -0.22 -0.26 -0.02 -0.03 -0.07 -0.01 -0.05 131 16@10 - 31@10 16@10 - 16@16 16@10 - 31@16 16@10 - 16@26 16@10 - 31@26 16@10 - 16@31 16@10 - 31@31 31@10 - 16@16 31@10 - 31@16 31@10 - 16@26 31@10 - 31@26 31@10 - 16@31 31@10 - 31@31 16@16 - 31@16 16@16 - 16@26 16@16 - 31@26 16@16 - 16@31 16@16 - 31@31 31@16 - 16@26 31@16 - 31@26 31@16 - 16@31 31@16 - 31@31 16@26 - 31@26 16@26 - 16@31 16@26 - 31@31 31@26 - 16@31 31@26 - 31@31 REFERENCES 132 REFERENCES Aksnes DL, Egge JK. 1991. A theoretical model for nutrient uptake in phytoplankton. Marine Ecology Progress Series 70:65-72. Armbrust EV, Berges JA, Bowler C, Green BR, Martinez D, et al. 2004. The genome of the diatom Thalassiosira pseudonana: ecology, evolution, and metabolism. Science 306:79–86. Atkinson D. 1994. Temperature and organism size-a biological law for ectotherms? Advances in Ecological Research 25:1-58. Atkinson D. 1995. Effects of temperature on the size of aquatic ectotherms: exceptions to the general rule. Journal of Thermal Biology 20:61-74. Atkinson D, Ciotti BJ, Montagnes DJS. 2003. Protists decrease in size linearly with temperature: ca. 2.5% °C-1. Proceedings of the Royal Society B-Biological Sciences 270:2605-2611. Corkrey R, McMeekin TA, Bowman JP, Ratkowsky DA, Olley J, Ross T. 2014. Protein thermodynamics can be predicted directly from biological growth rates. PLoS One 9:e96100. Cortese G, Gersonde R. 2007. Morphometric variability in the diatom Fragilariopsis kerguelensis: implications for Southern Ocean paleoceanography. Earth and Planetary Science Letters 257:526-544. Craig SE, Thomas H, Jones CT, Li WKW, Greenan BJW, et al. 2013. Temperature and phytoplankton cell size regulate carbon uptake and carbon overconsumption in the ocean. Biogeosciences 10:11255-11282. Edwards KF, Klausmeier CA, Litchman E. 2011. Evidence for a three-way trade-off between nitrogen and phosphorus competitive abilities and cell size in phytoplankton. Ecology 92:2085-2095. Eppley RW. 1972. Temperature and phytoplankton growth in the sea. Fishery Bulletin 70:1063- 1085. Eppley RW, Sloan PR. 1966. Growth rates of marine phytoplankton: correlation with light absorption by cell chlorophyll-a. Physiologia Plantarum 19:47-59. Fawley MW. 1984. The Effects of light intensity and temperature interactions on the growth characteristics of Phaeodactylum tricornutum (Bacillariophyceae). Journal of Phycology 20:67-72. Field CB, Behrenfeld MJ, Randerson JT, Falkowski PG. 1998. Primary production of the biosphere: integrating terrestrial and oceanic components. Science 281:237-240. 133 Finkel ZV, Beardall J, Flynn KJ, Quigg A, Rees TAV, et al. 2010. Phytoplankton in a changing world: cell size and elemental stoichiometry. Journal of Plankton Research 32:119-137. Gillooly JF, Brown JH, West GB, Savage VM, Charnov EL. Effects of size and temperature on metabolic rate. Science 293:2248-2251. Guillard RRL, Hargraves PE. 1993. Stichochrysis immobilis is a diatom, not a chrysophyte. Phycologia 32:234-236. Hare CE, Leblanc K, DiTullio GR, Kudela RM, Zhang Y, et al. 2007. Consequences of increased temperature and CO2 for phytoplankton community structure in the Bering Sea. Marine Ecology Progress Series 352:9-16. Harrison PJ, Conway HL, Holmes RW, Davis CO. 1977. Marine diatoms grown in chemostats under silicate or ammonium limitation. III. Cellular chemical composition and morphology of Chaetoceros debilis, Skeletonema costatum, and Thalassiosira gravida. Marine Biology 43:19-31. Joint IR, Pomroy AJ, Robinson GA, Morris RJ, McCartney MJ. 1987. Morphological changes in Skeletonema costatum (Bacillariophyceae) during a spring bloom in a marine ecosystem enclosure. British Phycological Journal 22:119-124. Jørgensen EG. 1966. Photosynthetic activity during the life cycle of synchronous Skeletonema cells. Physiologia Plantarum 19:789-799. Jørgensen