Core features of triacylglyceride production in Ettlia oleoabundans revealed by lipidomic and gene expression profiling under distinct induction conditions

Algal Research xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Algal Research journal homepage: www.elsevier.com/locate/algal

Core features of triacylglyceride production in Ettlia oleoabundans revealed by lipidomic and gene expression profiling under distinct induction conditions Falicia Qi Yun Goha,b, Justin Jeyakanib, Amaury Cazenave-Gassiotc, Phornpimon Tiptharab, Zhenxuan Yeoa,b, Markus Wenkc,d, Neil D. Clarkea,b,d,⁎ a

Yale-NUS College, Singapore 138527, Singapore Computational and Systems Biology, Genome Institute of Singapore, Singapore 138672, Singapore Department of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore, Singapore 117596, Singapore d Department of Biological Sciences, National University of Singapore, Singapore 117543, Singapore b c

A R T I C L E I N F O

A B S T R A C T

Keywords: Triacylglycerides Membrane lipids Gene expression Chlorophyll Osmotic stress Nitrogen limitation

Many algae produce triacylglycerides (TAG) when starved for assimilable nitrogen. In the model organism Chlamydomonas reinhardtii, and probably in other species, nitrogen starvation is also the cue to prepare for sexual reproduction. The ensuing process of gametogenesis entails a host of physiological and gene expression changes, which makes it difficult to distinguish the changes that are peculiar to TAG accumulation from other aspects of gametogenesis. As others have found, we show here that the chlorophyte Ettlia oleoabundans produces TAG under conditions of nitrogen deprivation and elevated NaCl concentrations. We exploit these two conditions, and the intermediate response of the organism to intermediate levels of stress, to identify physiological and gene regulatory features in common. Strikingly, TAG levels and chlorophyll concentrations are inversely correlated across both sets of conditions. Similarly, membrane lipids undergo related compositional changes under the two conditions. In contrast, RNA-seq analysis reveals substantially different expression profiles. This is useful because the gene expression changes that are in common, against this background of expression differences, are more likely to be relevant to the shared changes in physiology and lipid composition. Gene expression changes in common include transcripts related to fatty acid synthesis and degradation, TAG synthesis, and a putative TAG lipase that is substantially down-regulated. The identification of properties that are shared by distinct TAGinducing conditions could facilitate the engineering of algal strains with improved TAG production properties.

1. Introduction The production and storage of triacylglycerides (TAG) by algae has been the subject of intensive study [1–3]. The impetus for much of this research has been the prospect of algae becoming a feedstock for biodiesel production that is more sustainable than the likes of soybean and palm oil [4,5]. Substantial challenges remain in achieving this goal [6–10]. However, the production and accumulation of TAGs by algae is a fascinating metabolic phenomenon in its own right. Why is it that cells starved for a nutrient accumulate substantial quantities of such an energy-rich molecule? Why is it that nitrogen starvation appears to be a universal inductive signal for algal TAG accumulation, more so than the limitation of other essential compounds such as phosphate or sulfate? The TAG-producing species about which we know the most is Chlamydomonas reinhardtii [3]. Unfortunately, nitrogen starvation in

⁎

Chlamydomonas is not just an inducer of TAG synthesis, it is also the signal for gametogenesis, a complex set of physiological, morphological, biochemical and gene regulatory processes that prepare Chlamydomonas cells for sexual reproduction [11,12]. Whether TAG production and storage is an adaptive part of the sexual life cycle is not clear. For most algal species that produce TAGs upon nitrogen starvation we know nothing about the conditions under which sexual reproduction can occur. It is possible that all of these organisms initiate a process similar to Chlamydomonas gametogenesis, but it is also possible that nitrogen starvation and sexual reproduction are uncoupled (if sexual reproduction happens at all). Perhaps TAG synthesis is simply a way of dealing with excess photosynthetic reducing power that, under normal growth conditions, would be put into nitrogen-containing macromolecules such as proteins and nucleic acids. Without TAG synthesis as a sink for photosystem-generated reducing power, reactive

Corresponding author at: Yale-NUS College, Singapore 138527, Singapore. E-mail address: [email protected] (N.D. Clarke).

http://dx.doi.org/10.1016/j.algal.2017.06.014 Received 14 January 2017; Received in revised form 8 June 2017; Accepted 20 June 2017 2211-9264/ © 2017 Elsevier B.V. All rights reserved.

Please cite this article as: Goh, F.Q.Y., Algal Research (2017), http://dx.doi.org/10.1016/j.algal.2017.06.014

Algal Research xxx (xxxx) xxx–xxx

F.Q.Y. Goh et al.

pass filter.

oxygen species could be more of a problem for the organism. The challenge in studying TAG synthesis is to separate TAG production and storage from the other physiological and gene regulatory events that occur during nitrogen limitation. Here, we take advantage of a chlorophyte alga, Ettlia oleoabundans, that produces substantial quantities of TAGs under two very different conditions: nitrate starvation and elevated concentrations of NaCl. The discovery that Ettlia oleoabundans produces TAGs, including the observation that TAGs are produced at moderate salt concentrations and elevated pH, was made previously [13–15]. A partial analysis of lipid composition has been reported for Ettlia grown under conditions of elevated pH and ionic strength, and gene expression changes have been reported for cells starved for nitrate [16,17]. However, to our knowledge there has not yet been a comprehensive comparison of lipid metabolism, morphological and physiological changes, and gene regulation performed under different TAG-producing conditions. This is true not just for Ettlia, in fact, but for any algal species. In part this is because few species are known to produce substantial amounts of TAG under any condition except nitrogen starvation. The experiments described here have allowed us to identify a limited number of features that are associated with TAG synthesis in two very different physiological backgrounds. The dissimilarity of the conditions used to induce TAG accumulation, and the differences in the response of the organism to those stresses, gives added weight to those aspects of the response that are in common.

