Nitrogen-induced metabolic changes and molecular determinants of carbon allocation in Dunaliella tertiolecta

Certain species of microalgae are natural accumulators of lipids, while others are more inclined to store starch. However, what governs the preference to store lipids or starch is not well understood. In this study, the microalga Dunaliella tertiolecta was used as a model to study the global gene expression profile regulating starch accumulation in microalgae. D. tertiolecta, when depleted of nitrogen, produced only 1% of dry cell weight (DCW) in neutral lipids, while starch was rapidly accumulated up to 46% DCW. The increased in starch content was accompanied by a coordinated overexpression of genes shunting carbon towards starch synthesis, a response not seen in the oleaginous microalgae Nannochloropsis oceanica, Chlamydomonas reinhardtii or Chlorella vulgaris. Genes in the central carbon metabolism pathways, particularly those of the tricarboxylic acid cycle, were also simultaneously upregulated, indicating a robust interchange of carbon skeletons for anabolic and catabolic processes. In contrast, fatty acid and triacylglycerol synthesis genes were downregulated or unchanged, suggesting that lipids are not a preferred form of storage in these cells. This study reveals the transcriptomic influence behind storage reserve allocation in D. tertiolecta and provides valuable insights into the possible manipulation of genes for engineering microorganisms to synthesize products of interest.


Results
Physiological response of D. tertiolecta under Nitrogen depletion. In the Dunaliella genus which dominates list of top strains for carbohydrate production, D. tertiolecta is one of the most researched species due to its ease of culture, tolerance to varied environmental conditions, fast growth rate, high accumulation of storage compounds per dry weight (protein, starch, lipids), and overall biomass productivity 12,31 , making it an attractive microalgal feedstock for commercial mass cultivation 32 . D. tertiolecta strain UTEX LB 999 was grown in ATCC-1174 DA medium with an initial nitrate concentration of 5 mM and 0.5 mM for N-replete and N-deplete cultures respectively. From day 1 to 15, cell density for N-replete cultures increased from 0.75 × 10 6 to 14.4 × 10 6 cells/mL, before dropping to 13.88 × 10 6 cells/mL (Fig. 1a), suggesting that cell division had ceased. On the other hand, N-deplete cultures could only reach a maximum cell density of 4.28 × 10 6 cells/mL, on day 8. The nitrate concentrations in the culture medium was depleted after 3 days for N-deplete cells, and 9 days for N-replete cells (Fig. 1a), but cell growth continued for N-replete cells, likely driven by intracellular nitrogen stores. For the purpose of this study, cells were harvested for metabolite analyses at days 3,5,8,12, and 17 to capture the passage through exponential and stationary growth phases.
Although cell division has slowed for N-deplete cells from the fourth day onwards, the increase in dry weight ( Fig. 1d) was largely due to the accumulation of storage compounds such as starch and glycerol, as the cells continue to accumulate organic carbon (Fig. 1b,c). Over the course of the experiment, N-depletion also led to reduced total chlorophyll content (Fig. 1e) and photosynthetic yield (F v /F m ) (Fig. 1f). Although there was a decrease in chlorophyll content for both N-replete and N-deplete cultures on day 3, N-replete cultures subsequently recovered chlorophyll levels to the original state (i.e. Day 0) during the exponential growth phase, while those of N-deplete cultures continue to fall (Fig. 1e). The health of photosystem II as determined by F v /F m , held steady at around 0. 68-0.74 for N-replete cultures. On the other hand, N-deplete cultures experience a drop in photosynthetic yield from 0.59 to 0.41, indicative of cell stress and possible damage to photosystem II (Fig. 1f).
