Ostreococcus tauri (clonal isolate from OtH95) wild-type and transgenics were grown and monitored by flow cytometry as previously described (Degraeve-Guilbault et al., 2017). Vancomycin (1 mg/ml) was used to reduce bacterial contamination to less than 1% before experiments. Artificial sea-water base contained either 5 μM NaH₂PO₄ (phosphate limitation) or 35 μM. Cultures were grown in incubator-shaker (New Brunswick Innova 42R) with constant agitation (80 RPM) under white light (75 mmol photons m-2 s-1, 6 × T8 fluorescent bulbs 15 Watt each (Sylvania Gro-Lux). For screening of FA of O. tauri transgenics, cells were grown in T25 aerated culture flasks (Sartstedt, Nümbrecht, Germany) at 20°C. For O. tauri lipid analysis cells were grown in 200 mL of medium in 500 mL Erlenmeyer flasks. For temperature shift experiments cells were grown in either Erlenmeyer flasks or aerated T75 vertical flasks (100 mL medium in 250 mL flasks). O. tauri were acclimated at least for 10 generations (sub-cultivated twice) at a given temperature. Synechocystis sp. PCC6803 was grown accordingly to Kotajima et al. (2014). Nicotiana benthamiana plants were cultivated in a greenhouse under controlled conditions (16–8 h photoperiod, 25°C). Agrobacterium tumefaciens strain GV3101 was grown in Luria Broth medium at 30°C as previously described (Degraeve-Guilbault et al., 2020).

Sequences of putative ω3-desaturases were retrieved from genomic and transcriptomic data from NCBI and those from Mamiellophyceae species were manually checked for completion of Nt sequences; cTP were predicted from PredAlgo; alignment was performed using Snapgene trial version (Clustal omega). A codon-optimized sequences without the cTP (sequence start MTYNET) was used for expression Synechocystis sp. PCC6803 (Genewiz, Europe).

The O. tauri ω3-desaturase full-length sequence was amplified by PCR cycles from cDNA matrix using Q5^® Polymerase by two-step PCR (5′-ATGCGCGCCGCGACGTC-3′ and 5′-CTAGTCGCCCCGCTCCCAGAC-3′), cloned in pGEM^®-T Easy (Promega, Madison, WI, United States) and sequenced (Genwiz, Leipzig, Germany). Amplification from plasmid DNA was achieved with adapted primers to allow further cloning in pOtox-Luc (Moulager et al., 2010) (Restriction sites ApaI, AvrII) and using Gateway^® system according to manufacturer instruction (pDONR 221, pVT102-U-GW for Saccharomyces cerevisiae and pK7W2G2D or pK7YWG2 for N. benthamiana). Primers are provided in Supplementary Tables.

Overexpression in Synechocystis sp. PCC6803 was performed using the pTHT2031S vector after ligation to introduce the synthetic gene using the In-Fusion^® HD cloning kit (Takara Bio, Kusatsu, Japan). Primers are available from the Supplementary Data File.

For every RNA extraction FAs were analyzed in parallel. RNeasy-Plus Mini kit (Qiagen, Hilden, Germany) was used for RNA purification; DNase I was used to remove contaminating DNA (DNA-free kit, Invitrogen, Carlsbad, CA, United States) and cDNA obtained using the reverse transcription iScript^TM supermix kit (Bio-Rad, Hercules, CA, United States). Real-time RT quantitative PCR reactions were performed in a CFX96^TM Real-Time System (Bio-Rad) using the GoTaq^® qPCR Master mix (Promega, Madison, WI, United States). Bio-Rad CFX Manager software was used for data acquisition and analysis (version 3.1, Bio-Rad). Ct method was used to normalized transcript abundance with the references mRNA EF1α (elongation factor), CAL (calmodulin), and ACTprot2 (Actin protein-related 2). PCR efficiency ranged from 95 to 105%. Primers are available from the Supplementary Data File.