2.5. Lipid analyses Triacylglycerol (TAG), phosphatidyl-glycerol (PG), phosphatidylcholine (PC), phosphatidyl-ethanolamine (PE) sulfoquinovosyl-diacylglycerol (SQDG) and monogalactosyl-diacylglycerol (MGDG) were quantified by mass spectrometry (details provided as Supplemental Information). Lipids were extracted as described but using 2:1 chloroform:methanol [19]. Both standards and dried algal extract were dissolved in chloroform:methanol 1:1 (v:v) with a final injection volume of 20 μL. Quantifications were determined using area under curve (AUC) of the relevant ion chromatogram peaks for (a) a SIM analysis for TAG or (b) MRM analysis based on precursor ion to head group fragment transitions for PG and SQDG (81 MRM transitions) and PC and MGDG (22 MRM transitions). Normalizations were performed to internal standards except for MGDG and SQDG for which appropriate standards were unavailable. Standards were purchased from Avanti Polar Lipid: deuterated tripalmitoyl-glycerol (d5-TAG) for TAG; dimyristoylphosphatidic acid (DMPA), dimyristoylphosphatidyl glycerol (DMPG) and dimyristoylphosphatidyl ethanolamine (DMPE) for PG and dimyristoylphosphatidylcholine (DMPC) for PC. For MGDG, MRM transition intensities were normalized to the sum of the intensities for the standard-normalized species. For SQDG the signal was normalized to DMPE which elutes at a similar retention time. Essentially all TAG mass species could be explained by one C16 and two C18 fatty acids of varying saturation, or by two C16 and one C18 fatty acid, consistent with the preponderance of C16 and C18 fatty acids in Ettlia [15,20]. The methodology could have detected TAGs with three C16 or three C18 fatty acids had such TAGs been present at reasonable levels but none were found. Further details on the chromatography and mass spectrometry are provided as Supplemental Information.

2. Materials and methods 2.1. Strains and cultivation conditions Neochloris oleoabundans (UTEX # 1185) (now Ettlia oleoabundans) was obtained from the Culture Collection of Algae at the University of Texas (Austin, TX, USA). Cultures were grown photo-autotrophically in a modified Bold Basal Medium with 3-fold nitrogen (9 mM NaNO3) plus vitamins B1 and B12. Cultures were inoculated with 1 × 106 cells/ml late log phase cells into 250 ml Erlenmeyer flasks containing 100 ml of medium and grown at 25 °C under white light irradiation (22–28 μmol photons m− 2 s− 1), while being rotated at 100 rpm. NaNO3 and NaCl concentrations were varied as described in the text.

2.6. Replication of assays For each of the ten growth conditions, three cultures, grown months apart, were characterized by cell count, mean cell diameter, chlorophyll content, and Nile Red fluorescence. Aliquots were frozen for subsequent analysis by mass spectrometry; preparation of lipids from the frozen aliquots was done in parallel for the different biological replicates to minimize the effects of variation in solvent extraction. The values reported for lipid species, chlorophyll, and Nile Red fluorescence are the means of the three biological replicates, normalized (where described in the text) by cell count, or by both cell count and average cell volume. For each biological replicate, chlorophyll and Nile Red fluorescence values were themselves the means of at least three technical replicates. RNA-seq was performed on a single biological replicate for each of the ten conditions. However, the transcriptome analysis is effectively based on multiple samples because PCA was used to reduce a set of nine values for each transcript (fold-differences in expression relative to control) to just two (PCA dimensions 1 and 2). The values that characterize a transcript's gene expression are, in a sense, based on all nine of the fold-difference values, or an average of 4.5 values for each PCA coordinate.

2.2. Chlorophyll measurements Chlorophyll fluorescence was measured in dimethyl sulfoxide (DMSO) using a fluorescence plate reader (SpectraMax® M5 MultiMode Microplate Reader; Molecular devices LLC, California, USA). The excitation wavelength was 430 nm and the detection wavelength was 690 nm. The concentration was determined from a standard curve generated from chlorophyll-a obtained from Anacystis nidulans. 2.3. Quantitation of neutral lipids by Nile Red Nile Red fluorescence was used to estimate the quantity of triacylglycerides in cells [18]. Cells were fixed with Prefer solution (Anatech Ltd., Michigan, USA) for 10 min, washed and re-suspended in 50 mM Tris-HCl pH 7.0, and stained with Nile Red (Sigma, Saint Louis, USA) for 10 min in the dark at a final concentration of 2.5 μg/ml. Fluorescence was measured at an excitation wavelength of 488 nm and detected at 575 nm.

2.7. Construction of sequencing libraries Cell pellets were obtained by centrifugation at 10,000g for 5 min at 4 °C, snap-frozen in liquid nitrogen and transferred immediately to − 80 °C until ready for RNA extraction. RNA was extracted using a hot phenol method [21], and quality assessed with an Agilent 2100 bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Purification of poly-adenylated mRNA RNA was carried out using two rounds of hybridization to Dynal oligo(dT) magnetic beads (Invitrogen, Carlsbad, CA, USA). The resulting mRNA was used to construct Illumina sequencing libraries using the mRNA-Seq Kit (Illumina, San Diego, CA,

2.4. Microscopy Lipid bodies stained with Nile Red were visualized using a Nikon A1Rsi Fast Laser Scanning and Spectral Confocal microscope (Nikon, Tokyo, Japan). Excitation was at 488 nm; detection employed a 560–615 nm band pass filter. Calcofluor white fluorescence was detected by excitation at 633 nm and emission at 650 nm using a long 2

Algal Research xxx (xxxx) xxx–xxx

F.Q.Y. Goh et al.

Fig. 1. Triacylglyceride (TAG) production following six days of phototrophic growth under control and stress conditions. (A) Media differed in the concentration of nitrate and NaCl, as indicated. The symbols shown for growth conditions are used throughout this paper; control media is indicated by an open diamond, variations in nitrate concentration are indicated by squares, and variations in NaCl concentration are indicated by circles. Increased shading of the symbol indicates increasing stress (decreasing nitrate or increasing NaCl). Micrographs show cells grown under each condition using DIC (differential interference contrast) or fluorescence staining. Nile Red stains lipid droplets (red); calcofluor white stains cell walls (green). The squares below the micrographs represent relative amounts of TAG as inferred from Nile Red fluorescence (red) and mass spectrometry (yellow). Areas of these squares are directly related to the quantities measured, scaled for each assay to the amount found in 0 mM nitrate. The extreme nitrate and salt stresses are labeled N and S for reference to panel B. (B) Quantitation of TAG by mass spectrometry. The labels for the TAG species refer to the sum of fatty acid carbon atoms and the double of double bonds. For example, 50:3 has 50 carbons (probably 16, 16 and 18 per fatty acid) and 3 double bonds. Values are the means of three biological replicates; error bars show the standard error. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

A

B

families). As a consequence of these filters, all of the transcripts encoded protein sequences and for all there was at least one condition in which the transcript (or ≥ 1 member of the scaffold family) had a significant expression difference from the control (edgeR as described above; FDR = 0.05).

USA), according to manufacturer's protocol. Adaptor-ligated libraries were PCR-amplified for 15 cycles and the amplicons purified (QIAquick PCR purification kit, Qiagen Inc., Valencia CA, USA). The size distribution and concentration of the libraries were determined on an Agilent 2100 bioanalyzer. Eight of the libraries were constructed and sequenced in an initial set, and used for transcriptome assembly as described below. The remaining two libraries, corresponding to growth in 0.375 mM NaNO3 and 150 mM NaCl, were prepared on a later occasion. 75 bp paired-end sequencing was performed on all ten libraries.