EG. 1968. The adaptation of plankton algae. II. Aspects of the temperature adaptation of Skeletonema costatum. Physiologia Plantarum 21:423-427. Lenth RV. 2016. Least-squares means: the R package lsmeans. Journal of Statistical Software 69:1-33. Litchman E, Klausmeier CA, Miller JR, Schofield OM, Falkowski PG. 2006. Multi-nutrient, multi-group model of present and future oceanic phytoplankton communities. Biogeosciences 3:607-663. Martiny AC, Pham CTA, Primeau FW, Vrugt JA, Moore JK, et al. 2013. Strong latitudinal patterns in the elemental ratios of marine plankton and organic matter. Nature Geoscience 6:279-283. Menzel DW, Corwin N. 1965. The measurement of total phosphorus in seawater hased on the liberation of organically bound fractions by persulfate oxidation. Limnology and Oceanography 10:280-282. 134 Montagnes DJS, Franklin DJ. 2001. Effect of temperature on diatom volume, growth rate, and carbon and nitrogen content: reconsidering some paradigms. Limnology and Oceanography 46:2008-2018. Neori A, Vernet M, Holm-Hansen O, Haxo FT. 1986. Relationship between action spectra for chlorophyll a fluorescence and photosynthetic O2 evolution in algae. Journal of Plankton Research 8, 537-548. O’Donnell DR, Hamman CR, Johnson EC, Kremer CT, Klausmeier CA, Litchman E. 2018. Rapid thermal adaptation in a marine diatom reveals constraints and tradeoffs. Global Change Biology doi: 10.1111/gcb.14360. [Epub ahead of print] Padfield D, Yvon-Durocher G, Buckling A, Jennings S, Yvon-Durocher G. 2016. Rapid evolution of metabolic traits explains thermal adaptation in phytoplankton. Ecology Letters 19:133-142. Pinheiro J, Bates D, DebRoy S, Sarkar D and R Core Team. 2018. nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.1-137, https://CRAN.R- project.org/package=nlme. Polovina JJ, Howell EA, Abecassis M. Ocean's least productive waters are expanding. Geophysical Research Letters 35:1-5. Ratkowsky DA, Olley J, Ross T. 2005. Unifying temperature effects on the growth rate of bacteria and the stability of globular proteins. Journal of Theoretical Biology 233:351-362. Rhee G-Y, Gotham IJ. 1981. The effect of environmental factors on phytoplankton growth: temperature and the interactions of temperature with nutrient limitation. Limnology and Oceanography 26:635-648. Sarmiento JL, Slater R, Barber R, Bopp L, Doney SC, et al. 2004. Response of ocean ecosystems to climate warming. Global Biogeochemical Cycles 18:1-23. Schlüter L, Lohbeck KT, Gutowska M, Gröger JP, Riebesell U. Adaptation of a globally important coccolithophore to ocean warming and acidification. Nature Climate Change 4:1024-1030. Thomas MK; Kremer CT; Klausmeier CA; Litchman E. 2012. A global pattern of thermal adaptation in marine phytoplankton. Science 338:1085-1088. Thompson PA, Guo M, Harrison PJ. 1992a. Effects of variation in temperature. I. On the biochemical composition of eight species of marine phytoplankton. Journal of Phycology 28:481-488. Toseland A, Daines SJ, Clark JR, Kirkham A, Strauss J, et al. 2013. The impact of temperature on marine phytoplankton resource allocation and metabolism. Nature Climate Change 3:1-6. 135 Weber T, Deutsch C. 2012. Oceanic nitrogen reservoir regulated by plankton diversity and ocean circulation. Nature 489:419-422. Winder MR, Reuter JE, Schladow SG. 2009. Lake warming favours small-sized planktonic diatom species. Proceedings of the Royal Society B-Biological Sciences 276:427-435. Yvon-Durocher G, Dossena M, Trimmer M, Woodward G, Allen AP. 2015b. Temperature and the biogeography of algal stoichiometry. Global Ecology and Biogeography 24:662-570. Yvon-Durocher G, Schaum C-E, Trimmer M. 2017. The temperature dependence of phytoplankton stoichiometry: investigating the roles of species sorting and local adaptation. Frontiers in Microbiology 8:1-14. 136