Ostreococcus tauri transformation was achieved using the pOtOXLuc vector and electroporation and transgenics were pre-screened accordingly to their luminescent level as previously described (Degraeve-Guilbault et al., 2020). Control lines are transgenics of empty vectors.

Nicotiana benthamiana were transformed by agroinfiltration of leaves from five-week old plants as previously described (Degraeve-Guilbault et al., 2020). Co-infiltration of RNA-silencing inhibitor P19 (equal volume of a bacterial suspension harboring pBin61-P19), was used in all experiments (Shah et al., 2013). DNA constructs were transferred by electroporation into the Agrobacterium tumefaciens GV3101 strain. Briefly, A. tumefaciens transformants were selected with antibiotics (gentamycin 25 μg/mL with spectinomycin 100 μg/mL or kanamycin 50 μg/mL). A. tumefaciens transformants were grown overnight, diluted to an optical density at 660 nm of 0.1, and grown up to 0.6–0.8. Cells were re-suspended in 5 mL sterilized H₂O for a final OD of 0.4 and 0.2 for overexpression and subcellular localization experiments, respectively. and 1 mL was agroinfiltrated using a synringe without needle. Plants were analyzed 2 and 5 days after A. tumefaciens infiltration for subcellular localization experiments and for overexpression, respectively.

Synechocystis sp. PCC6803 transformation was achieved by homologous recombination. Briefly, the plasmid was transformed into ten-time concentrated cells collected at mid-log phase. Subsequently, the cell was incubated at 30°C under white fluorescent lamps for 16–18 h and selected by 25 μg/mL chloramphenicol and 5 μg/mL spectinomycin on BG-11 solid media (1.5% w/v Bacto-agar).

For all organisms, fatty acid analyses and for O. tauri further lipid analysis were achieved accordingly to Degraeve-Guilbault et al. (2017). Organic solvents all contained butylhydroytoluene as an antioxidant (0.001%) and glassware was used. To gain resolution on O. tauri FA analysis, a minimum of 50 mL culture (approx. 1.5 × 10⁹ cells) was pelleted and extracted (glass beads beating and 1 h at 80°C) in 1 mL acidic methanol (2% v/v H₂SO₄) containing heptadecanoic acid (2 or 10 μg/ml) as internal standard; phase separation was achieved using 1 mL of NaCl 2.5% (or water for O. tauri) and 1 mL of hexane. The upper phase was collected in a new tube and concentrated to 100 μL under nitrogen stream. Four μL were injected for GG-FID analysis (Hewlett-Packard 5,890 series II, Agilent, Waldbronn, Germany) on a15 m × 0.53 mm × 1.2 μ Carbowax column (Altech, Deerfield, IL, United States). This procedure allowed for increasing the resolution of minor FA detection without any column saturation with major FA. Lipid extraction was performed as previously described (Degraeve-Guilbault et al., 2020). Briefly, the material was extracted using glass beads in chloroform:methanol (2:1 v/v), pelleted and extracted again until no pigment could be extracted. Phase separation was performed adding 0.5 v of NaCl 0.9%. Lipid developments were achieved by HP-TCL under 33% humidity in the ADC2-chamber system, (CAMAG). For O. tauri polar lipid were separated using methyl acetate/isopropanol/chloroform/methanol/KCl 0.25% (25:25:25:10:4 v/v/v/v) and neutral lipid using hexane/diethyl-ether/glacial-acetic-acid (60:10:1.22 v/v/v). For Synechocystis sp. PCC6803, polar lipids were separated using chloroform/methanol/glacial acetic acid/water (85:12:12:1 v/v/v/v). Lipids were stained with a solution of 0.02% primuline in 80:20 acetone/water (deeping for 1 min, air dried for 20 min).

