Ocean-related global change alters lipid biomarker production in common marine phytoplankton

. Lipids, in their function as trophic markers in food webs and organic matter source indicators in water column and sediments, provide a tool for reconstructing the complexity of global change effects on aquatic ecosystems. It remains unclear how ongoing changes in multiple environmental drivers affect the production of key lipid biomarkers in marine phytoplankton. Here, we tested the responses of sterols, alkenones and fatty acids (FAs) in the diatom Phaeodactylum tricornutum , the cryptophyte Rhodomonas sp. and the haptophyte Emiliania huxleyi under a full-factorial combination of 15 three temperatures (12, 18 and 24°C), three N:P supply ratios (molar ratios 10:1, 24:1 and 63:1) and two p CO 2 levels (560 and 2400 µatm) in semi-continuous culturing experiments. Overall, N and P deficiency had a stronger effect on per-cell contents of sterols, alkenones and FAs than warming and enhanced p CO 2 . Specifically, P deficiency caused an overall increase in biomarker production in most cases, while N deficiency, warming and high p CO 2 caused non-systematic changes. Under future ocean scenarios, we predict an overall decrease in carbon-normalized contents of sterols and polyunsaturated 20 fatty acids (PUFAs) in E. huxleyi and P. tricornutum , and a decrease in sterols but an increase in PUFAs in Rhodomonas sp. Variable contents of lipid biomarkers indicate a diverse carbon allocation between marine phytoplankton species in response to changing environments. Thus, it is necessary to consider the changes in key lipids and their consequences for food web dynamics and biogeochemical cycles, when predicting the influence of global change on marine ecosystems. alkenones and compare these with those of


Introduction 25
Ocean phytoplankton has profoundly responded to and driven natural climatic variability throughout Earth's history (Riding, 1992;Falkowski and Oliver, 2007;Falkowski, 2015). In the contemporary ocean, human-induced physical and chemical modifications are complex and concurrent, including warming, acidification, deoxygenation, and changes in nutrient availability (Doney et al., 2012;Moore et al., 2013;DeVries et al., 2017). The ocean-related global change fundamentally affects marine ecosystems (Hoegh-Guldberg and Bruno, 2010). These include especially global 30 phytoplankton biomass decreases (Boyce et al., 2010;Moore et al., 2018;Lotze et al., 2019) and plankton communities changes (Richardson and Schoeman, 2004;Jonkers et al., 2019), which consequently alters food-web dynamics (Kortsch et al., 2015;du Pontavice et al., 2020) and biogeochemical cycles (Hofmann and Schellnhuber, 2009;Gruber, 2011;Doney et al., 2012). A major challenge is the lack of a better understanding of the complexity of biological impacts of global change, which has hindered the prediction of potential feedbacks between marine ecosystems and projected environmental changes. 35 Some of phytoplankton-produced biomolecules (biomarkers), functioning as indicators of nutritional food quality (Müller-Navarra, 2008) and tracers of organic matter sources (Volkman et al., 1998), have provided crucial insight into the trajectory of ecological responses to changing environment along food webs in the present-day ocean (Ruess and Müller-Navarra, 2019), and over geological time (Brocks et al., 2017).
Lipids are amongst the most important and widely used biomarkers, because they have far-reaching biochemical and 40 physiological roles in cells and are sensitive to environmental changes (Arts et al., 2009), but also because of their dominance in the geological record as fossil molecules to reveal life's signatures on Earth (Falkowski and Freeman, 2014).
There are also growing applications of lipids as proxies for global climate and marine ecosystem change. Of all biomarkers, fatty acids (FAs), the basic constituents of most algal lipids, have received the most intense attention. Polyunsaturated fatty acids (PUFAs) are essential for many animals and have been applied as nutritional components to study trophic interactions 45 (Brett and Müller-Navarra, 1997;Müller-Navarra et al., 2000;Dalsgaard et al., 2003;Kelly and Scheibling, 2012;Ruess and Müller-Navarra, 2019). The impact of environmental changes on phytoplankton FAs has been well studied, mostly with a focus on the effects of temperature and nutrient changes (reviewed by Guschina and Harwood, 2009;Galloway and Winder, 2015;Hixson and Arts, 2016), while the interplay between different environmental drivers has been recently tested (Bermúdez et al., 2015;Bi et al., 2017Bi et al., , 2018. However, determining how phytoplankton lipids respond to global change still 50 faces substantial challenges, partly because data on other important lipid classes such as sterols and alkenones are scarce.
Understanding the impact of environmental change on these lipid classes is critical to achieve a better application of lipid biomarkers to contemporary issues and to the past record of marine ecosystems.
Given the multiple biochemical roles and source specificity of sterols, their composition and biosynthetic pathways in phytoplankton have been identified in different phyla (Fabris et al., 2014;Villanueva et al., 2014;Volkman, 2016). It has been observed that sterol contents (per carbon or dry weight or percentage of total lipids) in phytoplankton vary with 65 environmental conditions. Vé ron et al. (1996) found a dramatic decrease in total sterol contents (per dry weight) in Phaeodactylum tricornutum as temperature increased. In contrast, a significant increase in carbon-normalized sterol contents with increasing temperature was found in other algal species (Piepho et al., 2012;Ding et al., 2019). Enhanced pCO2 caused an increase in the percentage of sterols (% of total lipids) in Dunaliella viridis (Gordillo et al., 1998), but no clear change in per-cell contents of sterols in Emiliania huxleyi (Riebesell et al., 2000). Although significant interactions between two 70 environmental factors have been observed on carbon-normalized or per-cell sterol contents in certain phytoplankton species (Piepho et al., 2010(Piepho et al., , 2012Ding et al., 2019), the impacts of multiple environmental drivers on phytoplankton sterol contents have not been thoroughly investigated.
