The efficacy of N starvation in inducing a rapid and massive transition from vegetative cells to resting spores in C. socialis, first shown by Pelusi et al. ([19], [21]), is here confirmed. After three days of exponential growth, C. socialis cells grown in N depleted medium entered in stationary phase and vegetative cells (Figure 1a) progressively turned into spores (Figure 1b), reaching ~30% and ~75% of the total cell number on T3 and T4, respectively (Figure 1c). In control conditions, spores were present with a much lower percentage (~15%) only at T4 (Figure 1d). The experiment was conducted with exactly the same set-up illustrated in Pelusi et al. ([19]), where concentrations of inorganic nutrients in the medium and of organic N and C in cells were monitored in both treatment and control conditions. The cell and spore concentrations over the four days of the experiment here illustrated matched the overall dynamics observed in Pelusi et al ([19]), where spore formation in the treatment coincided with a marked increase of intracellular C, while the intracellular N pool remained almost constant, notwithstanding the fact that inorganic N decreased in the medium, suggesting a qualitative shift in intracellular nitrogenous compounds in the cells.
Transcriptome assembly
RNA-seq experiments were performed on C. socialis samples collected in N deplete medium before the formation of spores (T2), when spore formation started (T3), and when spores reached > 75% of the whole population (T4); control samples were collected from cells in mid-exponential growth phase on day 2 (C2). Since the C. socialis genome is not available yet, raw reads have been assembled with a de novo approach. The assembled transcriptome accounted for a total of 32,224 transcripts (Table S1, S2), with a mean GC content of 43.98%, an average and median contig length of 776 and 536 bp, respectively, and a N50 of 1140 bp (Table S1). In silico encoded protein sequences were 19,153. Among these, 6693 (34.9%) were full-length proteins, 5401 (28.2%) were 5’ partial proteins lacking a stop codon, 2328 (12.2%) were 3’ partial proteins lacking a start methionine, and 4731 (24.7%) lacked both a starting methionine and a stop codon (Table S1). The function of 8,070 protein sequences (42.1% of the total predicted proteins) was successfully annotated (Table S2). Among these, 7,974 proteins were associated with at least to one gene ontology (GO) term.
The comparisons between each time point (T2, T3 and T4) and the control (C2) were used to detect the molecular pathways involved in spore formation. More than 65% of the total transcripts (21,873) resulted to be differentially expressed genes (DEGs), with a fold change |FC| ≥1.5, and the majority of them belonged to T3 and T4 (Table 1, Table S2). The hierarchical clustering of the DEGs dataset highlighted a suite of genes specifically characterizing T3 and T4 compared to the control condition (Figure 2).
GO enrichment
A summary of the 20 most significantly enriched biological processes detected by the GO enrichment analysis at the different time points of each comparison is presented in Figure 3. A total of 3,057 GO terms were significantly enriched in T2, T3 and T4 versus control (Table S3). The enriched GO terms for both up- and downregulated DEGs in T2 versus control were very few and redundant suggesting that the major metabolic rearrangements started at T3. In fact, the presence of spores at T3 and T4 resulted in an increase in the number of the enriched GO terms, both for up- and downregulated DEGs (Table 1, Table S3), highlighting a higher level of transcriptional regulation at these sampling points. In addition to terms related to metabolic processes (cellular aromatic compounds and cellular nitrogen compounds) and to nuclear activity (nucleobase-containing compound metabolic process, and nucleic acids metabolic processes), an enrichment of terms relative to upregulated DEGs linked to cell cycle was present both at T3 and T4 comparisons and to cell division at T4 comparison versus control. These latter two observations are paralleled by the high expression of cyclin D3-2 (Table 2), whose homolog in Phaeodactylum tricornutum was suggested to play a role in G2-to-M transition ([25]). The enrichment of terms related to cell division seems at first counterintuitive because N limitation should negatively affect cell duplication. However, this finding is supported by observations in time-lapse microscopy demonstrating that two consecutive acytokinetic nuclear divisions immediately precede the transformation of vegetative cells into spores ([21]). After the mitotic division, one nucleus degenerates and one new thick theca of the spore is synthesized; the process is repeated when the second theca of the spore is produced.