Live cell imaging was performed using a Leica SP5 confocal laser scanning microscopy system (Leica, Wetzlar, Germany) equipped with Argon, DPSS, He-Ne lasers, hybrid detectors, and 63x oil-immersion objective. N. benthamiana leave samples were transferred between a glass slide and coverslip in a drop of water. Fluorescence was collected using excitation/emission wavelengths of 488/490–540 nm for chlorophyll, 488/575- 610 nm for YFP, and 561/710- 740 nm for m-cherry. Co-localization images were taken using sequential scanning between frames. Experiments were performed using strictly identical confocal acquisition parameters (e.g., laser power, gain, zoom factor, resolution, and emission wavelengths reception), with detector settings optimized for low background and no pixel saturation.

GraphPad Prism version 9.0.0 for Windows (GraphPad Software, San Diego, CA, United States¹), was used to compute statistical analysis following the recommendation of the user guide. PCA analyses was performed using the standardize method with parallel analysis option. The unpaired t-test with Welch correction (no assumption made about the variance of each group) was used when normality could be assessed otherwise the non-parametric Mann–Whitney test was used. Note that only the Shapiro–Wilk test for normality provided results when n = 3.

We previously showed that during O. tauri batch-growth the proportion of 18:3n3 was readily increased at the expense of 18:4n3 and therefore might mask some specific impact of temperature on these FAs (Degraeve-Guilbault et al., 2017). Consequently, preliminary experiments were conducted to determine high and low temperatures for which minimal differences were observed with regards to the growth rate between temperatures and maximal differences for FA. Temperatures tested corresponded to those were O. tauri has been detected in the environment (Supplementary Figure 1; Limardo et al., 2017). In our conditions, cell-growth and/or viability appeared to be impaired at 10 and 32°C. Growth did not appear to be impacted at 15 and 25°C and monitoring growth in more details at 14°C and 24°C confirmed that acclimated cells displayed a similar daily growth rate with a μmax (14°C 0.91 ± 0.1 d^–1 and 24°C 0.99 ± 0.146 d^–1) that was comparable to the one previously observed at 20°C (0.99 ± 0.006) (Degraeve-Guilbault et al., 2017; Supplementary Figure 1A and Figure 1A). Principal component analysis of FA and temperature values from preliminary experiment unveiled an inversed correlation of 18:5n3 and temperature, and a direct correlation of most ω6 FAs and 18:3n3 (Supplementary Figures 1C,D). To gain further insight into FA variations, the glycerolipid FA profile of cells acclimated at 14 and 24°C collected in mid-exponential growth (3.5 × 10⁷ cell mL^–1) was analyzed (Figure 1). The proportion of most ω3-PUFAs was overall increased at the expense of ω6-PUFAs at 14°C. The proportion of 14:0 was also reduced. It was particularly noteworthy that 18:3n3 and 18:5n3 varied in a reverse way while 18:4n3 remained overall stable (Figure 1B). the most statistically relevant differences were detected for the ω3 18:3, 18:5, 20:4, 20:5, 22:5, and for the ω6 16:2, 18:3 as well as for the monounsaturated FA 18:1n9; an obvious drop occurred for the ω6 18:2 and 20:4 though computed P values were higher (Figure 1C).

Glycerolipid analysis from acclimated cells showed that the proportion of 18:5n3 was increased at low temperature in both MGDG and DGDG (Figures 1D,E). For MGDG this seemed to occur at the expense of both 18:3n3 and 18:4n3 while only at the expense of 18:3n3 in DGDG. The 18:4n3 proportion was higher in both SQDG and PG which are lacking 18:5n3, and this increase appeared to occur at the expense of 18:3n3 in SQDG (Figures 1F,G). A higher proportion of 22:6n3 was observed in DGTA but neither in PDPT/PS nor in TAG (Figures 1I,H,J). On the other hand, variations of other minor VLC-PUFAs such as 20:4n6, 20:5n3, and 22:5n3 were reverberated into TAG, in which 20:4n3 was also detected only at low temperature (Figure 1J). These changes translated into a higher unsaturation degree of the bulk of glycerolipids, including TAG, as well as a higher ω3/ω6 ratio in most glycerolipids, that was the highest for PG (Supplementary Figures 2A,B).