2.9. Annotation Transcripts were annotated by sequence homology searches, principally using BLAST2GO [26]. Transcripts were associated with Gene Ontology (GO) terms, enzyme commission (EC) numbers (KEGG) and protein sequence signatures (InterproScan). We also used blastx and blastn to identify homologs in the genomes of Chlorella variabilis and Chlamydomonas reinhardtii rRNA genes were identified through the RNAmmer server [27,28]. For transcripts whose automated annotation suggested membership in functional groups of interest, such as the KEGG-defined carbon and lipid metabolism pathways described in the text, we manually inspected the output of blastx searches against the NCBI database. This was partly to get a sense for the consistency of annotations and partly to eliminate a small number of transcripts that contained more than one ORF, almost certainly the consequence of assembly errors. Lipase candidates were identified through an additional search procedure. Transcript sequences were translated by TransDecoder and used as input to HMMER to search for matches to ten Pfam protein domain HMMs (Abhydro_lipase, Hydrolase_4, Lipase, Lipase_2, Lipase_3, Patatin, PEARLI-4, Phospholip_A2_1, Phospholip_A2_3, and Phospholip_B) [29–31]. Hits to these HMMs were then evaluated by manual inspection of the output from blastx searches of the transcript run against the NCBI nr database. All BLAST and HMM searches were conducted using default parameters.

2.8. Transcriptome assembly and RNA-seq analysis An Ettlia oleoabundans transcriptome was assembled with Trinity using 20% of the ~109 reads obtained across eight of the growth conditions [22]. The initial assembly contained ~113,000 scaffolds. Reads from each of the experimental conditions were mapped back to the scaffolds using Bowtie (v0.12.8) and transcript abundances determined with the RSEM package (1.2.0) [23,24]. Differential expression analysis was carried out using edgeR, comparing each of the nine ‘stress’ conditions to the control (v 3.0.8 from Bioconductor 2.11) [25]. TMM normalization and an edgeR dispersion value of 0.1 were used for this analysis. Scaffolds that share the first two identification numbers assigned by Trinity (i.e., differing only in the third number in ID's that are of the form < num > _c < num > _seq < num > ) were deemed to belong to the same scaffold family. Some of these may be authentic variants, but others are likely to be sequencing or assembly artifacts; consequently, we treated scaffold families as individual transcripts, summing their RPKM values for purposes of gene expression analysis. Additional filters were applied as described in Supplemental Information, yielding a final set of 10,122 transcripts (scaffolds or scaffold 3

Algal Research xxx (xxxx) xxx–xxx

F.Q.Y. Goh et al.

3. Results

A

3.1. Screening algae for alternative TAG-accumulation cues Before settling on Ettlia oleoabundans for this work, we first performed a screen of 25 algal species previously selected for analysis as part of an on-going project on plant and algal transcriptomes (https:// sites.google.com/a/ualberta.ca/onekp/; Supplemental Information Table 1). The algal species were assessed for photo-autotrophic growth rate and for TAG accumulation. All cultures were grown in non-carbon containing medium suggested by the culture collection from which the species was obtained; nitrogen starvation entailed washing and resuspending cells in the same medium except for the omission of whatever nitrogen source(s) were present in the standard media (Materials and methods). The production of TAG was determined by Nile Red fluorescence [18]. Strains that grew photo-autotrophically and produced TAG when switched to media lacking nitrogen were subjected to further screening to identify additional stress conditions that might induce TAG synthesis. Our preference was to find conditions as unrelated to nutrient starvation as possible. In the course of this screen, we discovered that Ettlia oleoabundans (Neochloris oleoabundans; UTEX 1185) produces TAGs at moderate ionic strengths (600 mM NaCl) in media that is otherwise standard (Materials and methods; Fig. 1). Ettlia was the only species screened that showed this effect. In preliminary studies, we found that TAG accumulation and cell growth plateaued at around six days under both nitrate starvation and elevated NaCl conditions. We therefore fixed the induction time at six days, varying the degree of stress over that period. In particular, all experiments described here were performed on cultures grown for six days in normal media (9 mM nitrate and 42 mM NaCl), at five lower nitrate concentrations (3.0 mM, 1.5 mM, 0.75 mM, 0.375 mM, 0 mM), and at four higher salt concentrations (75 mM, 150 mM, 300 mM, 600 mM).

B

3.2. Quantitation of TAG levels by mass spectrometry Both salt stress and nitrate starvation induce lipid droplet formation as determined by Nile Red staining. Quantitation of the fluorescence signal shows a ~10× stronger signal for extreme nitrate starvation than was observed at the highest salt concentration (Fig. 1A). However, mass spectrometry shows that TAG levels per cell are indistinguishably high at the two extreme conditions (Fig. 1A). The fatty acid compositions of TAG are also indistinguishable under these two conditions (Fig. 1B). In particular, TAGs under both conditions have an apparent fatty acid composition of 16-16-18 or 16-18-18, and the amount of each, and the degree of saturation, is not significantly different. There are reasons to believe that the mass spectrometric quantitation of TAG is more reliable than Nile Red fluorescence [21] (Fig. 1). First, Nile Red localizes to concentrated regions of non-polarity (i.e. lipid droplets); to the extent there may be TAGs that are not localized to lipid droplets, these will be poorly detected by Nile Red staining. Second, there is no reason to expect that the Nile Red signal should increase linearly with lipid droplet volume. Third, as a fluorescence probe, the signal is affected by other chlorophyll and other chromophores whose absorbance spectrum overlaps with the absorption or emission spectra of Nile Red. This is especially problematic because the concentration of chlorophyll itself decreases under increasing stress (Fig. 2A).

Fig. 2. Chlorophyll-a and TAG levels. (A) Bar-chart showing the amount of chlorophyll-a per cell, normalized to cell volume, as calculated from the mean cell diameter. The unnormalized amount of chlorophyll per cell is represented by the area of the squares above the bar-chart. Values are the means of three biological replicates; error bars show the standard error. (B) For both conditions, chlorophyll-a and TAG levels are inversely correlated (R2 = 0.84; p = 10–4). The dashed lines define a 95% confidence envelope for the linear fit.

unexpected. Across the full range of salt concentrations, we find an inverse relationship between cell volume and the number of cell divisions that took place, implying a trade-off between cell division and cell growth that shifts gradually toward fewer divisions and large cell volume at high salt concentrations (Fig. 3).