Long-chain alkenones are major lipids produced only by certain species of haptophytes, e.g., oceanic species E. huxleyi and Gephyrocapsa oceanica (Volkman et al., 1980c;Conte et al., 1995) and coastal species Isochrysis galbana (reviewed by 75 Conte et al., 1994). Alkenones in E. huxleyi are believed to be used for energy storage (Epstein et al., 2001;Eltgroth et al., 2005), while little is known about the entire biosynthetic pathway of alkenones and their evolutionary and ecological functions (Rontani et al., 2006;Kitamura et al., 2018). Alkenones may have fitness and trophic benefit for their producers, because these unusual lipids are not only more photostable than other neutral lipids such as triacylglycerols, but also resistant to digestion, perhaps making alkenone producers less suitable for grazers (Eltgroth et al., 2005). Moreover, alkenones are 80 well preserved in sediments over millions of years and thus their unsaturation ratios [e.g., the 37 ′ index (= C37:2/(C37:2 + C37:3)) (Brassell et al., 1986;Prahl and Wakeham, 1987)] are widely applied for reconstructing sea surface temperatures (Rosell-Melé and Prahl, 2013;Herbert et al., 2016). A long-standing issue for the use of alkenones to infer paleo-ocean surface temperature is how the production of these compounds is influenced by other environmental factors such as nutrients.
Thus, culture studies have been conducted to test alkenone contents (mostly per-cell contents) in several species of 85 haptophytes such as E. huxleyi under different growth phases (Wolhowe et al., 2009;Pan and Sun, 2011;Wolhowe et al., 2015), salinity (Sachs et al., 2016), temperature , and nutrient concentrations (Rokitta et al., 2014;Wördenweber et al., 2018). Conflicting results have been observed in different studies, e.g., independence of C37 -C39 alkenone contents on temperature (Prahl et al., 1988) versus significant responses of C37 alkenone contents to temperature changes in E. huxleyi . More empirical evidence appears necessary to determine how the total contents and 90 the ratios of specific alkenone isomers respond to multiple environmental drivers, which would allow us to better understand their roles in ecology and biogeochemistry.
Here, we present data from semi-continuous culture experiments to tackle the question of how important lipid biomarkers (FAs, sterols and alkenones) respond to the changes in multiple environmental drivers (temperature, N:P supply ratios and pCO2) in three phytoplankton species (the diatom P. tricornutum, the cryptophyte Rhodomonas sp. and the haptophyte E. 95 huxleyi). Specifically, we analyze the changes of particulate organic carbon (POC)-normalized (carbon-normalized hereafter) and per-cell contents of major sterols and alkenones in the three species, and compare these responses with those of published FA data from the same experiments. Our aims are to determine (i) how sterols and alkenones respond to the changes of multiple environmental drivers, and (ii) how the responses of sterols, alkenones and FAs differ between each other. The goal of this study is to generate a better understanding of the impact of ocean-related global change on lipid 100 biomarker productions in marine phytoplankton, which will help to quantitatively apply lipid biomarkers as proxies for ecosystem change, and to finally scale up from specific physiological roles of lipids to their effects on energy flow in food webs in the changing ocean.

Experimental design 105
The three phytoplankton species used in the experiments were the diatom P. tricornutum [SAG, 1090-1b;isolated from Plymouth, UK (De Martino et al., 2007)], the cryptophyte Rhodomonas sp. (isolated from the Kiel Bight, Baltic Sea and identified by the working group at GEOMAR; maintained and still available from GEOMAR for laboratory culture), and the haptophyte E. huxleyi (internal culture collection reference code: A8; isolated from waters off Terceira Island, Azores). P. tricornutum and Rhodomonas sp. are model species widely used in studies of diatom genomes, cryptophyte photosynthesis 110 and planktonic trophic dynamics (Bi et al., 2017), and E. huxleyi is one of the major calcifying organisms in the pelagic ocean (Winter et al., 2014). The algal cultures were non-axenic. Because the biomass of bacteria was very low, the bacterial influence on the chemical composition of phytoplankton was negligible. Over the course of the experiments, the cultures of all species were exposed to a salinity of 37 psu and a light intensity of 100 µmol photons m -2 s -1 following a light:dark cycle of 16:8 h in temperature-controlled rooms of 12, 18 and 24°C. The culture medium was prepared according to the modified 115 Provasoli's medium (Provasoli, 1963;Ismar et al., 2008), with enrichment nutrient solutions added to sterile filtered (0.2 µm pore size, Sartobran ® P 300; Sartorius, Goettingen, Germany) North Sea water. Sodium nitrate and potassium dihydrogen phosphate were added to achieve the molar ratios of 10:1 (35.2 µmol L -1 N and 3.6 µmol L -1 P), 24:1 (88 µmol L -1 N and 3.6 µmol L -1 P) and 63:1 (88 µmol L -1 N and 1.4 µmol L -1 P). Sodium silicate pentahydrate was also added to diatom cultures at a concentration of 88 µmol L -1 . Initial pCO2 was manipulated by bubbling with CO2-enriched air (560 and 2400 µatm). 120 Subsequently, the culture medium was transferred into sealed cell culture flasks with a 920-mL culture volume. Each treatment was replicated three times. All culture flasks were carefully agitated twice per day at a set time to minimize sedimentation.
At the onset of the experiments, each species was grown in batch cultures across a fully factorial combination of three temperatures (12, 18 and 24°C), three N:P supply ratios (molar ratios 10:1, 24:1 and 63:1) and two pCO2 levels (560 and 125 2400 µatm) (Fig. S1). The chosen levels of temperature, N:P supply ratio and pCO2 cover the ranges of typical changes of the three factors in natural environments, and they are in general agreements with projections. The temperature regimes broadly conform to sea surface temperatures in the source regions for the three taxa studied: Plymouth, UK for P. tricornutum (~9 -17°C) (Highfield et al., 2010), the Kiel Bight for Rhodomonas sp. (~3 -18°C) (Hiebenthal et al., 2013), and the Azores for E. huxleyi (16 -22°C; http://dive.visitazores.com/en/when-dive). The 6°C elevation also mimicks the 130 largest projected warming under climate change scenarios (Sommer and Lengfellner, 2008). N:P molar ratio of 24:1 was selected as the balanced ratio under which phytoplankton cultures are typically maintained (Guillard, 1975). Surface ocean inorganic N and P concentrations are highly depleted throughout much of the low-latitude oceans (Moore et al., 2013).
