On the proteome level, the temporal adaption to N- can be separated into three phases and eight functional clusters
To characterize the physiology upon switch from normal culture condition to N- in N. oceanica IMET1, proteomes were compared among culture samples from 3 h, 6 h, 12 h, 24 h, 48 h and 10 days, which provided proteome profiles of N. oceanica IMET1 cells under differing nitrogen availability. By employing label-free relative quantification (LFQ) for the proteome data, a total of 4114 protein sequences were identified and quantified. This represents about 41% of all theoretically predicted 9915 proteins encoded in the genome of N. oceanica IMET1 [6, 9, 33, 34] After filtering out proteins identified under one treatment only in less than half of the samples, 1795 proteins remained. For these, hierarchal clustering was performed on the log2 fold change of label free-quantified, normalized protein intensities (i.e. log2LFQ(N-/N+), as LogetP) [35, 36] (Figure 1A). Functional category was based on homology from gene annotation data obtained by Blast P searches against Non-redundant protein sequences (nr) database from NCBI (Figure 1B and C) [29, 37]. On the time dimension, proteins were distributed into three phases over six timepoints, i.e. from 3 h to 12 h, from 12 h to 48 h, and until 10 days, with 12 h as distinct phase switch point. PCA analysis of protein LFQ-intensities of all samples also proved this separation from another angle (Supplementary Figure 1). In general, proteome profiles in the first 48 h were more similar, yet 12 h was an outlier and considered as physiology switching point as discussed later. Another significant distinction from other proteomes is obvious for 10th day (Supplementary Figure 1). This observation is partly mirrored by the clustering based on spearman’s correlation (Figure 1A and Figure 1B): On the level of changes in protein abundance, these 1795 proteins (N- vs N+) formed eight clusters with timepoints 12 h and 10th day displaying the strongest changes in individual clusters, e.g. clusters 7 and 8 exhibit strong up/down-regulation of proteins at 12 h and in most clusters 10th day exhibits the strongest up/down-regulation. To probe whether a given cluster was linked to specific protein functions, proteins were divided into twelve functional categories, and the five most frequent categories in each cluster (including genes of unknown function) designated as its primary functional proteins (Figure 1B and C). Cluster K1 was functionally enriched with protein synthesis, carbon metabolism, gene expression, and other metabolism proteins, which were downregulated at 3 h and 6 h and then upregulated until 10th day (Supplementary Table 1). Although the primary functional genes of K4, K7 and K8 were all involved in carbon metabolism, gene expression, other metabolism, and protein synthesis, their tendencies were different (Figure 1B): The proteins in K5 and K6, showing a downregulation trend, were related to protein synthesis, photosynthesis, carbon and other metabolism, but their tendencies were slightly different. K5 started with downregulation at 6h whereas in K6 proteins were upregulated first then downregulated at 24 h. In K2, a zigzag trend for protein ratios was displayed, energy-related genes (ATP related genes) were enriched, as were proteins related to carbon metabolism, transport, and protein synthesis. In K3 proteins with functions in protein synthesis, transport, carbon metabolism, and energy were upregulated about one log2 fold change.
On the level of functional categories, magnitude of changes was smaller for proteome compared to transcriptome
For the various functional categories, a corresponding previous time series of transcriptomes fold changes (retrieved from [6] including five time points 3 h, 6 h, 12 h, 24 h and 48 h) was contrasted with the proteome fold changes over the six timepoints of 3 h, 6 h, 12 h 24 h 48 h and 10th day (Figure 2). Underlying ratios of proteins LogetP and the ratios of transcripts (log2FPKM(N-/N+), as LogetT) were listed in one table to facilitate further analysis (Supplementary Table 2). On the level of functional categories, temporal changes in proteome and transcriptome were mostly similar, yet some functional groups showed differences: At 12 h and 24 h transcripts of photosynthesis genes were more downregulated compared to their proteome counterpart. The strongest downregulation of transcripts of photosynthesis genes was at 12 h, whereas it was at 48 h for proteins. The chromatin group showed upregulation on transcriptome at 24 h and 48 h (factor around 0.5), whereas on the proteome, upregulation manifests at 10th day (Figure 2 and Supplementary Table 2). In the nitrogen metabolism group, the downregulation for transcriptome is stronger than proteome. The proteins related to cell structure were downregulated more than their transcripts for 3 h and 12 h, whereas upregulated at 48 h (Figure 2). In general, in most functionally enriched groups, the magnitude of changes for transcriptome was larger than for proteome, pointing towards a generally more stable proteome in N. oceanica facing N-.