3.4. Physiological similarity: chlorophyll concentrations and TAG amounts are inversely associated As noted above, absolute chlorophyll levels differ with the type of stress, increasing several fold in moderate salt concentrations and decreasing in nitrate starved cells (Fig. 2A). However, when normalized to cell volume, a monotonic decrease in chlorophyll is observed for both stresses, the sole exception being the first step away from the control (Fig. 2B). Intriguingly, we find that chlorophyll concentration is inversely correlated with TAG concentrations across the full set of ten conditions (R2 = 0.84, p = 1.1 × 10− 4). This inverse correlation holds true for each of the stresses individually (R2nitrate = 0.81; R2salt = 0.66).

3.3. Physiological difference: cell division and cell volume For the most extreme stress conditions, there was no change in cell number over the six-day period, but the volume of the cells increased. The increase in cell size was modest in the case of nitrate starvation (~ 1.5-fold) but dramatic in high salt (~ 10-fold). Nutrients are no more limiting in the salt stress condition than they are under normal growth conditions so an increase in biomass for this condition is not 4

Algal Research xxx (xxxx) xxx–xxx

F.Q.Y. Goh et al.

A

analyses [39] (see Materials and methods for details). Expression profiles for the ten conditions were compared by hierarchical clustering and by principal components analysis (PCA). Cultures grown in the three highest nitrate concentrations (1.5 mM, 3 mM and the control, 9 mM) were very similar to one another, clustering together by correlation coefficient and by PCA (Fig. 5A and B). The three lower nitrate concentrations are distinct from one another but cluster together when compared to the salt conditions. An advantage of PCA is that transcripts can be projected onto the same coordinate system as the growth conditions. This allows transcripts to be identified that play an exceptionally large role in defining the various transcriptional states. For example, transcripts involved in nitrogen transport and assimilation are up-regulated under conditions of nitrate starvation (Fig. 5C and D; purple circles) and, as a result, the transcripts project onto the first two PCA dimensions in the same direction as the nitrate starvation conditions. (squares in Fig. 5B). In contrast, transcripts encoding light harvesting proteins are down regulated under nitrate starvation; these down-regulated transcripts project onto the PCA plane on the opposite side of the origin from the upregulated nitrogen assimilation transcripts (Fig. 5C and D; yellow circles). Similarly, transcripts related to cell wall catabolism are upregulated in salt stress and map in the same PCA direction (Fig. 5C and D; blue circles) as the salt-stressed conditions themselves (Fig. 5A, gray and black circles). In contrast, two transcripts of unknown function are strongly down-regulated in salt; on the PCA plane these transcripts map opposite to the cell-wall, salt-up-regulated, genes (Fig. 5C and D; green circles). As a final example, there are a couple of transcripts involved in metal transport that are strongly down-regulated under both stress conditions. These transcripts map onto the PCA plane in a direction nearly identical to the control condition (Fig. 5A; open diamond). That is, they map in a direction opposite to that which would be associated with up-regulation under both conditions. These are examples of transcripts that contribute most to the total variance in gene expression across the two stress gradients. However, gene expression changes can be biologically important even when the magnitude of the change is much smaller than it is for these transcripts. We therefore turned our attention to transcripts whose predicted functions are relevant to carbon metabolism regardless of the magnitude of their expression change. We started with transcripts that had been annotated with enzyme names or EC numbers found in any of the following KEGG-defined pathways: glycerolipids, fatty acid biosynthesis, fatty acid elongation and degradation, glycolysis and gluconeogenesis, TCA cycle, pentose-phosphate pathway, carbon fixation in C3 organisms, dicarboxylic acid cycle (C4), and starch metabolism [40]. Nearly 500 such scaffolds or scaffold families were identified. Each was further assessed by manual inspection of BLAST searches. In cases where more than one transcript shared the same name or EC number, their records were merged and their RPKM values summed. In total, 116 merged transcripts for metabolic enzymes were projected onto the first two PCA axes (Fig. 5E). As a group, these 116 transcripts encoding metabolic enzymes have unexceptional gene expression changes. Purely by chance, we might expect 5% of these 116 (~ 6) to be among the top 5% of differentially regulated transcripts and 25% (~ 30) to be among the top 25%, but in fact none are in the top 5% and only 16 are in in the top 25% (Fig. 5E and F; p = ~0.002). However, this may partly be an artifact of having averaged the expression values for transcripts with the same annotation. To the extent metabolic genes are regulated, they are seemingly more sensitive to salt stress: of the sixteen transcripts among the top 25% of all transcripts, eleven lie in the same quadrant of the PCA projection as the salt stress conditions; only one lies in the same quadrant as the nitrate starvation conditions. In addition to these upregulated transcripts, two of the remaining four have expression profiles that are characterized mostly by repression in salt. One of these is acetyl-CoA carboxylase (ACCase), a key enzyme in the initiation of fatty acid synthesis (Fig. 5E); we return to ACCase in the context of fatty acid

B

Fig. 3. Growth and cell volume. (A) Mean cell diameters were determined from measurement of micrograph images following six days of phototrophic growth. The number of doublings in that time was determined from cell counts. (B) Inverse relationship between the number of cell doublings and cell volume.

3.5. Membrane lipid changes associated with TAG accumulation Given the numerous interconnected pathways for the synthesis and degradation of lipids, changes in TAG levels cannot be understood in isolation [32–35]. We therefore used mass spectrometry to quantify several lipids in addition to TAG. These were phosphatidyl-glycerol (PG), phosphatidyl-choline (PC), phosphatidyl-ethanolamine (PE), sulfoquinovosyl-diacylglycerol (SQDG), and monogalactosyl-diacylglycerol (MGDG). MGDG is the major membrane lipid in thylakoid membranes and, given the abundance of thylakoid membranes, it is also the predominant membrane lipid in the cell [36–38]. Fig. 4A depicts the sum total of these lipids per cell and the fraction of each. It is clear that there are substantial differences in the lipid profiles under the two stress conditions, but there are notable similarities as well. For example, under both of the extreme conditions (0 mM nitrate or 600 mM NaCl), the increase in TAG relative to the control is seemingly compensated for by loss of PC and PG. (By definition, of course, increases in TAG fraction must come at the expense of a decrease in the fraction of one or more of the other lipids, but which lipid it is could differ). This compensation of TAG accumulation through a relative reduction in PC and PG loss does not occur under less stringent conditions. What does seem to compensate for TAG increases under less stringent conditions is a decrease in MGDG. This is true for both stress conditions. Thus, there seem to be gross similarities in the way lipid composition is restructured as stress increases, regardless whether the stress is nitrate starvation or high salt. Finally, we also found that changes in the distribution of fatty acid chain lengths and saturation, while not identical, were broadly similar across the two stresses, again indicating common features in the physiological response (Fig. 4B).