Nutrient availability in the future will be altered by increasing external nutrient inputs especially in coastal oceans, changes in surface ocean chemistry driven by anthropogenic increases in atmospheric CO2, and changes in ocean circulation (Moore 135 et al., 2013). Therefore, inorganic N:P ratios show a strong spatial variation in the oceans, e.g., a low ratio of 6:1 mol mol -1 in the center of the South Pacific Gyre versus a global scenario of an increase driven by the high N:P atmospheric deposition of ~ 370:1 mol mol -1 (Bonnet et al., 2008;Peñuelas et al., 2012). N:P ratios of 10:1 and 63:1 selected in this study cover large-scale spatial patterns of nutrient status, and thus our prediction of lipid biomarker production changes is based on both future open ocean and coastal conditions. Partial CO2 pressure of 560 µatm is double the pre-industrial value and is a 140 standard level for determining climate model sensitivity to pCO2 forcing (e.g., IPCC, 2014). The value of 2400 µatm is at the mid-range of the projected values (1371 -2900 µatm) by 2150 (RCP8.5 scenario; IPCC, 2014). Also, a high pCO2 has been observed in the areas where one of the studied algae was isolated, e.g., 375 -2309 µatm in the Kiel Bight (Thomsen et al., 2010).
The observed maximal growth rate (µmax, d -1 ) was calculated from the changes of cell numbers within the exponential 145 growth phase in batch cultures (Bi et al., 2012). Once the early stationary phase was reached, semi-continuous cultures were started with the algae from batch cultures, and the gross growth rate (µ, d -1 ) was set as 20% of µmax. We calculated the volume of the daily renewal incubation water by multiplying daily renewal rate [D, d -1 ; D = 1-e -µ· t , where t is renewal interval (here t = 1 day)] with the incubation volume of 920 mL. According to the gross growth rate of 20% of µmax, the renewal volumes were about 10% of the total culture volume (100 -200 mL out of 920 mL), and renewal interval was 1 d. 150 Steady state was reached at about 10 d with slight differences between the individual cultures. Renewal of the cultures was carried out at the same hour every day using fresh filtered seawater pre-acclimated to target pCO2, and CO2-enriched water.
The difference between the gross growth rate and the loss rate, i.e., the net growth rate [r (d -1 ); r = µ -D] was used to assess the steady state, at which r was zero, and µ was equivalent to D.

Sampling and measurements 155
At steady state in semi-continuous cultures, samples were collected for the measurements of cell density, dissolved inorganic carbon (DIC), total alkalinity, pH, POC, FAs, sterols and alkenones. Cell density was measured daily in batch and semi-continuous cultures using an improved Neubauer hemacytometer (Glaswarenfabrik Karl Hecht GmbH, Rhön Mountains, Germany) under a microscope (Hund, Wetzlar, Germany). Also, pH measurements were carried out daily in semi-continuous cultures, and the electrode was calibrated using standard pH buffers (pH 4 and pH 7; WTW, Weilheim, 160 Germany).
DIC samples were taken on sampling days with 10-mL glass vials (Resteck, Germany) filled using a peristaltic pump and an intake tube containing a single-use syringe filter (0.2 µm, Minisart RC25; Sartorius, Goettingen, Germany). Vials were immediately sealed and stored in the dark at 4°C. DIC was measured according to Hansen et al. (2013) using a gas chromatographic system (8610C; SRI-Instruments, California, USA). For total alkalinity analysis, samples were filtered 165 (GF/F filters; Whatman GmbH, Dassel, Germany) and analyzed with the Tirino plus 848 (Metrohm, Filderstadt, Germany).
The remaining carbonate parameter pCO2 was calculated from DIC and total alkalinity using CO2SYS (Pierrot et al., 2006) and the constants of Hansson (1973) and Mehrbach et al. (1973) that were refitted by Dickson and Millero (1987) (Table S1; Bi et al., 2017Bi et al., , 2018. DIC and pH did not differ substantially between different temperature and N:P ratio treatments in our study (Table S1). 170 POC and FA samples (15 -50 ml) were taken on pre-combusted and pre-washed (5-10% HCl) GF/F filters (Whatman GmbH, Dassel, Germany). After filtration, filters for POC analysis were immediately dried at 60℃ and stored in a desiccator, and those for FA measurements were stored at -80℃. For POC analysis in E. huxleyi, particulate inorganic carbon was removed by exposing filters to fuming hydrochloric acid for 12h. POC was determined using an elemental analyzer (Thermo Flash 2000; Thermo Fisher Scientific, Schwerte, Germany) after Sharp (1974). FAs were measured as FA methyl esters 175 (FAMEs) using a gas chromatograph (Trace GC-Ultra; Thermo Fisher Scientific, Schwerte, Germany). Analytical procedure of FAs was modified after Christie (1989). Briefly, FAs were extracted with a solvent mixture of chloroform/dichloromethane/methanol (1:1:1, volume ratios). The FAME 19:0 was added as an internal standard and 21:0 as a esterification control. The extracted FAMEs were dissolved in 100 μL n-hexane for analysis. Sample aliquots (1 μL) were injected onto the GC with hydrogen as the carrier gas. Individual FAs were identified with reference to the standards 180 (Supelco 37 component FAME mixture and Supelco Menhaden fish oil), and the peaks were integrated using Chromcard software (Thermo Fisher Scientific, Schwerte, Germany), which included saturated, monounsaturated and polyunsaturated FAs (Bi et al., 2017(Bi et al., , 2018. Samples for sterol and alkenone analysis were filtered (GF/F filters; Whatman GmbH, Dassel, Germany) and measured according to the procedure of Zhao et al. (2006). The filtration volume of sterol and alkenone samples was 100 -200 mL, 185 while only 10% of the extractions were used for GC quantification and the rest was stored for later use. Lipids were extracted ultrasonically eight times from the freeze-dried filter samples utilizing dichloromethane and methanol (3:1, volume ratios) as extraction solvent, with C19 n-alkanol added for quantification. After hydrolysis with 6% potassium hydroxide in dichloromethane, the lipids were separated into a polar fraction and a non-polar fraction using silica gel chromatography.