Accumulation pattern of storage compounds upon N-depletion. Lipid, starch and glycerol synthesis pathways share common carbon precursors fed from the glycolytic pathway, but the regulation of carbon allocation into these routes is not well understood 33,34 . Under the favorable condition where nitrogen is available, N-replete D. tertiolecta maintained basal levels of starch (7-8 pg/cell) during the first 5 days of cultivation; consequently, when nitrogen is depleted, it increasing to 19 pg/cell on day 8 (Fig. 2a). In contrast, N-deplete cultures had starch rapidly accumulated on the third day (27 pg/cell), and the starch content of the latter remained higher during the entire process. Interestingly, starch was accumulated much earlier and in greater absolute amounts compared to lipids (neutral lipids and total lipids) (Fig. 2b,c). Notably, despite a large increase in neutral lipids (Figs 2b and 3c), N-depletion did not result in a corresponding increase in total lipids (Fig. 2c). On the contrary, it led to a cessation in its production (5.8% in N-deplete vs. 9.5% in N-replete on Day 17). Intracellular glycerol was relatively constant for both cultures throughout the experiment (Fig. 2d), with only extracellular glycerol showing marked increase, especially by N-deplete cells (Fig. 2e). This is consistent with findings in current literature implicating the continued release of glycerol by D. tertiolecta into the culture medium 35,36 . The amount of extracellular glycerol produced was able to reach higher quantities as it is not restricted by cell volume, enabling it to deliver microgram levels per cell not achievable with intracellular glycerol, starch or lipids.
When presented on a percentage dry cell weight (DCW) basis, similar trends were observed with subtle differences. Starch content of N-deplete cells peaked sharply on day 3 (28% DCW) compared to N-replete cells where it moderately increased until the media is deprived of nitrogen on day 8 (25% DCW) (Fig. 2f). Intracellular glycerol accounted for 9-14% of DCW in N-replete cells and N-deplete cells. However, in the latter stages (Day 8-17), intracellular glycerol representation in N-deplete cells dropped to 2-3% DCW as the DCW increased while intracellular glycerol content remained the same. Total lipid production was greatly attenuated in N-deplete cultures, accounting for 5.8% DCW on day 17, which is 39% less than the N-replete culture (Fig. 2f). There were minimal changes to the fatty acid (FA) composition between either cultures; C18:3 is the most abundant FA in Scientific RepoRts | 6:37235 | DOI: 10.1038/srep37235 D. tertiolecta, making up to 66% of its profile (Fig. 3a,b). The proportion of C16:0, however, appears to increase over time particularly for cells experiencing N-depletion, while those of C18:2 and C18:3 levels declined.
Analysis of gene expression by transcriptomics. The growth of D. tertiolecta in either cultures presented two distinctive phases: Exponential and Stationary (Fig. 1a). To assess the transcriptional regulation of selective carbon partitioning, we carried out transcriptome analyses at 2 times points representing N-deplete exponential phase (N-replete vs. N-deplete; both in Day 3 exponential phase) and stationary phase (N-replete exponential vs. N-deplete Day 5 stationary phase). Gene expression was expressed as fold change relative to N-replete samples. Out of all the annotated genes for Day 3 (D3) and Day 5 (D5) samples, we filtered for significantly expressed genes using a False discovery rate (FDR)-corrected p-value ≤ 0.05. This translated to 3,962 and 1,234 significantly expressed genes for D3 and D5, accounting respectively for ~17 and 6% of all genes annotated (Table S2). Among this filtered set of genes, 96% (D3) and 89% (D5) of genes were over/under-expressed (≤ − 2x or ≥ 2x fold-change) in N-deplete samples relative to N-replete samples.
Functional annotation and enrichment of differentially expressed genes. Gene Ontology (GO) enrichment of the differentially expressed genes showed that in D3, upregulated genes form the majority of categories while in D5 most genes were downregulated (Figs S2, S3). Nitrogen compound metabolic process was upregulated in both D3 and D5, an indication that the cells were increasing capacity for utilizing nitrogen in response to the nutrient's depletion. The top 10 most highly expressed or repressed genes were related to nitrogen-scavenging or photosynthesis (Table S3). For instance, THB1, a truncated hemoglobin highly expressed in the presence of nitric oxide 37 , was repressed by almost 300-fold. Likewise, three subunits of the urea active transporter were upregulated between 52-and 72-fold, representing one of the many enzymes whose expression are known to increase to harvest nitrogen from nitrogen-containing compounds such as ammonium, nitrate, nitrite, urea, purines, pyrimidines and amino acids 38 . In D3, the upregulated genes frequently represented those involved in central carbon metabolism (CCM) 39 , which includes the pathways of TCA cycle (75% enriched), glycolysis (28.6% enriched), and the oxidative pentose phosphate pathway (OPPP) (enrichment of glucose-6-phosphate dehydrogenase activity) (Fig. S2). Moreover, mitochondrial electron transport and enzymatic activity on NADH or NADPH (85-88%) were significantly upregulated, as well as pyruvate kinase activity tertiolecta was cultured in 50-mL batch cultures supplemented with 5% CO 2 . Cell growth and media nitrogen concentration (a) was measured every day. Total organic carbon (b,c) and dry cell weight (d), chlorophyll (e) and photosynthetic yield (f) were measured at different time points (Day 0, 3,5,8,12,17) to illustrate the different phases of growth. Error bars represent standard deviations from three independent biological replicates. Asterisks indicate statistically significant differences between N-replete and N-deplete samples after two-tailed t-tests (*p value ≤ 0.05; **p value ≤ 0.01).