Given that mRNA synthesis precedes protein synthesis, it can be comprehended that the proteome response to changes in transcription may be delayed. To account for such delay, contrasting the tendencies of transcriptome and proteome rather than making a point to point comparison would be more appropriate [38]. For this purpose, Paparrizos and Gravano introduced a statistical method, so called “k-shape” [39]. Intriguingly, despite high point to point differences between protein and mRNA for photosynthesis genes in Figure 2, according to k-shape only 18% of proteins exhibited tendencies highly unsimilar to their transcriptome data. Thus, for photosynthesis, majority of qualitative differences between protein and mRNA levels can be explained by a delayed proteome response to transcriptional changes (Figure 3 and Figure 4A).
Transcriptomics overestimates downregulation of light harvest complex (LHC) proteins under N-
In the N. oceanica IMET1 genome, 20 genes are predicted as light-harvesting complex or VCP antenna proteins [40] of which nine proteins (IMET_6482, 4633, 5763, 1484, 6457, 228, 9096, 4345 and 9281) were downregulated on proteome and all except IMET_9281 downregulated on the transcriptome, respectively (Figure 3). Moreover, based on the k-shape distance, five LHC proteins had change tendencies highly similar to their transcriptome (highly similar is defined as k-distance smaller than lower quartile of all the k-distances), which were also continuously downregulated. Four protein tendencies (IMET_9116, 5303, 1484 and 8913) were highly dissimilar to their transcripts (bigger than upper quartile of all the k-shape distances); the protein abundance of PSII Lhc4 (IMET_9116) was increasing under N- at 3, 6 and 24 h, and the PS II Lhc7 (IMET_9116) was increasing under N- at 48 h. The unclassified light harvest complex protein Lhcv8 was relatively stable over time. Generally, most LHC transcripts were considerably downregulated under N- from 3 h to 48 h except IMET_9281. However, corresponding proteins were relatively stable at beginning with onset of downregulation usually after 24 hours (Figure 3).
Under N-, carbon fixation and transfer, as well as main carbon source for TAG synthesis change over time
Nannochloropsis under N- will accumulate lipids, with most being TAGs [6, 14]. The relevant processes of carbon fixation, transfer, and conversion to lipids were depicted both for transcriptome and proteome (Figure 4A and B). Carbonic anhydrases (CAs) play a key role in biophysical carbon concentrating mechanisms (CCM) to catalyze the inter-conversion of CO2 and HCO3- [34]. Five putative CAs (CA1: IMET_2109 CA2: IMET_5775; CA3: IMET_4525; CA4: IMET_1930, CA5: IMET_1034) are present in N. oceanica IMET1 [6, 23] and predicted to reside in the chloroplast according to results from ChloroP, TargetP, Signal P and former research [41-44] (Supplementary Table 3). Under N-, on proteome level, CA1 and CA3 strongly increased in abundance at 10th day (log2LFQ(N-/N+) = 0.9 and 1.4), while CA2 increased during the first 48 h then decreased to -1 (LogetP at 10th day (Figure 4A). In contrast, transcripts of the three CAs (CA1, CA2, CA3) were all slightly downregulated. It is noteworthy that, despite the transcript of CA4 was increasing over time, the corresponding protein was not detected. Regarding Bicarbonate transporters (BCTs) responsible for chloroplast HCO3- transport, protein abundances increased at 48 h to 10th day (BCT1) and 6h (BCT2), yet the transcript levels of these two bicarbonate transporters were quite stable (Figure 4A).