The polar lipid fractions containing sterols from the extracts of all the three species, as well as C37 -C39 alkenones from the E. 190 huxleyi extracts, were eluted with 22 mL dichloromethane and methanol (95:5, volume ratios). Subsequently, the polar lipid fractions were silylated with 80 μL N,O-bis(trimethylsilyl)-trifluoroacetamide at 70℃ for 1h. The sterol and alkenone fractions were analyzed and quantified using an Agilent 7890A GC with flame ionization detection (50 m HP-1 capillary column, 0.32 mm i.d., 0.17 μm film thickness) by comparing analyte peak area to known amount of the internal standard C19 n-alkanol. The oven temperature held initially at 80℃ for 1 min, increased to 200℃ at 25℃ min -1 and then programmed to 195 250℃ at 4℃ min -1 and to 300℃ at 1.7℃ min -1 holding for 12 min, and finally to 315℃ at 5℃ min -1 with an 8 min isothermal period.
Alkenones were identified using an Agilent GC 7890B (30 m HP-5MS column, 0.25 mm i.d., 0.25 μm film thickness) connected to an Agilent MSD 5977B mass selective detector (70 eV constant ionization potential, ion source temperature 205 230℃). The temperature program started with a 1 min hold time at 80 ℃ and then increased to 200℃ at 25℃ min -1 , followed by a 4℃ min -1 ramp to 250℃ and 1.8℃ min -1 to 300℃ holding for 10 min, and finally increased to 315℃ at 5℃ min -1 holding for 5 min. Alkenone identifications were performed by comparing sample mass spectra generated by GC-MS to previous published EI mass spectra, based on the molecular ion and prominent ions of each alkenone component (de Leeuw et al., 1980;Volkman et al., 1980c;Lopez and Grimalt, 2006). Note that the molecular ions for C39:3 ethyl ketone 210 (C39:3 Et) (at m/z 556) and C39:2 ethyl ketone (C39:2 Et) (at m/z 558) were below detection, only prominent ion fragments were detected. In addition, the comparisons of the GC retention time of alkenone molecules between our study and previous work (Volkman et al., 1980c) indicated the presence of C39:3 Et and C39:2 Et.

Statistics
Generalized linear mixed models (GLMMs; Bolker et al., 2009) were used to test the effects of temperature, N:P supply 215 ratios and pCO2 on carbon-normalized and per-cell contents of brassicasterol/epi-brassicasterol and C37 -C39 total alkenones (as μg mg C -1 and pg cell -1 ), per-cell contents of C37 alkenones, C38 alkenones, C38 ethyl ketones (C38 Et) and C38 methyl ketones (C38 Me), C37/C38 alkenone ratios and C38 Et/C38 Me ratios (C38 Et/Me), with temperature, N:P supply ratios and pCO2 as fixed effects. Target distributions were tested before GLMMs were taken. Subsequently, appropriate link functions were chosen, e.g., identity link function for any distribution except for multinomial, and logit link function for the binomial 220 or multinomial distribution. To find the model that best predicted targets, we tested models containing first order effects, and second and third order interactions of temperature, N:P supply ratios and pCO2. The Akaike Information Criterion corrected (AICc) was used to select the best model for each response variable, with a lower AICc value representing a better fit of the model. When the changes of AICc values were 10 units or more, it was considered as a reasonable improvement in the fitting of GLMMs (Bolker et al., 2009). In case AICc values were comparable (< 10 units difference), the simpler model was 225 chosen. Based on the differences in AICc values, models containing only first order effects of temperature, N:P supply ratios and pCO2 were chosen as the best models for all response variables (bold letters in Table S2), while those containing second or third order interactions were not selected.
Principal component analysis (PCA) was conducted to visualize the responses of carbon-normalized contents of brassicasterol/epi-brassicasterol and total fatty acids (TFAs) in the three species, and total alkenones in E. huxleyi to the 230 changes of temperature, N:P supply ratios and pCO2. Data for TFAs were from previous studies (Bi et al., 2017(Bi et al., , 2018. GLMMs were performed using SPSS 19.0 (IBM Corporation, New York, USA). PCA were conducted using R package factoextra (Kassambara and Mundt, 2017) and FactoMineR (Le et al., 2008) in R version 3.5.1 (R Development Core Team, R Foundation for Statistical Computing, Vienna, Austria). All statistical tests were conducted at a significance threshold of p = 0.05. 235

Responses of brassicasterol/epi-brassicasterol to environmental changes
GLMM results showed that per-cell contents of brassicasterol/epi-brassicasterol responded significantly to changes in N:P supply ratios in the three species (bold letters in Table 1). Moreover, per-cell contents of brassicasterol/epi-brassicasterol in P. tricornutum also showed significant responses to temperature changes, while non-significant effects of pCO2 were 245 observed in all species. Specifically, higher per-cell contents of brassicasterol/epi-brassicasterol were observed at higher temperatures and higher N:P supply ratios in P. tricornutum ( Fig. 1a; Table S3), under the lowest and highest N:P supply ratios in Rhodomonas sp. (Fig. 1c), and under higher N:P supply ratios in E. huxleyi (Fig. 1e).
Carbon-normalized contents of brassicasterol/epi-brassicasterol responded significantly to pCO2 in Rhodomonas sp., and to temperature and N:P supply ratios in E. huxleyi (bold letters in Table 1), but with non-significant responses in P. 250 tricornutum. Carbon-normalized contents of brassicasterol/epi-brassicasterol in Rhodomonas sp. decreased as pCO2 increased (Fig. 1d), while those in E. huxleyi were generally higher at higher temperatures and under the balanced N:P supply ratio (N:P = 24:1 mol mol -1 ; Fig. 1e).

Responses of alkenones to environmental changes
Total alkenone content per cell in E. huxleyi increased with increasing N:P supply ratios (Fig. 2 a; bold letters in Table 1;  255   Table S4). However, carbon-normalized contents of total alkenones showed non-significant responses to changes in temperature, N:P supply ratios and pCO2.
C37/C38 alkenone ratios responded significantly to all three environmental factors (bold letters in Table 1), showing a clear increase with increasing temperature, a higher value under the lowest and highest N:P supply ratios, and a slight decrease at high pCO2 (Fig. 2 c and d). C38 Et/Me ratios had significant responses only to temperature changes, clearly higher at the 260 highest temperature (Fig. 2 e and f).