The glycerolipid composition was also impacted (Figure 1K and Supplementary Figure 1C). Considering the cellular amount, DGDG was increased at the expense of MGDG at 14°C (Figure 1K). This feature has been reported decades ago in both higher plants and algae grown at low temperature and is assumed to be related to the stabilization of plastid membranes in response to various stresses including phosphate deprivation (Lynch and Thompson, 1982; Kuiper, 1985; Li and Yu, 2018). More surprisingly, the amount of SQDG was significantly reduced while that of phospholipids increased, a feature recalling increased phosphate availability, though not coherent with the variation observed in galactolipids (Van Mooy et al., 2009). Noteworthy TAG were twice as abundant at low temperature.

Altogether, our results indicate that low temperature acclimation correlates with an overall increase of ω3-PUFAs and a concomitant decrease of ω6-PUFAs in all structural glycerolipids. A drop of 18:3n3 occurred in all plastidial lipids and was concomitant of a rise of 18:5n3 in galactolipids whereas 18:4n3 was increased only in PG and SQDG.

From experience we know that O. tauri rhythms are readily synchronized by external cues (Moulager et al., 2007; Monnier et al., 2010). In order to gain the most accurate insight into the kinetics of changes upon temperature shift, the cultures were synchronized by light-dark cycles (L/D) of 18–6 h. This experimental design aimed to restrain time resetting and cell re-synchronization by temperature shift. Indeed, continuous light does not abolish circadian rhythms but merely result in progressive desynchronization of rhythms between individual cell (Bieniawska et al., 2008). As temperature is a strong cue for time resetting, the cells switched to a novel temperature are most likely resynchronized while the control cell are not (McClung and Davis, 2010). This results in comparing different internal times. In contrast, under diurnal cycles internal rhythms and entrained, i.e., synchronized and set on the external time; temperature changes do not change the phase of the rhythms. Because temperature responses were reported to be augmented in the morning and since in O. tauri desaturase expression is known to peak from late night to mid-day, temperature-shifts were achieved [4.5 h after light on (T0)] (Supplementary Figure 3).

Cells acclimated at 14 and 24°C and collected at the time of the temperature shift displayed FA differences that were closely related to those observed under continuous light, though the P varied, with the further variation of 16:4n3 that was lower at 14°C (Figures 2A,B). Principal component analysis was used to test the relationship between FA variations and temperature in this homogeneous data set; it unveiled that minor ω3 PUFAs as well as 18:5n3, 16:0, 18:1, and were related and negatively correlated to temperature. In contrast, 20:4n6, 18:3n6 as well as 16:4n3 displayed correlation scores that were close to the temperature variable while 18:2, 16:2 as well as 18:3n3 variations were also explained by the second principal component (Figure 2C).

The swiftness of FA-variations upon chilling (24–14°C) and warming (14–24°C) was investigated achieving early sampling times after the temperature and following variations up to 36 h (Figure 3). The data were averaged from independent experiments. As the proportion of 18:3n3 and 18:4n3 were slightly different between experiments (due to the nutrient state), an individual kinetics is further provided as Supplementary Data, in order to better highlight the most earlier changes (Supplementary Figure 4).

As both FA and transcriptional variations strongly suggested the involvement of the ω3-Des in the temperature acclimation process, we next focused on the only ω3-Des candidate found in O. tauri (Otpω3-Des) and related species (Degraeve-Guilbault et al., 2020). These putative ω3-Des have been reported to be orthologous of the Δ15-Des of Emiliania huxleyi (EhΔ15-Des) and cluster apart from previously characterized Δ15-Des. The authors hypothesized that EhΔ15-Des was involved in the accumulation of 18:5n3 that occurs in E. huxleyii at low temperature (Kotajima et al., 2014).