Coordinated expression of central carbon metabolism genes for storage compound synthesis.
Photosynthetic components and antenna proteins were severely affected in N-deplete cells, as evidenced by the global downregulation of genes encoding for light-harvesting complexes, photosystem, and RuBisCO activase, an enzyme catalyzing the activation of RuBisCO (Figs 4, S8, S9). Nonetheless, carbon continued to be assimilated as seen from the upregulation of three isoforms of gamma carbonic anhydrase (CAG1, CAG2, CAG3) in D3 (Supplementary dataset), which would provide a source of bicarbonate from CO 2 assimilation. Pyruvate phosphate dikinase (PPD2) was upregulated by 23-fold on D3 and 4-fold on D5. PPD2 is used in the C4 carbon-concentrating mechanism pathway, where it converts pyruvate to PEP, which then reacts with bicarbonate to produce oxaloacetate (Fig. 4).
ATP:citrate lyase (ACLB1) and NADP-malic enzyme (MME2) were overexpressed, which would supply the cell with acetyl-CoA and NADPH respectively. In addition, glucose-6-phosphate dehydrogenase (G6PDH), a provider of reducing power from the OPPP, was also upregulated. Surprisingly, genes involved in the FA synthesis pathway were largely unchanged except for some downregulation; expression levels of malonyl-CoA-acyl carrier protein transacylase (MCAT) and acetyl-CoA carboxylase (ACCase) subunits were depressed.

Comparing the N-depletion response in D. tertiolecta and other oleaginous and non-oleaginous microorganisms.
To evaluate the differences in storage product accumulation between oleaginous and non-oleaginous microalgae during N-depletion, we compared the published physiological and transcriptomic data of three widely studied oleaginous species: Nannochloropsis oceanica 7,40 , Chlamydomonas reinhardtii 17,41,42 , and Chlorella vulgaris 43 (Figs 5 and 6). D. tertiolecta accumulates very low amounts of TAGs in N-replete and N-deplete conditions (0.2 and 1% DCW respectively), but stores a large amount of starch, thus making it a high starch and low TAG accumulator (Fig. 6, Table S4). Interestingly, although N. oceanica -a marine microalga alike D. tertiolecta -stores the same amount of TAGs under N-replete conditions, it rapidly accumulates TAGs under N-deplete conditions (up to 40% DCW); it however lacks starch as a storage compound 13,44 . C. reinhardtii and C. vulgaris, both heterotrophically grown microalgae, accumulates high amounts of starch and TAGs during N-depletion.
Analysis of the expression of key genes in the FA, TAG and starch synthesis pathways showed that upon N-depletion, the oleaginous microalga N. oceanica and moderate TAG accumulator C. vulgaris experienced downregulation in the FA synthesis pathway, a pattern similar to that in the low TAG-accumulating D. tertiolecta from this study (Fig. 5, Table S5). In contrast, C. reinhardtii had moderate overexpression of its FA synthesis genes. TAG synthesis genes had variable changes in each microalga: In D. tertiolecta GPAT was downregulated while LPAAT was upregulated, N. oceanica and C. reinhardtii genes were mostly unchanged, while C. vulgaris genes were upregulated. Noticeably, D. tertiolecta has most of its starch synthesis genes upregulated, suggesting a coordinated push towards starch accumulation during N-depletion.