3.6. Transcriptomic analysis In order to find gene expression correlates to the lipid changes under both conditions, we performed de novo RNA-seq analysis under the same ten conditions for which we performed lipid and chlorophyll-a 5

Algal Research xxx (xxxx) xxx–xxx

F.Q.Y. Goh et al.

Fig. 4. Quantitation of lipids by mass spectrometry. (A) The sum of measured lipid quantities normalized to cell number is indicated by the width of the stacked bars. Heights of the stacked bars show the fraction of the total represented by a particular class of lipid. Consequently, the area of each stacked bar represents the total amount of a particular lipid class per cell. Monogalactosyl-diacylglyceride (MGDG) is indicated by green bars, phosphatidylcholine (PC) red, phosphatidyl-glycerol (PG) purple and triacylglycerides (TAG) yellow. Phosphatidyl-ethanolamine (PE) is in gray but is at levels too low to be readily seen. Sulfoquinovosyl-diacylglycerol (SQDG) was measured as well but is not included here as we lacked a standard that would allow its levels to be compared to the other lipids. A breakdown of the lipid species that are summarized in this figure is provided in Supp Fig. 1. (B) The sum of the constituent fatty acid chain lengths and the degree of saturation can be inferred from the mass of each lipid species. The heatmaps show changes in fatty acid chain length (top) and saturation (bottom) relative to the control condition for six different lipid classes. The values are weighted averages across all mass-species of the lipid classes detected by mass spectrometry. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

A

B

else, and some transcripts annotated as something else might be lipases. Even if a sequence is identified correctly as a lipase, its specificity is likely to be uncertain. Thus, by relying on BLAST2GO we may have missed TAG lipases. To search more systematically for lipase homologs, we ran all the open reading frames (ORFs) encoded by the transcripts against eleven lipase-related Pfam Hidden Markov Models (HMMs). Interpretation of those searches was refined using the original transcript sequences as queries in blastx searches (Materials and Methods). The analysis resulted in a set of 33 putative lipase sequences, tentatively classified into five groups: TAG lipases (6 sequences), DAG or MAG lipases (5), phospholipases (13), galactolipase (1) and miscellaneous, generic or ambiguously annotated lipases (8). Expression of most of these transcripts differ from the control by less than 2-fold under either of the conditions. A striking exception is a lipase from the ambiguously annotated class that is down-regulated nearly 5-fold in 0 mM nitrate and nearly 20-fold in 600 mM salt (Supp. Fig. 2). It is tempting to speculate that this is actually a TAG lipase, whose down-regulation contributes to the accumulation of TAG under both stress conditions. However, neither the HMM searches nor blast-based searches provide strong support for TAG specificity, or indeed for any specificity beyond that of a being some kind of lipase. To confirm our speculation that this could be a TAG lipase, experimental characterization of the enzyme will likely be necessary.

metabolism below.

3.7. Glycerolipid metabolism We next narrowed our focus from carbon metabolism to glycerolipid metabolism as defined by KEGG. These enzymes cover activities related to membrane glycerolipid synthesis, TAG synthesis and lipases. Lipases break down lipids for a number of purposes including catabolic energy recovery, remodeling of cellular lipid distribution, and the production or degradation of signal transduction molecules. Among the glycerolipid-related transcripts, four are up-regulated similarly in both stress conditions: 1-acylglycerol 3-P acyltransferase (AGPAT), phosphatidate phosphatase (PP) and diacylglycerol acyltransferases 1 and 2 (DGAT1 and DGAT2) (Fig. 6A). These enzymes catalyze successive steps on the pathway from monoacylglycerolphosphate to triacylglycerides. Although the change in gene expression of even these transcripts is modest, they stand out within the glycerolipid group in terms of both the magnitude of induction and the similarity of their expression profiles in NaCl-stressed and N-starved cells (Fig. 6B). The induction of genes encoding the final steps in TAG synthesis appears to be a shared mechanism for TAG accumulation under both conditions. In principle, TAG accumulation could also be enhanced by downregulation of a TAG lipase. As it happens, none of the putative lipases in our original annotation was down-regulated substantially under either condition. However, annotation of lipases is problematic. Many lipases share a structural fold and active site with other esterases and with serine proteases [41]; thus some annotated lipases might be something

3.8. Fatty acid synthesis and oxidation The remodeling of membrane and neutral lipids that occurs under 6

Algal Research xxx (xxxx) xxx–xxx

F.Q.Y. Goh et al.

A

B −1.0

PC1 (36.0%)

1 0.9

−0.5

0.8 0.7 0.0

0.6 0.5

0.5

0.4 0.3 0.2

1.0 −1.0

−0.5

0.0

0.5

0.1

1.0

0

PC2 (24.3%)

C

D

−7 -3

3 7

PC1 (36.0%)

-20

log2(fold-change) (a)

-10

(b) (c)

0

(d) (e)

10

20 −20

−10

0

10

20

PC2 (24.3%)

E

F −3

-5.0

Fig. 5. Features of whole transcriptomes and selected gene sets across the ten conditions. (A) Principal component analysis (PCA); the first two dimensions account for ~ 60% of the variance in expression. PCA plots have been rotated 90 degrees to facilitate comparison to other visualizations in the paper as these show nitrogen starvation conditions on the left and salt stress on the right. Symbols are as described in Fig. 1. (B) Heatmap and dendrogram of Pearson correlation coefficients calculated on log(RPKM) values. (C) PCA projection of transcripts in the same coordinate space as the conditions in panel A. The two concentric red rings enclose 95% and 75% of the ~ 10,000 transcripts. The radii for the transcripts are scaled to distance from the origin in order to highlight transcripts that contribute most to the principal components. Colored circles correspond to transcripts with related annotations whose gene expression changes are shown in panel D. (D) Average expression differences for five sets of strongly regulated transcripts, labeled a–e. The inferred functions of these transcripts are related to (a) cell wall catabolism, (b) nitrogen transport and assimilation, (c) photosystem, (d) unknown and (e) metal ion transport. The heatmap scale bar is in units of log2(foldchange). (E) PCA projection of transcripts for 116 enzymes involved in lipid and central carbon metabolism (Materials and Methods). Transcript radii are scaled in a same manner to panel C; the red circles are the same as those in panel C and highlight the smaller PCA loading of these transcripts. Selected transcripts have been color-coded for reference to panel F. (F) Expression differences for the transcripts highlighted in panel E. (a): phosphoenolpyruvate carboxykinase (4.1.1.49); (b) glyceraldehyde-3-P dehydrogenase (EC 1.2.1.13); (c) ribose 5-phosphate isomerase (EC 5.3.1.6), (d) malonyl CoA-ACP transacylase (FabD) (EC 2.3.1.39), (e): phosphoglycerate kinase (EC 2.7.2.3). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