All organisms combined, the inverse correlation of FA unsaturation with temperature appears to be a universal trend. According to the homeoviscous hypothesis, the increase of FA unsaturation and thereby the decrease of the FA melting point, is necessary to maintain the fluid state of biological membranes at low temperature (Ernst et al., 2016). However, it should be emphasized that the addition of just one double-bound to saturated acyl-chain has the most drastic impact on the FA melting temperature and that the involvement of PUFAs for adjusting biomembrane physical properties has not been demonstrated; on the contrary omega-3 PUFAs failed to exhibit any peculiar fluidifying potency compared to oleic acid (De Santis et al., 2018). Although photosynthesis defects have been reported for mutants with altered UFAs content in plants and microalgae, it seems much more difficult to establish a causal link between these defects and impaired membrane fluidity (Vijayan and Browse, 2002; Falcone et al., 2004; Kugler et al., 2019). In cyanobacteria, ω3-PUFAs are only produced at chilling temperature and a direct link between membrane fluidity and DesB transcription has been demonstrated (Tasaka et al., 1996; Los et al., 2013; Mironov et al., 2018). However, neither membrane physical properties nor growth nor photosynthesis were significantly impacted in mutants lacking trienoic FAs. In contrast, plant mutants lacking trienoic FAs were reported to display distinct patterns of symptoms including sever thylakoid loss for cold temperature but also lower growth at high temperature (Routaboul et al., 2000). As trienoic PUFAs are the precursors of plant oxylipins and Nitro-FAs, it might be that these patterns are the result of impaired temperature signaling in plants (Mata-Perez et al., 2018; He and Ding, 2020; Yu et al., 2020). Though, establishing a relationship between PUFA membrane precursors and signaling down-products is a challenging issue, it should be kept in mind that oxylipins are also occurring in microalgae (Lauritano et al., 2016; Lupette et al., 2018).

As regards microalgae, temperature has been shown to impact FA profile in a species-dependent manner and it appears most difficult to discern a general trend from literature (Renaud et al., 1995; Boelen et al., 2013; Aussant et al., 2018; Gill et al., 2018; Balakrishnan and Shanmugam, 2020). Though it cannot be viewed as a general rule, the increase of ω3-PUFAs has been reported in response to chilling in various species of green microalgae while the increase of 18:5n3 and/or 20:5n3 was reported in Chromista species (see below) (Leblond et al., 2010; Kotajima et al., 2014; Aussant et al., 2018). The higher proportion of ω3 was in some case concomitant of a decline of the corresponding ω6-precursor (Nguyen et al., 2013; Zorin et al., 2017). The physiological relevance of PUFA variations have been only sporadically tackled. Works studying microalgal mutants with altered PUFA content at different temperature are scarce and overall provide mild evidences that mutations affect growth and/or photosynthetic processes (Sukenik et al., 1998; Nguyen et al., 2013; Zorin et al., 2017). For instance, the growth of CrFAD7 knock-out mutant was not impaired at low temperature; extreme high temperature (45°C) was necessary to highlight that the mutation was associated with a reduced impairment of PSII activity. In these studies, early variations have not been investigated.