Discussion
In microalgae, TAGs and starch are the two primary energy storage products and both synthesis pathways share common carbon precursors. It has been reported that oleaginous microalgae are able to accumulate   [78][79][80][81] is shown to depict an alternative route for carbon assimilation. Neutral lipid droplets found in microalgae consist mostly of triacylglycerols (TAGs), formed by combining FAs and glycerol. Legend: ACCase, acetyl-CoA carboxylase; ACD, acyl-CoA dehydrogenase; ACL, ATP-citrate lyase; ACS, acyl-CoA synthetase; AGPP, ADP-glucose pyrophosphorylase; AMY, amylase; CA, carbonic anhydrase; DGAT, diacylglycerol acyltransferase; DHAP, dihydroxyacetone phosphate; F1,6 P, fructose 1,6-bisphosphate; F6P, fructose 6-phosphate; FAT, fatty acylacyl carrier protein (ACP) thioesterase; G1P, glucose 1-phosphate; G6P, glucose 6-phosphate; G6PDH: G6P dehydrogenase, GAP, glyceraldehyde 3-phosphate; GPAT, glycerol-3-phosphate acyltransferase; GPDH, glycerol-3-phosphate dehydrogenase; MAL, malate; MDH, malate dehydrogenase; MME: NADP-malic enzyme; OAA, oxaloacetate; PDC, pyruvate dehydrogenase complex; PEP, phosphoenolpyruvate; PEPC, PEP carboxylase; PK, pyruvate kinase; Ru5P, ribulose 5-phosphate; Ru1,5BP, ribulose 1,5-bisphosphate; RuBisCO, Ru1,5BP carboxylase/oxygenase; TPT, triose phosphate translocator; 3-PGA, 3-phosphoglycerate; 6PGDH, 6-phosphogluconate dehydrogenase. Please refer to Supplementary Information for abbreviations of genes involved in starch synthesis and photosynthesis and antenna proteins.  Table S5. *ACCase denotes the Acetyl-CoA biotin carboxyl carrier subunit as it was the most reported gene among the literature. large amounts of TAGs (up to 60% of dry weight) under N-depletions [41][42][43][44][45][46] . Although these studies show that N-depletion enabled oleaginous cells to channel carbon to TAGs, little attention has been paid to non-oleaginous species which prefer alternative carbon stores such as starch and glycerol, and the transcriptomic basis governing these carbon fluxes. In this study, we tracked the storage product accumulation and global gene expression profile of the halophilic, high-starch accumulating microalga D. tertiolecta upon N-depletion, and compared its preference of storage products with other oleaginous microalgae. A coordinated upregulation of genes involved in central carbon metabolism was identified in exponentially growing cells, suggesting that the genetic response to N-depletion occurs prior to the cessation of cell growth (i.e. stationary phase). Surprisingly, the transcriptome profile of D. tertiolecta -specifically the upregulation of CCM genes contributing to the supply of acetyl-CoA and NADPH -were vastly similar to oleaginous microalgae including N. oceanica 7 and C. reinhardtii 47 . The increased transcript levels of TCA cycle genes and genes contributing to reducing power such as G6PDH and ME were correspondingly observed in heterotrophic fungi such as Mortierella alpina 48 and Mucor circinelloides 49,50 to be critical determinants of lipid synthesis.
Nitrogen is a key component for the synthesis of chlorophyll and essential proteins of the photosystem (e.g. LHCII apoprotein), thus the restriction of nitrogen supply could hinder both structural and physiological components of photosynthesis 51 . The observed drop in chlorophyll content and decreased PSII quantum yield (Fig. 1e,f) coincides with previous reports that demonstrated the degradation of chlorophyll and carotenoids in D. tertiolecta upon nitrogen starvation [51][52][53] . Corresponding downregulation of most light harvesting complex genes (Figs S8, S9) suggest that N-depletion triggered a response of the cell to reorganize its photosynthetic apparatus. Multiple studies of nutrient limitation in microalgae point to the degradation of thylakoid membranes in favour of the de novo synthesis of TAGs where intracellular membrane remodeling substantially contribute to neutral lipid accumulation [54][55][56][57] . Our present study shows that D. tertiolecta is able to accumulate up to six times more neutral lipids under N-deplete conditions (Fig. 2b), but concurrently experienced a substantial decrease in total lipid content relative to N-replete samples (9.5% of DCW in N-replete vs. 5.8% of DCW in N-deplete) (Fig. 2c). This is similar to other researchers who observe similar declines in total fatty acid content in D. tertiolecta under N-depletion (7.5% of DCW in N-sufficient vs 5.9% of DCW in N-deficient) 29 . The parallel increase in neutral lipids could be explained by a dramatic shift in intracellular processes, from the catabolic degradation of thylakoid membranes (which contribute to total lipids) to the anabolic reactions of storage compound accumulation (neutral lipids and starch content). This notion is supported by a recent report which showed that when cultured under N-depleted conditions, D. tertiolecta had significant decreases in the lipid classes of diacylglyceryltrimethylhomoserine (DGTS) and digalactosyldiacylglycerol (DGDG), a main component of chloroplast membranes 52 . This suggests the occurrence of a major remodeling of lipid membranes during nitrogen starvation in D. tertiolecta, a response akin to other microalgae [54][55][56][57] .