Photosynthesis was the most enriched term among the downregulated genes in T3 and T4 compared to the control, followed by protein-chromophore linkage and generation of precursor metabolites and energy generation (Figure 3, Table S3). The decline of photosynthesis during N starvation is a common response in photosynthetic organisms due to the tight connection between C and N metabolisms ([26]). N starvation induces an excess of light absorption relative to C fixation, resulting in photo-oxidative damage to a cell that could lead to cell death; thus, cells must control photosynthesis to mitigate the damage. Liefer et al. ([27]) demonstrated a significant decline of light-harvesting complexes and PSII reaction centers during N starvation in Thalassiosira pseudonana and T. weissflogii. We can thus conclude that the downregulation of genes related to photosynthesis is a general response in diatoms, well-fitting with the lower metabolism that characterizes resting stages ([2]). The enrichment of terms of downregulated genes such as the translation elongation and organonitrogen compound biosynthetic process, especially at T3 versus control, recalls the downregulation of the same processes in P. tricornutum during N starvation ([22]). Surprisingly, also the reactive oxygen metabolic process term was enriched at all the downregulated sampling points, being among the first 20 GO terms at T2 together with other stress-related terms (Figure 3). However, a number of transcripts specifically related to oxidative stress were found upregulated at T3 and T4 (see below).
Primary metabolism during spore formation
In the following, we provide evidence derived from gene expression analysis for several metabolic pathways detected in C. socialis during spore formation. The 10 annotated genes with the highest up- and downregulation values in all pairwise comparisons are reported in Table 2; upregulated genes had a FC between 13 and 12 both at T3 and T4 and downregulated ones were not differentially expressed at T2, indicating their involvement during spore formation. Some of the genes were present with a high number of isoforms, such as the high-affinity urea active transporters DUR3 and the serine/threonine-protein kinases TOR (see Table S4, S7), suggesting a still high level of redundancy among transcripts even though the filtering applied during the bioinformatic analysis.
Two and four out of the six DEGs tested were found significantly differentiated in APC12 and MCA6, respectively (Figure S1). The most differentially expressed high-affinity nitrate transporter NRT 2.6 and the silicon efflux transporter LSI3 confirmed their expression in both strains.
Nitrogen assimilation
Surprisingly, the high-affinity nitrate transporters (NRTs) in C. socials had a distinctive expression profile in N starved conditions as compared to the one detected in other diatoms ([28], [29]). The assimilation of nitrate was severely impacted by N stress: three out of the four NRTs were in fact highly downregulated, both at T3 and T4, with one of them (high-affinity nitrate transporter 2.6, NRT2.6) being the most downregulated transcript of the whole dataset (Figure 4, Table 2 B, Figure S1). NRTs can be constitutive or inducible in microalgae, as in plants, especially in response to low N concentration ([28]). Examples among diatoms are T. pseudonana, Pseudo-nitzschia multiseries and P. tricornutum, which showed upregulation of these genes in N starvation, with the only exception of one transcript of P. multiseries (Pm 261779) ([23]). In the haptophyte Tisochrysis lutea, the expression of one of the four NTRs decreased in N depletion, showing an expression profile matching the decline of the intracellular N:C ratio ([30]). This ratio markedly decreased also in C. socialis, due to C accumulation when cells turned into spores ([19]), suggesting that a similar mechanism could explain the observed expression profile in our dataset. In plants, light and C modulate NRTs expression as well and nitrate uptake is not only determined by nitrate availability and demand but also by C produced by photosynthesis ([31]).
We could not detect any low-affinity nitrate transporters (NPFs) in C. socialis, thus confirming the results recently obtained for various Chaetoceros species in which NPFs are missing ([32]); this could support the suggestion that the species of this genus can preferentially use ammonia, possibly from bacteria, as a source of N in case of fluctuating nitrate concentrations ([33]). While the expression of ammonium transporters was somehow variable in our dataset, the high-affinity urea active transporters DUR3 were upregulated at least in one sampling point, with two of them already differentially expressed at T2 (Figure 4, Table S4).
Several transcripts coding for glutamine synthetase (GS) and glutamate synthase (GOGAT), key enzymes in N assimilation, have been detected (Figure 4, Table S4). Among the GS transcripts, one transcript coding for GSIII showed a markedly positive FC at T3 and T4 versus control, as reported for T. pseudonana ([34]) and for the cyanobacterium Synechococcus ([35]) during the early phase of N starvation. Hockin et al. ([34]) suggested that in T. pseudonana this gene participates in the assimilation of ammonium obtained by intracellular catabolic processes, in combination with a NAD(P)H-GOGAT. In our dataset, only a chloroplastic ferredoxin-GOGAT enzyme displayed a similar positive trend at all sampling points, while the other enzymes using NADH as cofactor showed only slight positive regulation at T3 (Figure 4). It can thus be hypothesized that C. socialis, as other diatoms, recovers most of N by recycling the internal pools, as further suggested by the detection of autophagy-related proteins and the hydrolysis of urea by urease, over-expressed at T3 and T4 (Figure 4, Table S4). On the other hand, despite the upregulation of urease, the urea cycle transcripts did not appear differentially expressed (Figure 4). Another important source of N could be the ammonia obtained from the breakdown of phosphoethanolamines through the ethanolamine-phosphate phospho-lyase, an enzyme showing a FC greater than 12 when spores were detected at T3 and T4 (Table S4).