3 log2(fold-change)

PC1 (36.0%)

−2.5

(a) (b)

0.0

(c) (d) (e)

2.5

5.0 -5.0

−2.5

0.0

2.5

5.0

PC2 (24.3%) transcripts for this second step in fatty acid synthesis (FabD) is also repressed ~3-fold in salt and to a lesser extent in nitrate starvation. Taking the first two steps in fatty acid synthesis together, the data suggest that novo synthesis of malonyl-ACP, the building block of fatty acid synthesis, is likely to be reduced under both conditions. Unexpectedly, while transcripts for the first steps in fatty synthesis are down-regulated, transcripts for two subsequent steps are up-regulated. FabG, which catalyzes the first of a series of redox and dehydration reactions that complete the elongation cycle following the addition of a new acetyl unit to the growing fatty acyl chain, is fairly strongly induced under nitrogen starvation conditions (~ 4-fold) and more modestly in salt (~ 1.5-fold). FabH (KASIII), which initiates fatty acid synthesis through the condensation of acetyl and malonyl

stress conditions requires at least some oxidation and re-synthesis of constituent fatty acids. In an effort to gain some insight into these processes in the two stress conditions, we identified transcripts related to KEGG-defined pathways for fatty acid biosynthesis, elongation, oxidation and degradation. As was done for the glycerolipids (Fig. 6), for each of the two stresses we averaged the gene expression change over the two most extreme conditions. This summarizes the gene expression profile in two numbers, one for salt stress and one for nitrogen deprivation (Fig. 7). The most striking result is the ~3-fold reduction in transcript levels for acetyl-CoA carboxylase (ACCase), which catalyzes the first step in fatty acid synthesis, the production of malonyl-CoA. The next step is a trans-thioesterification reaction in which the CoA in malonyl-CoA is replaced by acyl carrier protein (ACP). Expression of 7

Algal Research xxx (xxxx) xxx–xxx

F.Q.Y. Goh et al.

A

A

B

B

Fig. 6. Glycerolipid gene expression and TAG synthesis. (A) Pathway to TAG synthesis and gene expression. Lipid schematics show the glycerol backbone with fatty acids (FA) and phosphate (P) when present. MAG-P: monoacylglycerol-phosphate. DAG-P: diacylglycerol-phosphate. DAG: diacylglycerol. TAG: triacylglycerol. Enzymes are AGPAT (1acyl-sn-glycerol-3-phosphate O-acyltransferase; EC2.3.1.51), PP (phosphatidate phosphatase; EC3.1.3.4) and DGAT1 and DGAT2 (diacylglycerol acyltransferase; EC2.3.1.20). (B) Gene expression changes in high NaCl and low nitrate for transcripts whose predicted enzyme activities are in the KEGG glycerolipid pathway. Values indicated are the average log2(fold-change) values for the two most extreme conditions of that type (i.e., 300 mM and 600 mM NaCl and 0 mM and 0.38 mM nitrate). As the scale of the axes differ, a dashed line with a slope of 1 running through the origin has been added to aid interpretation. Enzymes highlighted in orange are those in the direct pathway to triacylglyceride synthesis shown in panel A. Several other enzymes with large or condition-specific expression changes are labeled. With the exception of SQD1 (SQDG synthesis), these are lipases of differing predicted specificity.

Fig. 7. Expression changes for enzymes involved in fatty acid synthesis and degradation. (A) Values on the x-axis are the average of the log2 expression change from control for the two most extreme nitrogen starvation conditions (0 mM and 0.375 mM). Similarly, values on the y-axis are the average of the log2 expression change for the two most extreme NaCl conditions (300 mM and 600 mM). Transcripts shown in orange encode enzymes involved in fatty acid synthesis; those in green encode enzymes unambiguously involved in fatty acid oxidation; those in purple participate in both synthesis and oxidation. (B) Pathways of synthesis and oxidation. The synthesis enzymes (orange) that show the most greatest expression changes are indicated in bold and are labeled in panel A. The remainder are also shown in panel A but cluster at the origin and are unlabeled. Oxidation enzymes (green and purple) are labeled by EC number as in panel A. Oxidation enzymes shown but not labeled in panel A have annotations that are ambiguous or suggestive of more specialized functions. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

simultaneously, a reduction in the ability to synthesize malonyl-ACP units de novo. This suggests that there may be a mechanism for salvaging malonyl-ACP from fatty acyl oxidation pathways, allowing resynthesis of new, different fatty acids without the expense of de novo malonyl-ACP synthesis.

thioesters, is upregulated modestly under both conditions. The repression of ACCase under both conditions and the induction of FabG under nitrate starvation are unambiguous; changes in FabD and FabH expression are less certain, but in the context of ACCase and FabG they are consistent with stress conditions causing an increased capacity for fatty acid synthesis downstream of malonyl-ACP and, 8

Algal Research xxx (xxxx) xxx–xxx

F.Q.Y. Goh et al.

4. Conclusions

the relevant enzymes and corresponding changes in lipid composition. Perhaps the engineered repression of this gene under normal growth conditions might be a strategy for increased TAG production. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.algal.2017.06.014.