Glycolytic genes that shunt carbon precursors to de novo FA synthesis, including phosphofructokinase, pyruvate kinase and pyruvate dehydrogenase complex (PDC), are upregulated in N. oceanica and C. reinhardtii under N-depleted conditions correlating with TAG accumulation up to 40% DCW, despite a downregulation of the genes directly involved in TAG and FA synthesis 7,58 . This transcriptome response is unexpectedly similar to D. tertiolecta in this study (Fig. 4), which accumulates more starch (46% DCW) than lipids (Figs 2f and 5). Even more striking is the similarity in the overexpression of genes involved in the TCA cycle in the mitochondria (Fig. 4), which is also greatly enhanced in N. oceanica and was attributed to the utilization of carbon skeletons derived from membrane lipids to produce energy for TAG or its precursors 7 . As Nannochloropsis cells lack starch synthesis genes consistent with the absence of pyrenoid starch 13,59,60 , it may be logical the excess carbon and energy derived from the breakdown of membrane lipids are channeled to TAGs. However, in D. tertiolecta the response was to channel more carbon to starch instead of TAGs, as can be seen from the synchronized upregulation of genes in gluconeogenesis and starch synthesis genes (Table S5). Thus, the increased activity of CCM pathways and  Table S4 and S5 for more information on the composition of storage compounds and gene expression in N-replete and N-deplete conditions. Scientific RepoRts | 6:37235 | DOI: 10.1038/srep37235 genes contributing to the generation of acetyl-CoA and NADPH may merely be a general response of microalgae cells to N-depletion rather than a determinant of oleaginicity. Indeed, it has been previously proposed that the CCM could play varied roles upon N-depletion; the pathways could be involved in providing carbon skeletons, for nitrogen reassimilation, or for channeling excess carbon into FA synthesis 61 . Noting that most TCA cycle enzymes are bidirectional, Sweetlove et al. 62 proposed that the TCA cycle could catalyze reverse reactions and be highly adaptable to changing conditions of the cell. Hence, the TCA cycle may act as a central hub, balancing between demands for specific carbon skeletons for anabolic processes, and the energetic needs of the cell in absence of photosynthetic activity as they produce reducing equivalents for ATP synthesis.
The preference to store carbon as starch may stem from the fact that it is energetically more favourable to make than TAGs, and is the preferential reserve which is rapidly mobilized during switch from N-deplete to N-replete conditions 17 . On a per carbon basis, a 55-carbon TAG molecule requires 6.25 ATP and 2.93 NADPH, while 55 carbon units of starch requires 4.16 ATP and 2 NADPH 33 ; TAG synthesis require 50% more ATP and 45% more NADPH per carbon to make. While TAGs return more ATP than sugars during metabolism, the energy recovered is less than the energy invested in its synthesis 33,34 , chiefly due to the carbon lost during the conversion of pyruvate to acetyl-CoA by PDC. Much of this begs the question: Why do some microalgae prefer to store more TAGs and others prefer to store predominantly starch, while some are able to store moderately equal amounts of both? One possible explanation is that as lipids require more ATP and reducing power than starch for production, photoheterotrophically grown cells such as C. reinhardtii and C. vulgaris (Table S4) can generate more substrates needed for FA synthesis due to their ability to assimilate organic carbon sources on top of inorganic carbon fixation from photosynthesis. For instance, acetate can be directly converted into acetyl-CoA from acetyl-CoA synthetase (ACS), and incorporate immediately into the TCA cycle or used in FA synthesis by C. reinhardtii 34 . Likewise, Chlorella grown in acetate medium use ACS to directly incorporate acetate into acetyl-CoA, by-passing the PDC route for acetyl-CoA production and enabling a high rate of lipid synthesis under N-depletion 63 . Furthermore, glucose-fed Chlorella cultures generate more reducing power in the dark than photoautotrophically cultured Chlorella in the light, thus the former accumulates more lipids and less starch 64 . This explanation does not extend to Nannochloropsis however, due to its inability to synthesize starch 13,59,60 . Rather, it stores small amounts of carbohydrates (5-17%) while channeling majority of carbon to TAGs during N-depletion.