The Calvin cycle plays a vital role in transforming the light energy into carbon compounds by fixing CO2 as glyceraldehyde 3-phosphate (GAP). On proteome level, the abundances of triosephosphate isomerase (TPI) and transketolase (TL) (IMET_71 and 7516) increased at 24 h and 10th day, whereas most of the other enzymes related to the Calvin cycle decreased under N- condition: The protein abundance of seven enzymes (IMET_8093, 9697, 1575, 3984, 4415, 5352 and 4727) started to decrease at 48 h with tendencies highly similar to their transcripts. For other enzymes of the Calvin cycle, like IMET_2880, 6907, 2540 and 4720, only their transcripts were identified. Generally, except TPI (IMET_71), which increased at 10th day, all the proteins related to the Calvin cycle were stable before 48 h of N- then started their down regulation. Unlike proteins, the transcripts of genes were reduced earlier and more (Figure 4A) at 3 h, down to -3 and -2 LogetT after 24 h. Generally, unlike the transcripts indicating onset of reduced activity of Calvin cycle in Phase I (3 h-12 h), proteomics suggests that the decrease of the efficiency of Calvin cycle happened after 48 h of N- (Figure 5).
The synthesis of TAG under N- is not yet sufficiently understood at the proteome level. In chloroplast exists the fatty acid synthesis pathway that catalyzes the synthesis of fatty acids from GAP obtained in Calvin cycle, [4, 45-47] (Figure 4A). Alternatively, triosephosphates from Calvin cycle are used for biosynthesis of carbohydrates (Supplementary Figure 3). Therefore, both lipid- and carbohydrate-synthesis pathways are closely associated;
For carbohydrates, GAP is converted into glucose, and the reaction of glucose-1-phosphate to UDP-glucose is committed step for synthesis of cellulose and the storage carbohydrate laminarin. For the first 3 h, transcripts exhibited a slight increase for enzymes of glucose-1-phosphate synthesis, cellulose and laminarin. This was mirrored on the proteome level, though not significantly for glucose-phosphate mutase (IMET_2792, IMET_4273), UDP-glucose phosphorylase (IME_422, IMET_2792), cellulose synthase (IMET_4748, IMET_6523), except for laminarin branching enzyme (IMET_2418). Endoglycosidases were not detected on the proteome level – IMET_5082 mRNA was consistently upregulated, whereas IMET_7845 mRNA was downregulated at 24 h and 48 h.
For lipids, the GAP is converted to pyruvate, which is utilized for acetyl-coA synthesis by chloroplast pyruvate dehydrogenase (PDH, IMET_8724), which increased after 24 h both in transcript and protein amounts. Enzymes that catalyze conversion of malonyl-acp to 3-ketoacyl-acp were strongly upregulated (LogetP= 3) at 10th day under N-, with changes highly unsimilar to their transcripts. The protein abundance of ketoacyl synthase (KAS6, IMET_5417) under N- also increased at 10th day, but its transcription barely changed. In Nannochloropsis, the increasing abundance of fatty acid synthesis proteins on phase II (12 h-48 h) and phase III (10th day) indicated that after N- the photosynthesis switched from carbohydrate to fatty acid production to supply TAG after 48 h. On the other hand, the upregulation of lipase (IMET_5348) indicated the TAG synthesis until 24 h under N- may consume chloroplast lipids.
The upregulation of enzymes on Kennedy pathway suggests that most of the TAGs were de novo synthesized after 48 h of N- depletion
The Kennedy pathway involves the synthesis of TAG from acyl-CoA. At the proteome level, only a minor portion of related enzymes were identified. Most of these proteins showed no strong changes (Supplementary Table 2 and Figure 4B). The glyceraldehyde 3-phosphate dehydrogenase (GAPDH, IMET_ 3470) that provides glycerol-3-phosphate as backbone for fatty acid synthesis presented an increase on the protein level at 6 hours. Glycerol-3-phosphate acyltransferase (GPAT, IMET_9161) was regulated highly differently at proteome level vs transcript level; its protein abundance showed a slight upregulation at 6h, and strong increase up to 2 LogetP at 10th day, while its transcript remained unchanged. No lysophosphatidic acid acetyl transferase (LPAAT) was found at the proteome level, even though in the transcriptome seven isoforms were identified (Figure 4B). Long chain fatty acyl-CoA synthetase (LC-FACS) were reported to participate in lipid synthesis and transportation [48, 49]. Protein abundance changes of IMET_3596, 8191 and 5675 were highly unsimilar to their transcript changes. IMET_8191 was upregulated at 6h and 10th day and IMET_5675 was slightly downregulated from 3 h, then upregulated at 10th day. For DGAT, only two proteins were identified. On the protein level, unlike its transcript, DGAT (IMET_1645) was downregulated at 3 h and 12 h and at the other time points not changed. The other detected DGAT protein was IMET_9521, from 3 h to 48 h, its protein abundance was downregulated, however, it was strongly upregulated at 10th day under N-. The lipid body surface protein (LDSP, IMET_5506) displayed similar trend for proteome and transcriptome; both hardly changed over time under N-.