What we have sought to do in this work is to discover physiological, metabolic and gene regulatory correlates of TAG synthesis that are consistent across two quite different TAG-inducing conditions. The most intriguing of the correlates we observed was a decrease in chlorophyll concentrations. For the nitrogen-stressed series, this is unsurprising: chlorophyll contains four nitrogens per molecule, so its synthesis can be expected to become restricted under conditions of nitrogen limitation. Furthermore, nitrogen starvation is known (in Chlamydomonas, and presumably other species) to lead to the remodeling of chloroplast membrane lipids and the loss or bleaching of chlorophyll [42–44]. There is no reason to think, based on nitrate starvation alone, that chlorophyll loss and TAG accumulation are mechanistically linked rather than being independent consequences of nitrate limitation. What was surprising was the strikingly similar relationship between TAG and chlorophyll in NaCl-stressed cells. The decrease in chlorophyll occurred in NaCl-stressed samples even though they were in nitrogen replete media and, unlike nitrogen-starved cells, continued to add biomass at all concentrations. The remarkably similar relationship between TAG and chlorophyll under such different conditions increases the chances that the two phenomena are related mechanistically. Whether there is a direct, cause-and-effect relationship is uncertain, but the result does suggest that there is a common pathway to TAG accumulation that operates under both stress conditions, and that this pathway entails the remodeling of chloroplast membrane lipids. Consistent with the notion of a common pathway involving membrane remodeling are the similarities in the way lipid compositions change. These include the loss of MGDG at intermediate stress conditions and the loss of PG and PC under the most extreme stress conditions. Furthermore, at both extremes, fractional MGDG is substantially higher than it is under moderate conditions. Of course, there are substantial differences in the lipid profiles as well, which is not surprising given the radically different effects that the two growth conditions have on cell division and cell size. In a similar vein, differences in gene expression across the two stress gradients attests to some fundamental differences in the physiological response to those stresses. Surprisingly, perhaps, this is true even for genes involved in carbon metabolism: to the extent that such genes are regulated at all, they are typically rather different in their response to the two stresses. The most prominent exceptions are transcripts whose gene products are proximal to TAG synthesis and storage. Most notably, transcripts for the three final steps leading to TAG are all induced under both stress conditions. These are precisely the kind of obvious candidates one might expect to find in common under different TAG-inducing conditions, which is both reassuring and disappointing. In addition, there is a lipase of some kind that is substantially down-regulated under both conditions, a property that distinguishes it clearly from other annotated lipases (Section 3.7, Supp. Fig. 2). The down-regulation of this lipase under conditions in which TAG is accumulating would make sense if its function were to break down TAGs. Whether TAG breakdown is really the function of this transcript cannot be determined from sequence homology. Lipid-related enzymes are a particular challenge for annotation because of ambiguities in substrate specificity. We found, for example, no correlation between the length and saturation patterns of fatty acids (Fig. 4B) and the gene expression patterns of fatty acid elongases and desaturases. Whether this is because of problems with the inference of substrate specificities or whether there simply is not a straightforward relationship between expression values and fatty acid properties across numerous kinds of lipids and organelles is not clear. On the other hand, transcripts encoding enzymes for galactosyl lipids and sulfoquinovosyl lipids did show gene expression patterns that were well correlated with MGDG and SQDG (Supp. Fig. 4). Thus there are at least some cases in which a simple relationship seems to exist between gene expression changes of

Author contributions Conceptualization, NDC; methodology, FCYG, ACG, PT; formal analysis, FCYG, JJ, ZY, NDC; investigation, FQYG, ACG, PT; data curation, JJ, ZY; writing, FQYG, NDC; original draft, FQYG; visualization, NDC; supervision, MW, NDC; funding acquisition, MW, NDC. Acknowledgements Funding to NDC for the early stages of this work was provided by the Agency for Science, Technology and Research (Singapore) through the Genome Institute of Singapore and for later work by Yale-NUS College (R-607-265-053-121) and the National University of Singapore. Funding to MW was supported by grants from the National University of Singapore via the Life Sciences Institute (LSI), the National Research Foundation (NRFI2015-05) and a BMRC-SERC joint grant (BMRC-SERC 112 148 0006) from the Agency for Science, Technology and Research (A*Star). Raw sequence reads are available from the Sequence Read Archive (SRA: PRJNA305197). Assembled transcripts have been submitted to the Transcriptome Shotgun Assembly database (TSA at NCBI; submission SUB1246417, currently being processed as GEEU00000000). Expression data has been deposited at GEO (GSE78006). References [1] I.A. Guschina, J.L. Harwood, Lipids and lipid metabolism in eukaryotic algae, Prog. Lipid Res. 45 (2006) 160–186. [2] Q. Hu, M. Sommerfeld, E. Jarvis, M. Ghirardi, M. Posewitz, M. Seibert, A. Darzins, Microalgal triacylglycerols as feedstocks for biofuel production: perspectives and advances, Plant J. 54 (2008) 621–639. [3] S.S. Merchant, J. Kropat, B. Liu, J. Shaw, J. Warakanont, TAG, you're it! Chlamydomonas as a reference organism for understanding algal triacylglycerol accumulation, Curr. Opin. Biotechnol. 23 (2012) 352–363. [4] C.S. Jones, S.P. Mayfield, Algae biofuels: versatility for the future of bioenergy, Curr. Opin. Biotechnol. 23 (2012) 346–351. [5] R.H. Wijffels, M.J. Barbosa, An outlook on microalgal biofuels, Science 329 (2010) 796–799. [6] M.W. Fields, A. Hise, E.J. Lohman, T. Bell, R.D. Gardner, L. Corredor, K. Moll, B.M. Peyton, G.W. Characklis, R. Gerlach, Sources and resources: importance of nutrients, resource allocation, and ecology in microalgal cultivation for lipid accumulation, Appl. Microbiol. Biotechnol. 98 (2014) 4805–4816. [7] H.C. Greenwell, L.M.L. Laurens, R.J. Shields, R.W. Lovitt, K.J. Flynn, Placing microalgae on the biofuels priority list: a review of the technological challenges, J. R. Soc. Interface 7 (2010) 703–726. [8] D. Klein-Marcuschamer, Y. Chisti, J.R. Benemann, D. Lewis, A matter of detail: assessing the true potential of microalgal biofuels, Biotechnol. Bioeng. 110 (2013) 2317–2322. [9] S.A. Scott, M.P. Davey, J.S. Dennis, I. Horst, C.J. Howe, D.J. Lea-Smith, A.G. Smith, Biodiesel from algae: challenges and prospects, Curr. Opin. Biotechnol. 21 (2010) 277–286. [10] I.C. Woertz, J.R. Benemann, N. Du, S. Unnasch, D. Mendola, B.G. Mitchell, T.J. Lundquist, Life cycle GHG emissions from microalgal biodiesel—a CA-GREET model, Environ. Sci. Technol. 48 (2014) 6060–6068. [11] J. Abe, T. Kubo, Y. Takagi, T. Saito, K. Miura, H. Fukuzawa, Y. Matsuda, The transcriptional program of synchronous gametogenesis in Chlamydomonas reinhardtii, Curr. Genet. 46 (2004) 304–315. [12] L.M. Quarmby, Signal transduction in the sexual life of Chlamydomonas, Plant Mol. Biol. 26 (1994) 1271–1287. [13] L. Gouveia, A.E. Marques, T.L. da Silva, A. Reis, Neochloris oleabundans UTEX #1185: a suitable renewable lipid source for biofuel production, J. Ind. Microbiol. Biotechnol. 36 (2009) 821–826. [14] Y. Li, M. Horsman, B. Wang, N. Wu, C.Q. Lan, Effects of nitrogen sources on cell growth and lipid accumulation of green alga Neochloris oleoabundans, Appl. Microbiol. Biotechnol. 81 (2008) 629–636. [15] T.G. Tornabene, G. Holzer, S. Lien, N. Burris, Lipid composition of the nitrogen starved green alga Neochloris oleoabundans, Enzym. Microb.Technol. 5 (1983) 435–440. [16] H. Rismani-Yazdi, B.Z. Haznedaroglu, C. Hsin, J. Peccia, Transcriptomic analysis of the oleaginous microalga Neochloris oleoabundans reveals metabolic insights into

9

Algal Research xxx (xxxx) xxx–xxx

F.Q.Y. Goh et al.