Microalgae of the genus Dunaliella dominate the ranks of top carbohydrate producers, and D. tertiolectawith the ability to accumulate up to 63% DCW in carbohydrates -leads the list 12 . In this study, we show that D. tertiolecta rapidly accumulates starch in response to N-depletion, suggesting that D. tertiolecta cells tend to favor starch over TAGs as an efficient strategy for chemical energy accumulation (Fig. 2a,f). As an obligate photoautotroph that cannot use dissolved organic compounds 32,65 , efficient energy storage and utilization is of paramount importance, especially since photosynthesis is inefficient with less than 8% conversion of solar energy chemical energy 33 . Notably, starch reserves provide the energy required in the dark for DNA replication and general cell metabolism 66 . This is in line with the increase in gene expression of starch-degrading enzymes in D. tertiolecta during N-depletion (Fig. 4, Fig. S7), when photosynthesis is compromised. D. tertiolecta also specifically degrade starch in high salinity stress conditions to yield DHAP which can be used to produce glycerol 67 , making starch a more suitable storage compound as the cells can respond to both nutrient deficiency and salinity stress, conditions which could fluctuate greatly in their natural environments 68 . In light of these observations, it is reasonable to assume that carbon would be channeled to TAG production once starch synthesis is blocked. However, this notion was refuted as there was no increase in oil among three starchless C. reinhardtii mutants examined 17 , suggesting that blocking starch synthesis does not result in oil accumulation. Preference of starch and TAG accumulation in microalgae may instead stem from natural evolutionary patterns and exposure to different carbon sources.
D. tertiolecta was previously reported to accumulate and release glycerol into the external medium, and that the process was not affected by nutrient starvation or cell death 36 . In this study, we show that under N-depletion increased amounts of glycerol were present in the extracellular medium (Fig. 2e), indicating that glycerol synthesis and export is a response to nutrient stress. This is supported by the marked upregulation of a triose phosphate translocator, which exports photosynthetically fixed carbon or carbon derived from starch breakdown from the chloroplast to the cytosol, where the enzyme GPDH converts into glycerol (Fig. 4). D. tertiolecta is known to uptake exogenous glycerol in response to stress 69 , and as releasing glycerol to the external medium provides the cells with an expansive carbon sink not restricted by cell size, the extracellular glycerol may serve as a readily utilizable carbon source for the cells in anticipation of improving nutrient environments 36 .
The present study identifies the unique transcriptome signatures of D. tertiolecta, a high starch-accumulating microalga in response to N-depletion. In particular, its coordinated upregulation of genes involved in starch synthesis could be attributed to its preference to store starch over TAGs. However, D. tertiolecta also shares similar upregulation of CCM genes as the oleaginous microalga N. oceanica, where CCM pathways were suggested to provide substrates for FA and TAG synthesis during N-depletion. The perplexing similarities and differences in transcriptomic and metabolic responses of oleaginous and non-oleaginous microalgae thus demands a relook into the role of biochemical pathways governing the allocation of carbon and energy in the cell. Addressing these issues would advance our understanding of starch and lipid synthesis in microalgae as well as facilitate the genetic engineering of microorganisms designed to accumulate either storage product of interest.  (Table S1). The cells were grown in 50-mL batch cultures on a rotary shaker at 25 °C and illuminated with 30 μ mol photons m 2 /s under a photoperiod of 14 h Light/10 h Dark, with or without 5% CO 2 aeration. Cell densities were determined using an automated cell counter (TC20 TM automated cell counter, Biorad Laboratories). Prior to counting, D. tertiolecta cells were fixed with 2% paraformaldehyde. Optical density measurements (OD 680 ) were conducted with an UV spectrophotometer (Genesys 10 S UV-VIS, ThermoFisher Scientific). Biomass was determined by dry cell weight (DCW) (g/mL) measurement and combined with cell density (cells/mL) to get dcw per cell (g/cell). Ten-milliliter (10 mL) of cells were collected by filtration on preweighed Advantec GB-140 filter paper (0.4 μ m pore size; diameter 47 mm), and washed with isotonic 0.5 M ammonium formate (40 mL) to remove salts without causing the cells to burst. Cells captured on filter paper discs were dried in oven at 95 °C until the weight was constant. Work conducted throughout the paper is based on biological triplicates unless otherwise stated.