In the present work, experimental conditions were optimized for the identification of early temperature-specific FA variations. O. tauri growing temperature is commonly fixed at 20°C. In our conditions, O. tauri readily acclimated to 14°C and grew equally well at 24°C, in coherence with the range of temperature at which Ostreococcus species were identified in the environment (Limardo et al., 2017). Therefore, we can assume that temperature shifts between 14 and 24°C and reciprocally are moderate as regards to the thermoacclimation capacity of O. tauri. Finally, L/D entrainment was used to restrain internal time resetting by temperature reducing the chance that the differences observed are indirectly resulting from circadian rhythms shifts. It should be emphasized that early changes are more likely to be related to direct response to temperature whereas late variations may rather be related to indirect general metabolism adjustment. Chilling and warming triggered a swift and reverse adjustment of the ω6/ω3 PUFA ratio in all glycerolipids; the earliest variations occurred in C18-PUFAs (known to predominate in plastidic lipids), including 18:5n3, which is exclusively located in galactolipids (Degraeve-Guilbault et al., 2017). The progressive decrease of 20:4n6 upon chilling was also a robust trend. As evoked, compared to monounsaturated FA, the fluidizing potency of PUFAs have not been demonstrated to be better (De Santis et al., 2018). Though, the proportion of 18:1 is very low in O. tauri and even if it is increased in 14°C acclimated cells, other FAs, sterols and/or pigments might participate for acclimatizing membrane fluidity PUFAs are also known precursors of oxylipins and of Nitro-FAs in various organism ranging from microalgae to animals and their production could, to some extent, be related to the variation of membrane PUFAs and be involved in the signaling of temperature changes (Lupette et al., 2018; Wasternack and Feussner, 2018).

In the present work, the FA-phenotype of pω3-Des OE was weak and it was therefore not surprising that no growth defect could be detected. We assume that either compensatory mechanisms are taking place in pω3-Des OE and/or that our conditions are not appropriate and/or stringent enough to unveil any defects. Nevertheless, the fact that pω3-Des OE displayed lipid features closely related to those of low temperature acclimated cells was obvious and coherent with the involvement of pω3-Des in temperature acclimation. The reduction of 14:0 at low temperature might possibly be part of the homeoviscous process required at low temperature but in this case, would not be expected in the pω3-Des OE.

Most interestingly, the 18:5n3 variations observed in O. tauri were reported in evolutionary distant microalgae (Renaud et al., 1995; Kotajima et al., 2014). As evoked, prasinophyta (including Mamiellophyceae) are the only primary endosymbiotic organisms that encompass 18:5n3, which is commonly found in microalgae emerging from secondary endosymbiosis. In the haptophytes, E. huxleyi and Isochrysis species 18:5n3 was reported to increase at low temperature while it was found at higher percentage (together with 18:4n-3) in cold-adapted dinoflagellates (Leblond et al., 2006). We show here that the content of 18:5n3 is tightly regulated by temperature in a species from the green lineage and varies as early as 6 h after temperature-shift. This result points out the involvement of this peculiar FA in temperature responses may be an ancient feature. Since 18:5n3 is predominantly located in glycerolipids across all species, this regulation might be related to the fine-tuning of photosynthesis. The increase of 20:5n3 upon chilling has been reported for several species for which 20:5n3 is a major component of galactolipids and further evidences suggested that it could be involved in photosynthesis, in particular in non-photochemical quenching in Nannochloropsis gaditana CCMP526 (Renaud et al., 1995; Camacho-Rodriguez et al., 2013; Dolch et al., 2017; Gill et al., 2018). On the other hand, 20:5 derived isoprostanes were characterized from Phaeodactylum tricornutum (Lupette et al., 2018). Interestingly their content was inversely correlated to that of 20:5 FA. It is therefore possible that non-enzymatic oxylipin derived from pentaenoic FAs might be involved in a retrograde signaling for temperature stress.

Biosynthesis of 18:5n3 is an enigmatic issue; several hypotheses have been proposed including a yet not identified Δ3-Des or the shortening of 20:5n3 (Joseph, 1975). Though no Δ3-Des have been identified to date, a positional isomer of stearidonic acid encompassing a double bond at position three (18:4^Δ3,6,9,12) in a thermophilic cyanobacterium has been described; as this species, like the bulk of cynaobacteria, do not produce VLC-PUFAs, this result is in favor of Δ3-Des activity in cyanobacteria (Rezanka et al., 2012). In O. tauri, the recent discovery of plastidial Δ6-desturases showed that the plastidic C18-PUFA-pool is regulated independently of ER-Des and therefore supports the existence of a yet not identified plastidial Δ3-Des in this species. This activity could be carried by either one of the plastidial Δ6-Des, as previously discussed elsewhere or by the Δ5-Des whose expression is tightly regulated by the temperature (Degraeve-Guilbault et al., 2020). It should be recalled here that Bathycoccus prasinos lacks both the pΔ6-Des2 and 18:5 (Degraeve-Guilbault et al., 2017). On the other hand, several Δ5-Des from animals were reported to display extended Δ6 and Δ4 regiospecificities and plant Des regiospecificity was shown to switch with sub-cellular localization (Heilmann et al., 2004; Li et al., 2010). Noteworthy, a cTP is predicted for the Δ5-Des and the recent genomic sequence assembly from the O. tauri strain RCC1115 further pinpointed an additional ORF opening the possibility that alternative translation might be used to produce two differentially located isoforms of Δ5-Des (Hoffmann et al., 2008).