427–436. [33] E.C. Goncalves, J.V. Johnson, B. Rathinasabapathi, Conversion of membrane lipid acyl groups to triacylglycerol and formation of lipid bodies upon nitrogen starvation in biofuel green algae Chlorella UTEX29, Planta 238 (2013) 895–906. [34] T. Sakaki, K. Saito, A. Kawaguchi, N. Kondo, M. Yamada, Conversion of monogalactosyldiacylglycerols to triacylglycerols in ozone-fumigated spinach leaves, Plant Physiol. 94 (1990) 766–772. [35] K. Sakurai, T. Moriyama, N. Sato, Detailed identification of fatty acid isomers sheds light on the probable precursors of triacylglycerol accumulation in photoautotrophically grown Chlamydomonas reinhardtii, Eukaryot. Cell 13 (2014) 256–266. [36] C. Giroud, A. Gerber, W. Eichenberger, Lipids of Chlamydomonas reinhardtii. Analysis of molecular species and intracellular site(s) of biosynthesis, Plant Cell Physiol. 29 (1988) 587–595. [37] D.R. Janero, R. Barrnett, Cellular and thylakoid-membrane phospholipids of Chlamydomonas reinhardtii 137, J. Lipid Res. 22 (1981) 1126–1130. [38] A. Vieler, C. Wilhelm, R. Goss, R. Süss, J. Schiller, The lipid composition of the unicellular green alga Chlamydomonas reinhardtii and the diatom Cyclotella meneghiniana investigated by MALDI-TOF MS and TLC, Chem. Phys. Lipids 150 (2007) 143–155. [39] Z. Wang, M. Gerstein, M. Snyder, RNA-Seq: a revolutionary tool for transcriptomics, Nat. Rev. Genet. 10 (2009) 57–63. [40] M. Kanehisa, S. Goto, Y. Sato, M. Kawashima, M. Furumichi, M. Tanabe, Data, information, knowledge and principle: back to metabolism in KEGG, Nucleic Acids Res. 42 (2014) D199–D205. [41] M. Cygler, J.D. Schrag, J.L. Sussman, M. Harel, I. Silman, M.K. Gentry, B.P. Doctor, Relationship between sequence conservation and three-dimensional structure in a large family of esterases, lipases, and related proteins, Protein Sci. 2 (1993) 366–382. [42] M.T. Juergens, R.R. Deshpande, B.F. Lucker, J.-J. Park, H. Wang, M. Gargouri, F.O. Holguin, B. Disbrow, T. Schaub, J.N. Skepper, et al., The regulation of photosynthetic structure and function during nitrogen deprivation in Chlamydomonas reinhardtii, Plant Physiol. 16 (2015) 558–573. [43] X. Li, E.R. Moellering, B. Liu, C. Johnny, M. Fedewa, B.B. Sears, M.-H. Kuo, C. Benning, A galactoglycerolipid lipase is required for triacylglycerol accumulation and survival following nitrogen deprivation in Chlamydomonas reinhardtii, Plant Cell 24 (2012) 4670–4686. [44] E. Preininger, A. Kósa, Z.S. Lőrincz, P. Nyitrai, J. Simon, Structural and functional changes in the photosynthetic apparatus of Chlamydomonas reinhardtii, Photosynthetica 53 (2015) 369–377.

triacylglyceride accumulation, Biotechnol. Biofuels 5 (2012) 74–89. [17] A.M. Santos, P.P. Lamers, M. Janssen, R.H. Wijffels, Biomass and lipid productivity of Neochloris oleoabundans under alkaline–saline conditions, Algal Res. 2 (2013) 204–211. [18] W. Chen, C. Zhang, L. Song, M. Sommerfeld, Q. Hu, A high throughput Nile red method for quantitative measurement of neutral lipids in microalgae, J. Microbiol. Methods 77 (2009) 41–47. [19] E.G. Bligh, W.J. Dyer, A rapid method of total lipid extraction and purification, Can. J. Biochem. Physiol. 37 (1959) 911–917. [20] M.J. Griffiths, R.P. van Hille, S. Harrison, Lipid productivity, settling potential and fatty acid profile of 11 microalgal species grown under nitrogen replete and limited conditions, J. Appl. Phycol. 24 (2012) 989–1001. [21] U. Pick, T. Rachutin-Zalogin, Kinetic anomalies in the interactions of Nile red with microalgae, J. Microbiol. Methods 88 (2011) 189–196. [22] M.G. Grabherr, et al., Full-length transcriptome assembly from RNA-Seq data without a reference genome, Nat. Biotechnol. 29 (2011) 644–652. [23] B. Langmead, C. Trapnell, M. Pop, S.L. Salzberg, Ultrafast and memory-efficient alignment of short DNA sequences to the human genome, Genome Biol. 10 (2009) R25. [24] B. Li, C.N. Dewey, RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome, BMC Bioinf. 12 (2011) 323–338. [25] M.D. Robinson, D.J. McCarthy, G.K. Smyth, edgeR: a bioconductor package for differential expression analysis of digital gene expression data, Bioinformatics 26 (2010) 139–140. [26] A. Conesa, S. Götz, J.M. García-Gómez, J. Terol, M. Talón, M. Robles, Blast2GO: a universal tool for annotation, visualization and analysis in functional genomics research, Bioinformatics 21 (2005) 3674–3676. [27] G. Blanc, et al., The Chlorella variabilis NC64A genome reveals adaptation to photosymbiosis, coevolution with viruses, and cryptic sex, Plant Cell 22 (2010) 2943–2955. [28] S. Merchant, S. Prochnik, O. Vallon, E. Harris, The Chlamydomonas genome reveals the evolution of key animal and plant functions, Science 25 (2007) 245–251. [29] S.R. Eddy, Profile hidden Markov models, Bioinformatics 14 (1998) 755–763. [30] R.D. Finn, et al., The Pfam protein families database, Nucleic Acids Res. 38 (2010) D211–D222. [31] B.J. Haas, et al., De novo transcript sequence reconstruction from RNA-seq using the Trinity platform for reference generation and analysis, Nat. Protoc. 8 (2013) 1494–1512. [32] B. Antonny, S. Vanni, H. Shindou, T. Ferreira, From zero to six double bonds: phospholipid unsaturation and organelle function, Trends Cell Biol. 25 (2015)

10