Methods
Nitrogen Depletion and Cultivation. D. tertiolecta subjected to nitrogen depletion (N-depletion) were grown in nitrogen-limited media (ATCC-1174 DA medium containing 0.5 mM KNO 3 , equivalent to 10% of original KNO 3 concentration; K substituted with KCl) (Table S1). N-depletion was achieved by harvesting exponentially growing cells (cell density approximately 3 × 10 6 cells/mL) and twice washing them with fresh ATCC medium containing 10% KNO 3 . All experimental cultures began at an OD 680 of 0.1 for standardization. Nitrate concentration was determined using a UV spectrophotometric method measuring the difference between OD 220 and OD 275 , and converted with a standard curve (Fig. S1). Cells were harvested at regular time intervals corresponding to exponential, late-exponential and stationary phases for RNA extraction, neutral lipid analysis, fluorescence microscopy, and biomass measurements. Experiments were conducted as independent biological triplicates.
Total Organic Carbon, chlorophyll, photosynthetic yield, starch and glycerol measurements. Intracellular total organic carbon (TOC) content was measured using an automated High Temperature Combustion TOC Analyzer (LOTIX; Teledyne Tekmar), installed with a Non-Dispersive Infrared Detector. Cell densities of 3 × 10 7 cells were harvested, spun down and washed with 1 mL of 0.08 M NaCl, reconstituted in ultrapure water to a final volume of 20 mL, and transferred to the TOC analyzer for analysis.
Starch content was measured using the Starch Assay Kit (STA20; Sigma-Aldrich), performed according to the manufacturer's instructions. Glycerol content was measured using the Free Glycerol Determination Kit (FG0100; Sigma-Aldrich) as previously reported by Chow et al. 35 . Harvested cells were centrifuged (10,000 g for 10 mins, 4 °C) and the cell pellet and supernatant were separated for measurement of intracellular and extracellular glycerol respectively. For intracellular glycerol, the cell pellet was washed with equivalent volume of 0.5 M NaCl and centrifuged again to remove remaining extracellular glycerol. The cell pellet was then resuspended in 200 μ L of ultrapure water, vortexed and boiled for 15 mins. Subsequently, the samples were centrifuged to remove cell debris, and the supernatant collected for intracellular glycerol analysis.
Photosynthetic yields (maximum efficiency of photosystem II; F v /F m ) were evaluated using chlorophyll fluorescence measured with AquaPen-C fluorometer (AP-C 100/USB; Photon Systems Instruments). Cells were dark-adapted for 5 mins before using the fluorometer to measure F v /F m values according to the user's manual.