Comparisons of sterol, alkenone and fatty acid responses
PCA extracted four axes with eigenvalues > 1, and the first two axes in PCA explained 44.1% of total variance (Table   S5). PC axis 1 explains 26.4% of the total variance and largely differentiates between the highest (N:P = 63:1 mol mol -1 ) and lower N:P treatments (N:P = 10:1 and 24:1 mol mol -1 ). Along PC axis 1, N:P supply ratios correlated positively with carbon-265 normalized contents of TFAs in P. tricornutum and Rhodomonas sp., but negatively with brassicasterol/epi-brassicasterol in E. huxleyi (Fig. 3). PC axis 2 (17.7% of the total variance) differentiates between the highest and both colder temperature treatments, along which temperature showed positive correlations with brassicasterol/epi-brassicasterol in Rhodomonas sp. and alkenones in E. huxleyi, but a negative correlation with TFAs in E. huxleyi. Both Dim1 and Dim 2 loadings for pCO2 were very low (0.185 and 0.151, respectively), showing a weaker effect of pCO2 on lipid biomarkers compared to 270 temperature and N:P supply ratios. Along PC axis 2, pCO2 plays a negative role particularly on TFAs in E. huxleyi, while the contribution of pCO2 to other lipid biomarkers was weak.

Discussion
To our knowledge, this is the first study to disentangle the effects of multiple environmental drivers on sterol and alkenone productions, and to compare the responses of sterols, alkenones and FAs in marine phytoplankton. The mean 275 percent changes of the three lipid biomarkers were elucidated (Table 2), particularly showing obvious changes in per-cell lipid contents which underly specific modes of biosynthesis. Furthermore, the PCA results highlight that the responses of TFA carbon-normalized contents to N:P supply ratios were opposite to that of brassicasterol/epi-brassicasterol and alkenones ( Fig. 3; Table S5), e.g., strong positive correlations of TFAs with N:P ratios in P. tricornutum and Rhodomonas sp., but negative ones of brassicasterol/epi-brassicasterol with N:P ratios in E. huxleyi. Also, carbon-normalized alkenone contents in 280 E. huxleyi correlated positively with temperature, but TFAs contents showed a negative correlation with temperature. Such variable responses of the three classes of lipid biomarkers can be attributed to their specific physiological functions and biosynthetic pathways (Riebesell et al., 2000). Carbon-normalized contents of FAs were particularly sensitive to environmental changes, because FAs can be incorporated into different types of lipids, and thus play multiple roles within the cells such as energy storage, membrane components and metabolic regulations (Guschina and Harwood, 2009;Van 285 Mooy et al., 2009). Conversely, less pronounced changes in carbon-normalized contents of sterols and alkenones may reflect their major roles in membrane functions and energy storage, respectively. The varying production of lipid biomarkers indicate potential changes of energy flow in marine food webs in response to ocean-related global change.
Alkenones were only observed in E. huxleyi in our study, consistent with previous results showing that these compounds were only synthesized by a few haptophytes including E. huxleyi (Volkman et al., 1980b;Volkman et al., 1998). The 300 alkenone composition of E. huxleyi was characterized by the presence of four pairs of isomers, including eight alkenone compounds (Table S4) (Marlowe et al., 1984;Riebesell et al., 2000;Sachs et al., 2016). Moreover, higher abundance of several alkenone components have been also observed in some E. huxleyi strains under certain culture conditions, e.g., C37:4 Me in the strain 1742 (Eltgroth et al., 2005) or at low temperatures (our study; Prahl and Wakeham, 1987). In addition, two other compounds C38:4 Et and C38:4 Me were also found in one E. huxleyi strain (Marlowe et al., 1984). 305 Collectively, the results above highlight the similarity of sterol and alkenone composition in algal species in our study with those in conspecifics or congeneric phytoplankters in previous studies. Sterol and alkenone composition can vary between algal strains and can be affected by environmental changes (Conte et al., 1994;Volkman, 2003), which may explain the differences between our findings and previous results. In the following section, specific response patterns of brassicasterol/epi-brassicasterol and alkenones are evaluated and quantified, which are further compared with FA responses 310 under changing temperature, N:P supply ratios and pCO2.

Responses of brassicasterol/epi-brassicasterol contents
Increasing temperature caused an overall 12% increase in carbon-normalized contents of brassicasterol/epi-brassicasterol in E. huxleyi, but non-significant changes in P. tricornutum and Rhodomonas sp. (Table 1; Table 2). Consistent with our findings, positive correlations between increasing temperature and sterol carbon-normalized contents have been observed in 315 the dinoflagellates Karenia mikimotoi and Prorocentrum minimum , and the green alga Scenedesmus quadricauda (Piepho et al., 2012). High sterol contents at high temperature could be predicted based on its biochemical function, because increasing levels of sterols can reduce membrane fluidity to enable an organism's functional activity as temperature increases (Ford and Barber, 1983).
Enhanced partial CO2 pressure caused a 21% decrease in carbon-normalized brassicasterol/epi-brassicasterol contents in 320 Rhodomonas sp., and non-significant responses in P. tricornutum and E. huxleyi; however, per-cell contents of brassicasterol/epi-brassicasterol and POC showed non-significant changes in all three species (Table 1; Table 2). Minor effects of CO2 concentration on per-cell contents of sterols have been observed in another strain of E. huxleyi (PML B92/11) (Riebesell et al., 2000) and the Chlorophyceae D. viridis (Gordillo et al., 1998). While the mechanism underlying sterol responses to pCO2 is still unclear, our results indicate that enhanced pCO2 did not induce substantial changes in per-cell 325 contents of sterols in phytoplankton due to the role of sterols in membrane composition and functions (Riebesell et al., 2000).