Tricarboxylic acid cycle
The tricarboxylic acid (TCA) cycle is responsible for generating energy through the oxidation of pyruvate to form CO2, ATP, NADH, and carbon skeletons used for biosynthetic processes. This term was enriched in the GO enrichment analysis, although it was not among the first 20 terms, and presented several enzymes upregulated during the time course of the experiment (Table S4). The upregulation of two isocitrate lyases with FCs of 4 and 8, respectively, only when spores were present (i.e., T3 and T4) could imply that the glyoxylate cycle, a variant of TCA cycle that uses lipids as a source of energy, is preferred by diatom spores (Table S4). The upregulation of transcripts such as the mitochondrial short-chain specific acyl-CoA dehydrogenase and the peroxisomal multifunctional enzyme type 2 (MFE-2) that increased progressively in T3 and T4, suggest the involvement of the beta-oxidation of fatty acids in feeding the glyoxylate cycle with acetate molecules (Table S4). Indeed, a calcium-independent phospholipase A2-gamma responsible for membrane lipid degradation has been found among the most up regulated genes (Table 2). These metabolic changes are consistent with those involved in the formation of pellicle cysts in the dinoflagellate Scripsiella trochoidea ([36]).
Cell wall
One of the most relevant changes during the formation of spores is the deposition of two thick heteromorphic siliceous thecae that confer mechanical protection to the spores. Metabolites such as proline and long-chain polyamines (LCA) are involved in the synthesis of the organic component of the siliceous cell wall and their synthesis is connected to the urea cycle and the TCA cycle ([37]). Transcripts related to the proline biosynthesis were not particularly overexpressed at any sampling point, but evidence of spermidine production, another polyamine, came from the upregulation of several spermidine synthases, especially when spores were present (Table S4). Furthermore, polyamines transporters and polyamine oxidases, involved in the regulation of their intracellular concentration, showed high FC at T3 and T4. The deposition of new silica thecae of the spores was supported also by the simultaneous increase of two silicon efflux transporters (LSI), with FC between 4 and 5, and of several other transcripts presenting the InterPRO domain ‘Silicon transporter’ (IPR004693) with extremely high FC (up to 11) (Table S4, Figure S1).
Carbon skeletons
Part of carbon skeletons for sustaining the cells were obtained from the degradation of chrysolaminaran, the principal storage compound of diatoms. The exo-1,3,-beta-glucanases, presumed enzymes responsible for its degradation ([38]), were found with ten different transcripts, eight of which were upregulated particularly when spores were present with FC ranging from 1.9 to 6 (Table S4). These skeletons most probably feed the glycolysis and the pentose phosphate pathways, both with upregulated enzymes (Table S4).
The pentose phosphate pathway produces NADH and pentose sugars in the oxidative and reductive phases, respectively. Among the sugars produced, the ribose 5-phosphate is the precursor of nucleotides and thus essential for DNA replication and transcription. In analogy to what was observed in N starved P. tricornutum ([22]), enzymes involved in the oxidative pathway - two glucose-6-phosphate 1-dehydrogenases and two 6-phosphogluconate dehydrogenase decarboxylating 1, one chloroplastic and one cytosolic - were upregulated at T3 and T4 (Table S4). Transcripts of the non-oxidative part were instead generally downregulated, except for one chloroplastic transketolase, which is responsible for the production of D-xylulose 5-phosphate and D-ribose 5-phosphate; this latter molecule is the precursor of nucleotide biosynthesis, essential for the formation of spores.
Chemical signaling
Pathways related to oxylipin production
There are several examples of chemically mediated communication in unicellular organisms ([39]), with some of them having a critical role in dormancy ([40]). In diatoms, the most studied infochemicals related to intra- and inter-specific communication are compounds belonging the oxylipin family, which include polyunsaturated aldehydes (PUAs) and linear oxygenated fatty acids (LOFAs), both derived from the oxygenation of fatty acids ([41]). The production of LOFAs has been reported for C. socialis although the enzymatic pathway involved is still unknown ([42]). The first step of their synthesis derives from the oxidation of membrane lipids, through several enzymes among which a crucial role is played by lipoxygenases. A single-copy lipoxygenase with an extremely high FC (>10) has only been detected when spores were present (i.e. T3 and T4), supporting the hypothesis of oxylipin-mediated signaling as the trigger of spore formation (Table S5). The formation of spores under stress conditions – i.e., N depletion, high cell density or viral attack ([19],[20]) – in which oxylipins can be produced following cell lysis or breakage, further corroborates the involvement of these infochemicals in determining this life cycle transition in C. socialis. In addition to the overexpression of the lipoxygenase, several aldehyde dehydrogenases (ALDHs), plausibly linked to the detoxification from oxylipins, were highly overexpressed especially at T3 and T4 (Table S5). These enzymes are very conserved all over the phylogenetic tree of life and have a variety of functions, spanning from detoxification against oxidative stress ([43]), to being markers for highly proliferating stem cells and cancer cell phenotypes ([44] and reference therein). We hypothesize that the simultaneous production of LOFAs and ALDHs in C. socialis may be attributed to the presence of two different cell types in the culture: LOFAs could be produced by N starved cells in poor physiological conditions as a response to stress, while the ALDHs by healthy cells that could ‘react’ to stress starting a series of intracellular cascade signals that lead to spore formation.