Mitochondrion contributed carbon precursors to the fatty acid synthesis via the protein degradation
The mitochondrion may possess a potential biochemical pathway for fixing inorganic carbon (HCO3-) postulated previously by us [23, 29]. The carbon was fixed by phosphoenolpyruvate carboxylase (PEPC, IMET_4839) converting phosphoenolpyruvate (PEP) to oxaloacetate (OAA) (Figure 4B). From 24 hours an increase in protein abundance began for PEPC (IMET_4839). It reached a LogetP of 1 3 at 48 h. Meanwhile, its transcript has been already upregulated since 3 h. In the TCA cycle, OAA is transformed to malate by malate dehydrogenase (MDH IMET_7304, 8706) to generate energy or transferred to chloroplast via C4 like pathway [6]. There was a strong upregulation on the protein level after 48 h for most OAA-related enzymes (MDH1: IMET_7304; pyruvate phosphate dikinase (PPDK2: IMET_5132; PEPC: IMET_4839, and phosphoenolpyruvate carboxykinase (PEPCK, IMET_6484), while their transcripts increased after 24 h of N-. For the other enzymes related to the TCA cycle, all proteins except citrate synthase (CS1, IMET_1911) were strongly upregulated at 10th day, succinate dehydrogenase (SDH, IMET_1564) was upregulated up to factor 3.6 LogetP, transcripts of genes IMET_706, 9373, 7535, 1564 and 1900 have been strongly upregulated from 24 hours of N- (Figure 4B).
In cytoplasm, the enzyme catalyzing conversion of pyruvate (PYR) to PEP (IMET_3227) strongly increased from 24 h to 10th day and has been at transcriptome level upregulated from 3 h of N-. On the contrary, the enzyme catalyzing conversion of PEP to PYR, pyruvate kinase (PK, IMET_8583), was strongly downregulated at 24 h and 10th day on transcription level with no big changes in protein abundances. The other genes involved in the so called C4-like pathway, such as IMET_3280, were also upregulated, though here protein abundances increased later than that of transcripts. In summary, the upregulation of TCA cycle enzymes points to increasing MAL and PYR concentrations from 48 h, which provided the carbon precursors for fatty acid synthesis (Figure 5).
Nitrogen metabolism differs in L-glutamate synthesis between proteome and transcriptome
Several proteins involved in nitrogen transport and metabolism were not detected, but glutamate dehydrogenase (IMET_1767) and glutamate synthase (IMET_277). Glutamate dehydrogenase catalyzes the synthesis of L-glutamate from ammonia and alpha-ketoglutarate. Whereas glutamate synthase converts L-glutamine and alpha-ketoglutarate in two molecules L-glutamate and increased on the protein level. Both enzymes IMET_1767 and IMET_277 were identified with dissimilar changes on the protein and transcript levels (Figure 6) from 12 h to 48 h. Ammonium importer (IMET_7297) was strongly upregulated at 48 h on the protein level. Nitrate reductase (IMET_1590) hardly changed on both protein and transcript level. Proteasome is the first step of protein degradation, in Figure 6, the proteasome subunits (IMET_1975, 3213 and 8302) were upregulated at 12 h. Later, at 10th day, more proteasome subunits like IMET_1975, 2420, 4664, 5113, 6614 and 8302 were upregulated. On the contrary, the transcriptome of proteasome was stable or slightly downregulated. Proteasome regulators are considered as activities regulator of proteasome. On the protein level, most proteasome regulators (IMET_1280, 1764, 3358, 3821, 4773, 5451, 6079, 6531 and 8002) were stable until 48 h and downregulated at 10th day, except the IMET_8951, which was upregulated at 3 h and 10th day.