Nevertheless, enhanced pCO2 might change carbon metabolism in phytoplankton (Gordillo et al., 2001), as revealed by variable carbon-normalized contents of brassicasterol/epi-brassicasterol in Rhodomonas sp. in our study. N and P deficiency caused overall 8% and 37% decreases in carbon-normalized brassicasterol/epi-brassicasterol contents, respectively, in E. huxleyi, but non-significant changes in other two species (Table 1; Table 2). Carbon-normalized or dry-330 weight contents of sterols in phytoplankton generally reduced in response to N or P deficiency Piepho et al., 2010;Ding et al., 2019). Furthermore, the relatively higher per-cell contents of sterols in response to P deficiency than N deficiency have been also found in the three species in this study and in the freshwater diatom Stephanodiscus minutulus (Lynn et al., 2000). Lipid modifications triggered by nutrient deficiency have been well studied in the plant Arabidopsis and more recently elucidated in typical phytoplankters (Van Mooy et al., 2009;Abida et al., 2015;Shemi et al., 2016). In P. 335 tricornutum, N deficiency exerted more severe stress on membrane glycerolipids than P deficiency which caused a stepwise adaptive response, resulting in undetectable phospholipids and instead the increase in the synthesis of non-phosphorus lipids (Abida et al., 2015). Also in plants, P deficiency resulted in the replacement of phospholipids by non-phosphorous glycolipids such as glucosylceramide, sterol glucoside and acylated sterol glucoside (Siebers et al., 2015). Consequently, sterols are synthesized and accumulate in the plasma membrane in response to P deficiency. Thus, N deficiency may inhibit 340 the capacity of the cells to synthesize sterols, while upon P deficiency membrane glycerolipid remodeling with the accumulation of non-phosphorous lipids may explain the relatively higher per-cell contents of sterols in response to P deficiency in our study.
In summary, our study shows that temperature, N:P supply ratios and pCO2 had significant separate effects on per-cell and per-carbon contents of brassicasterol/epi-brassicasterol in certain algal species. Previous studies have shown significant 345 interactions of two environmental factors on carbon-normalized contents of sterols in phytoplankton (Piepho et al., 2012;Chen et al., 2019;Ding et al., 2019). We here found potentially confounding effects of multiple environmental drivers on phytoplankton sterol contents. For example, the responses of E. huxleyi brassicasterol/epi-brassicasterol carbon-normalized contents to increasing temperature varied with N:P supply ratios, i.e., an increase under lower N:P supply ratios, but no clear changes under the highest N:P ratio (Fig. 1e), suggesting that nutrient availability may potentially alter the effects of 350 warming on sterol contents in E. huxleyi.

Responses of alkenone contents and ratios
Carbon-normalized contents of total alkenones showed non-significant responses to the changes in temperature, N:P supply ratios or pCO2 in E. huxleyi (Table 1), which can be attributed to similar response patterns of per-cell contents of alkenones and POC (Table 2). We observed that per-cell contents of alkenones changed significantly in response to N and P 355 deficiency, but showed non-significant responses to warming or enhanced pCO2. In the following, the responses of per-cell contents of alkenones and the ratios of certain alkenone isomers are discussed.
N deficiency in semi-continuous E. huxleyi cultures grown at 20% of µmax led to a 35% decrease in per-cell contents of alkenones in our study (Table 2). However, all published culture studies we could find in which E. huxleyi was grown under N deficiency were performed with batch cultures and reported either an increase in per-cell alkenone contents or a non-360 significant change (Epstein et al., 1998;Prahl et al., 2003;Bakku et al., 2018;Wördenweber et al., 2018). Several possible explanations exist for these contradictory findings. An increase in growth rate has been shown to reduce alkenone concentrations (ng mL -1 ) in continuous cultures of E. huxleyi (Sachs and Kawka, 2015). Thus, the higher growth rate of E.
huxleyi (20% of μmax) in semi-continuous cultures in our study is a possible cause of lower alkenone contents compared to the batch culture studies where cells were harvested at or near the stationary phase of growth (i.e., growth rate approaching 0) 365 (Epstein et al., 1998;Prahl et al., 2003;Bakku et al., 2018;Wördenweber et al., 2018). Because growth rate and growth phase strongly affect sterol contents (Sachs and Kawka, 2015;Chen et al., 2019), the conjunctions above-described may change if the maximum growth rates of batch cultures were used for comparison. In addition, phytoplankton in a continuous culture has a constant growth rate under a given dilution rate, while the growth rate at the stationary phase of a batch culture is zero. Thus, energy-availability for sterol remodeling differs between the two culture systems. Contradictory results 370 obtained in batch and semi-continuous cultures could indicate different alkenone contents produced by E. huxleyi during the bloom period and summertime growth of this species, respectively (Lampert and Sommer, 2007). Another possibility is that gene complements within the species of E. huxleyi vary considerably, which may explain different phenotypic variations, including differences in N and P uptake in this species (Read et al., 2013). N deficiency severely impairs the synthesis of nucleotides, amino acids and ultimately all enzymatic machinery, consequently resulting in a decrease of most central 375 metabolites (Wördenweber et al., 2018). Intense lipid turnover with the reduction of most central metabolites have been reported in E. huxleyi under N deficiency based on transcriptomic and metabolomic studies (Rokitta et al., 2014;Wördenweber et al., 2018), which may also result in lower per-cell alkenone contents in our study.
In contrast, P deficiency caused an increase (49%) in per-cell contents of alkenones in E. huxleyi (Table 2), as well as in other strains and life-cycle stages of E. huxleyi (Wördenweber et al., 2018). Experimental data presented here agree with a 380 metabolic model predicted from transcriptomic data (Rokitta et al., 2016), and the findings in a comprehensive metabolome study showing a significant accumulation of several key metabolites, especially neutral lipids such as triacylglycerols, alkenones and alkenes in response to P deficiency (Wördenweber et al., 2018). E. huxleyi contains only very small amounts of triacylglycerols and hence alkenones have been suggested to have a storage role (Volkman et al., 1980c;Bell and Pond, 1996;Eltgroth et al., 2005). Our results support this view and suggest that P deficiency can induce the accumulation of 385 alkenones which can serve as storage molecules in E. huxleyi cells. The increased abundance of metabolites in response to P deficiency is likely derived from the arrest of cell-cycling due to decreased nucleic acid synthesis, and the reduction equivalents are preserved by lipogenesis as enzymatic functionality (Wördenweber et al., 2018).