Programmed cell death and oxidative stress
ALDHs are also associated with programmed cell death (PCD) in addition to stress oxidative response, as shown for N starved diatoms ([45]). In other microalgae, these two processes are involved in cell cycle arrest and/or differentiation ([46], [47], [48]). The exposure to a sub-lethal concentration of reactive oxygen species modifies a broad range of cellular processes, activating or inhibiting transcription factors, membrane channels, etc., and thus acting as secondary messengers for various physiological processes aimed to increase cell survival ([49]). The high expression of antioxidant enzymes, counteracts the undergoing oxidative stress during spore formation: for example, the peroxiredoxin-6 and the quinone oxidoreductase PIG3 were detected among the most overexpressed genes in the dataset (Table 2). Interestingly, the quinone oxidoreductase PIG3 has a double role in inducing apoptosis or prolonged cell cycle arrest in cancer mammalian cells, depending on the physiological state of cells ([50]). Thus, it is possible that this enzyme is produced by both resting and dying cells of C. socialis in our experimental set-up.
Diatoms have a surveillance system based on Ca2+ and nitric oxide (NO) production, which helps to monitor the stress levels within the population ([51], [52]). Although only one out of the three nitric oxide synthases present in C. socialis was weakly up regulated in the early stage of spore formation (T3), several Ca2+-dependent protein kinases have been found with FCs ranging from 2 to 12, with the highest FCs detected when spores were present (Table S5). The expression among the antioxidants of two chloroplastic enzymes - a glutathione synthetase and a thioredoxin-like 2-1 with FCs ranging from 5 to 7 especially at T3 and T4, hinted an important role for the chloroplast redoxome in defining which cells were dying or forming spores. However, the contemporary presence of the two types of cells in the culture is most probably at the base of the contrasting expression profile of many antioxidants and PCD related genes within the dataset, which hampers a precise interpretation of the biological meaning of these genes.
Genes involved in quiescence
Quiescence, i.e. a reversible state in which a cell does not divide but retains the ability to re-enter cell proliferation, presents very conserved molecular features among evolutionary distant organisms ([3]). For example, both quiescent mammalian cells and yeasts arrest the cell cycle in G1, condense chromosomes, reduce rRNA synthesis and translation and activate autophagy mechanisms becoming more resistant to different stresses ([3] and references therein). The same features were recorded in the C. socialis transcriptome, as shown by the presence of genes related to autophagy and low translation, together with the nuclear rearrangement observed during spore formation. Among conserved pathways that have been related to quiescence in yeasts, there are several serine/threonine-protein kinases TOR; the activity of this enzyme decreases in N starved Saccharomyces cerevisiae, inducing the formation of spores ([53]). Different transcripts related to the TOR showed a marked positive expression (FCs from 3.5 to 6) when spores were present (T3 and T4) (Table S6). However, the same trend was observed for its inhibitor GATOR supporting once more the fact that different signals are produced by cells undergoing different fates, i.e. dying cells and cells that turn into spores.
Recent studies on resting cells in the dinoflagellate Scrippsiella trochoidea and the diatom P. tricornutum reported the possible involvement of the phytohormone abscisic acid (ABA) in regulating the transition between active cell division and quiescence ([54], [36], [55]). This molecule is a well know signal initiator in seed dormancy but it is still poorly studied in microalgae ([36]). In C. socialis, the expression of 9-cis-epoxycarotenoid dioxygenase, the rate-limiting enzyme of the ABA biosynthetic pathway, increased concomitantly to spore production (FCs from 2.6 to 6 in T3 and T4 versus control comparisons); the farnesyltransferase subunit beta is an essential part of the farnesyltransferase complex, an ABA negative regulator and maintained high FCs (from 5.3 to 6.7) at the same sampling points (Table S6).