For chlorophyll measurements, cells were centrifuged (10,000 g for 10 mins, 4 °C), supernatant were removed, and the cell pellet resuspended in 0.5 mL DMSO and vortexed for 1 min until the pellet disintegrated. An equivalent volume of 90% acetone was added and mixed well, followed by centrifugation to remove cell debris. The supernatant (extraction volume of 800 μ L) was used for measuring absorbances at 630 nm, 647 nm, 664 nm, 665 nm, and 750 nm using a UV spectrophotometer. Correction of pheopigments was done by adding 40 μ L of 1 M HCl to To the mixture and incubating at room temperature for 90 secs. Absorbances were measured again at 665 nm and 750 nm. Chlorophyll amounts were calculated according to trichromatic equations 70  R= maximum absorbance ratio of OD665o/OD665a in the absence of pheopigments = 1.7 K = R/(R− 1) = 2.43 Total chlorophyll (μ g/mL) = Chlorophyll b (μ g/mL) + Corrected chlorophyll a (μ g/mL)

Neutral Lipid Quantification. A modified Nile Red staining method 28 was used to quantify intracellular
TAGs. Briefly, cells were harvested by centrifugation (3000 g for 10 min at 4 °C), supernatant was removed and the pellet resuspended in fresh 0.5 M ATCC-1174 DA media to an OD 680 of 0.3. Triolein was used as a standard for determining neutral lipid concentrations by Nile Red (Fig. S10). Two hundred microliters of triolein standards (40, 20, 10, 5, 2.5, 0 μ g/mL) and cell suspensions were loaded as technical triplicates onto a 96-well black, clear bottom plate (CLS3603; Sigma-Aldrich). Prior to staining, Nile red stock is diluted in acetone to obtain a working solution (25 μ g/mL), and 2 μ L of the Nile red working solution is added to each well of sample and standard, followed by a 5 min incubation in the dark. Fluorescence of each sample was detected using a microplate reader (Infinite M200 PRO, Tecan) at excitation and emission wavelengths of 524 nm and 586 nm. Fluorescence imaging of Nile Red-stained cells was performed with an automated fluorescence microscope (Olympus BX63). Acquisition and processing of data was done using the cellSens software.

Total Lipid Analysis by Gas Chromatography-Mass Spectrometry.
To analyze the accumulation of total lipids, cells were harvested, snap-frozen in liquid nitrogen and stored at − 80 °C until analysis. Frozen culture samples were lyophilized by freeze-drying and lipids were extracted by hexane using direct transesterification 72 as it was reported to be a convenient and accurate method for analyzing total fatty acids 73

RNA-Seq and Differential Gene Expression Analysis.
The cDNA libraries were normalized to 10 nM, pooled in equal volumes, and sequenced for 2 × 300-bp runs (paired-end) using Illumina MiSeq Sequencer (Illumina). A set of four cDNA libraries were sequenced per run in order to generate sufficient reads for each sample. FASTQ datasets generated from Illumina Miseq were uploaded into a Partek ® Flow ® web server (version 4.0, Partek Inc.). To ensure quality, the raw data were trimmed from both ends with the following criteria: Phred-equivalent quality score of more than 20, minimum read length of 25, quality encoding = autodetect. The filtered reads were aligned with STAR aligner (version 2.3.1j; default parameters) 75 to an assembled D. tertiolecta transcriptome database previously created by our colleagues 28 . Data aligned to the transcriptome from STAR were used to test for differential expression of genes between N-replete and N-deplete samples. This was performed at transcript level with the Gene-specific analysis (GSA) approach from Partek ® Flow ® (Poisson model was selected) and normalized to RPKM (Reads Per Kilobase per Million mapped reads) values. Genes were considered to be significantly differentially expressed if their expression values had at least a fold change greater than ± 2, a false discovery rate (FDR)-corrected p value of ≤ 0.05 (Benjamini-Hochberg step-up correction), and the read coverage at either of the culture conditions were ≥ 10. The RNA-seq raw data were deposited in the Sequence Read Archive database (http://www.ncbi.nlm.nih.gov/Traces/sra/) under the accession numbers SRR4011621, SRR4011622, SRR4011623, SRR4011624, SRR4011625, SRR4011626, SRR4011627, and SRR4011628.
Functional Annotation and Biological Interpretation of RNA-seq data. Functional annotation of cDNA reads was performed with the Partek ® Genomics Suite ® (PGS) software (version 6.6, Partek Inc.). The GSA file containing filtered genes and their associated information was imported into PGS and merged with the assembled D. tertiolecta transcriptome annotation file according to the name of D. tertiolecta contigs. The annotated data was subsequently used to perform Gene Ontology (GO) enrichment with a modified C. reinhardtii GO annotation file 28 downloaded from JGI website (http://jgi.doe.gov/). Representative pathways were discovered by mapping the gene names to the C. reinhardtii Kyoto Encyclopedia of Genes and Genomes (KEGG) database (http://www.genome.jp/) using the pathway analysis tool 76,77 . Both GO enrichment and KEGG pathway analyses were conducted using the Fisher's Exact test; the analysis was restricted to pathways with more than 2 genes, and results were filtered by enrichment p-value of less than 0.05.