The carbon chain-length distribution of alkenones (C37/C38 alkenone ratios) showed a 13 -21% increase from the cold to warm treatments and in response to N and P deficiency, and a slight decrease (6%) with enhanced pCO2 (Table 1; Table 2). 390 Previous studies have shown that C37/C38 alkenone ratios not only varied with temperature and physiological stages (such as growth stage), but also differed between alkenone-producing species (Conte et al., 1998;Pan and Sun, 2011;Nakamura et al., 2014). In agreement with our findings, a slight increase in C37/C38 alkenone ratios at higher temperatures has been also found in four E. huxleyi strains in exponential phase cultures (Conte et al., 1998). In contrast to our results, it has been reported lower C37/C38 alkenone ratios occurred under nutrient deficiency at the stationary phase of E. huxleyi in comparison to those 395 at the exponential phase (Conte et al., 1998;Pan and Sun, 2011). As discussed above, different culturing approaches may cause conflicting results in different studies, as the effects of nutrient deficiency and growth rate cannot be well distinguished in the batch approach. The proposed biosynthetic pathways of classical C37 -C40 alkenones show that biosynthesis of C37 Me involve chain elongation with malonyl-CoA, while C38 Et are formed by the condensation of methylmalonyl-CoA and C38 Me are produced after the involvement of an additional α-oxidation (Rontani et al., 2006). Our 400 results suggest that warming, N and P deficiency and enhanced pCO2 may have independent effects on the synthesis of C37 Me and C38 alkenones, that ultimately result in changes in C37/C38 ratios.
The relative abundance of C38 homologs (C38 Et/Me ratios) showed an 82% increase from the cold to warm treatments (Table 1; Table 2). The prominent increase in C38 Et/Me ratios resulted from non-significant changes in per-cell contents of C38 Et and the decrease in C38 Me. An increase in C38 Et/Me ratios with increasing temperatures has been found in four E. 405 huxleyi strains in mid-exponential phase of batch cultures (Conte et al., 1998). More importantly, our experimental results agree well with the findings in the sedimentary records back to ~ 120.5 Ma (Brassell et al., 2004), showing the absence of all C38 Me but the occurrence of C38 Et in Cretaceous sediments (warm climate, indicating a high C38 Et/Me ratio) but the presence of C38 Me from Cretaceous to Quaternary ages (warm to cold climate, suggesting a potentially declined C38 Et/Me ratio). The strong increases in C38 Et/Me ratios with warming in our study may reflect the differences in biosynthetic 410 pathways between C38 Et and C38 Me (Rontani et al., 2006). Therefore, the distribution of alkenones in sediments over time can be linked to evolutionary adaption of alkenone biosynthesis in response to global climate change (Brassell, 2014).
The results discussed above show significant changes in per-cell contents of alkenones in response to N and P deficiency, indicating the important role of alkenones as storage molecules. Another type of biomolecules, alkenoates, has been identified in E. huxleyi and may biochemically link with alkenones (Marlowe et al., 1984;Conte et al., 1994). However, 415 alkenoates were converted to FAs by saponification in our sample preparation steps and thus not evaluated. There might be interesting variations in alkenone/alkenoate ratios with changing multiple environmental conditions, which can be assessed in future studies.

Implications for ecology and biogeochemistry
There has been evidence that carbon allocation in algal cells is highly responsive to environmental changes (Palmucci et 420 al., 2011;Halsey and Jones, 2015). Our new study demonstrated that, under future ocean scenarios (warming, N and P deficiency and enhanced pCO2), carbon-normalized contents of brassicasterol/epi-brassicasterol, alkenones and FAs have differential responses, i.e., significant but non-systematic changes in sterols and FAs, and non-significant changes in alkenones (Table 1). Our results further suggest rearrangements of cellular carbon pools under future ocean scenarios, and such variations would have important impacts on marine ecological functions and biogeochemical cycles. 425 Our study revealed an overall decrease (~ 20%) in carbon-normalized contents of brassicasterol/epi-brassicasterol in Rhodomonas sp. and E. huxleyi under ocean-related global change scenarios ( Table 2). The low availability or absence of dietary sterols has been shown to constrain growth, reproduction and survival in Daphnia (Martin-Creuzburg et al., 2005; Martin-Creuzburg and von Elert, 2009a), and development and egg production in copepods (Hassett, 2004;Klein-Breteler et al., 2005). The potential influence of sterol deficiency on ecosystem functioning is the reduction of carbon transfer efficiency 430 across autotroph-herbivore interface, leading to a low production of higher trophic levels (von Elert et al., 2003;Martin-Creuzburg and von Elert, 2009b). Brassicasterol/epi-brassicasterol are cholesterol precursors and can be converted to cholesterol by most crustaceans (Martin-Creuzburg and von Elert, 2009b;Kumar et al., 2018), and thus can efficiently support somatic growth of crustacean zooplankton such as Daphnia magna . It is therefore possible that reduced brassicasterol/epi-brassicasterol under projected future ocean conditions may have deleterious 435 ecological consequences in plankton communities, particularly where Rhodomonas or E. huxleyi is dominant.
Carbon-normalized contents of PUFAs showed an overall increase (~ 65%) in Rhodomonas sp., and an overall decrease (~ 10 -20%) in P. tricornutum and E. huxleyi (Table 2). FA composition and contents in phytoplankton have shown significant effects on zooplankton production and trophic carbon transfer from phytoplankton to zooplankton (Müller-Navarra et al., 2000;Jónasdóttir et al., 2009;Rossoll et al., 2012;Arndt and Sommer, 2014). An example of this is the 440 positive effect of increased docosahexaenoic acid (22:6n-3; DHA) content in the diatom Skeletonema marinoi on egg production rates of the calanoid copepod Acartia tonsa (Amin et al., 2011). These impacts of FA production remodelling have been discussed in detail in our previous studies (Bi et al., 2014;Bi et al., 2018;Bi and Sommer, 2020). The varying PUFA contents observed in the present study may thus potentially influence zooplankton nutrition. For example, an increase in PUFA contents in the diatoms in cold periods may have positive effects on zooplankton production; in contrast, a decrease 445 in PUFA contents in E. huxleyi at enhanced pCO2 may reduce trophic transfer efficiency at phytoplankton-zooplankton interface. Moreover, we found that the overall responses of brassicasterol/epi-brassicasterol were opposite to PUFAs in Rhodomonas sp., while both showed an overall decrease in E. huxleyi and only brassicasterol/epi-brassicasterol decreased in P. tricornutum under future ocean scenarios (Table 2). Co-limitation of sterols and PUFAs has been well studied in a freshwater herbivore D. magna, showing a negative effect on the growth of this herbivore (Martin-Creuzburg et al., 2009;450 Sperfeld et al., 2012;Marzetz et al., 2017). Less is known about how sterols and PUFAs regulate the performance of marine herbivorous zooplankton. The differential responses of sterols and PUFAs we observed may have tremendous implications for the study of marine food webs, especially in habitats where phytoplankton succession is highly dynamic with environmental changes.
Carbon-normalized contents of total alkenones in E. huxleyi showed non-significant changes ( Table 2). E. huxleyi is 455 mainly grazed by heterotrophic protists in the context of pelagic food webs (Braeckman et al. 2018;Nejstgaard et al. 1997).
Also, early studies on copepod feeding clearly showed ingestion of E. huxleyi (Harris, 1994;Nejstgaard et al., 1997;Vermont et al., 2016) and excretion of alkenones (Volkman et al., 1980a), indicating faecal pellet transport of these compounds to sediments (Volkman et al., 1980a). In the present study, we observed non-significant changes in carbonnormalized contents of total alkenones in E. huxleyi under variable temperature, N:P ratios and pCO2, demonstrating 460 quantitative applicability of alkenones as proxies for E. huxleyi biomass in biogeochemical cycles.
From a paleoceanographic perspective, lipid biomarker data have been reported on the basis of both dry sediment weight and total organic carbon (TOC) content (Zimmerman and Canuel, 2002;Zhao et al., 2006;Xing et al., 2016). Dry sediments contain variable organic and inorganic components, thus biomarker contents normalized to TOC can partially eliminate the influence of sedimentation rates, and between-site or temporal changes in the amount of organic carbon deposition or 465 preservation (Zimmerman and Canuel, 2002). The application of sediment biomarker contents for paleoproductivity reconstruction is based on two key assumptions: (1) relatively constant ratios of biomarker per cell or per POC in a specific phytoplankton group, and (2) non-significant changes in biomarker/POC ratios during post production and deposition degradation. In laboratory culture studies, carbon-normalized and cell-normalized lipid contents provide broadly similar responses to environmental parameter changes in most cases, but there are exceptions this study). 470 However, cell-normalization is difficult for sediment reconstruction, as phytoplankton cell counting is often not quantitative (Piepho et al., 2010;Ahmed and Schenk, 2017). In the present study, we focus on the implications of our findings based on organic carbon (POC)-normalized contents of lipid biomarkers. For example, our study revealed an overall 20% decrease in carbon-normalized contents of brassicasterol/epi-brassicasterol in Rhodomonas sp. and E. huxleyi in ocean-related global change scenarios, but not in the diatom P. tricornutum. Because smaller ranges are expected in temperature, N:P supply ratio 475 and pCO2 in individual locations over time, our results provide additional support for the applicability of using sterols in paleoproductivity reconstruction, especially in diatom-dominated areas. Furthermore, we observed that C37/C38 alkenone ratios varied significantly with the changes in temperature, N:P supply ratios and pCO2, indicating that besides temperature, other environmental factors may also significantly influence C37/C38 alkenone ratios. In contrast, C38 Et/Me ratios in our study responded significantly only to temperature changes. These results denote the importance to consider the effects of 480 multiple environmental factors on C37/C38 alkenone ratios, and further underline the importance of temperature in geological application of alkenone ratios.

Conclusions
The responses of sterols, alkenones and FAs to projected future scenario changes in temperature, N:P supply ratios and pCO2 were experimentally examined in three phytoplankton species. Our results reveal that N and P deficiency had a 485 stronger effect on per-cell contents of the three lipid biomarkers, while the effects of warming and high pCO2 were relatively moderate. We also show that P deficiency caused an increase but N deficiency led to a decrease in per-cell contents of lipids in most cases. Our results provide important new evidence to previous transcriptomic and metabolomic studies, which showed that key metabolites were up-regulated in response to P deficiency while most central metabolites were downregulated in response to N deficiency. Such transcriptomic and metabolomic rearrangements are linked to the regulation of 490 lipid biosynthesis-related genes (Read et al., 2013;Wördenweber et al., 2018). Future studies are suggested to consider the influence of these and other environmental changes on the composition of major and minor sterols, alkenoates and other energy storage molecules such as triacylglycerols, for example, whether varying environmental conditions influence C-24 alkylation in sterols.
Our study demonstrates that, under future ocean scenarios, the overall carbon-normalized contents of brassicasterol/epi-495 brassicasterol and PUFAs decreased in most cases in the three algal species; however, non-significant changes were also observed in brassicasterol/epi-brassicasterol and alkenones, and a significant increase was found in PUFAs in one of the three species (Rhodomonas sp.). This result highlights that a diverse allocation of carbon would potentially occur between lipid biomarkers and between phytoplankton taxa when they acclimate to large fluctuations in environmental conditions. Such variations in the contents of essential lipids (sterols and PUFAs) and in carbon allocation strategies may influence the 500 structures and functions of food webs and the future ocean ecosystems.
Data availability: Data supporting the conclusions will be publicly available at PANGAEA at the time of publication of this Kiel. This is MCTL contribution 209.
Triplicates were set for each treatment. fatty acids (μg mgC -1 ) in Phaeodactylum tricornutum, Rhodomonas sp. and Emiliania huxleyi, and C37 -C39 total alkenones in E. huxleyi under different temperatures, N:P supply ratios and pCO2. Blue, black and red symbols represent 12, 18 and 24℃, respectively. Open triangles, open circles and closed circles represent N:P molar ratios of 10:1, 24:1 and 63:1, respectively. The first two dimensions (Dims) account for 44.1% of the total variance. The length of each vector reflects the combined loading of each variable in the first two Dims (Table S5). 915