The 20:5 shortening hypothesis cannot be excluded and is supported for O. tauri by: (1) the existence of sn-1/sn-2 20:5/16:4 DGTA, (2) the occurrence of high level of 20:5 in the acyl-CoA pool and the detection of 18:5n3, and (3) the occurrence of 18:5 in DAGs (Degraeve-Guilbault et al., 2017). From these observations, it can be speculated that either 18:5-CoA is transferred to the plastid after 20:5-CoA shortening in the cytosol and specifically esterified to lyso-16:4 galactolipids, or that 18:5-DAG arising from remodeled DGTA 20:5/16:4 DGTA serve as precursors for 18:5-galactolipid synthesis. Until a Δ3-activity can be demonstrated the two hypotheses remain equally speculative.

Ostreococcus tauri biological processes have been previously shown to highly rely on orchestrated transcriptional regulation (Monnier et al., 2010). Clustering of genes according to their temporal waveforms suggested that master transcriptional regulations are at work to coordinate lipid metabolism with chloroplast and carotenoids biogenesis at late night and with photosynthesis, oxidative stress, and DNA repair at mid-day. Transcriptional rewiring in response to temperature, especially chilling, has been extensively studied in freshwater cyanobacteria and plants and are beginning to be studied in marine cyanobacteria (Sinetova and Los, 2016; Shi et al., 2018; Breton et al., 2020; Guyet et al., 2020). However, there is a large gap of knowledge with regards to microalgae. Desaturase transcriptional induction in response to temperature and/or membrane fluidity has been demonstrated in freshwater cyanobacteria (Los et al., 1997). In Arabidopsis, the up-regulation of plastidial ω3-Des FAD8 upon cold/chilling was shown to involve both transcriptional and post-transcriptional regulations (Gibson et al., 1994; Matsuda et al., 2005). Higher transcript level of ω3-Des occurred in Chlorella vulgaris and C. reinhardtii at chilling temperature and early transcriptional activation of Δ6-Des has been reported for Isochrysis sp. (Suga et al., 2002; Nguyen et al., 2013; Wang et al., 2016). Part of our aim was to identify whether desaturases expression was regulated by temperature and how far these regulations were coherent with FA variations. Chilling was shown to swiftly and sustainably up-regulated most desaturase gene under both L/D and continuous light conditions while warming had an overall reverse impact. Under L/D, the plastidial desaturase genes pΔ6-Des2 and pω3-Des as well as Δ5-Des were the most up-regulated genes. Conversely, warming repressed pω3-Des, pΔ6-Des1, andΔ5-Des and to a greater extent Acyl-CoA-Δ6-Des. These regulations are overall coherent with the activation of the ω3 pathway upon chilling and its repression upon warming. Since we previously showed that overexpression Acyl-CoAΔ6-Des in O. tauri mostly impacted TAG-FA profile, the temperature dependent regulation of Acyl-CoA-Δ6-Des expression might possibly be related to the accumulation of highly unsaturated TAG at low temperature. The differential transcriptional regulation of the plastidial Δ6-desaturases pΔ6-Des1 and pΔ6-Des2, demonstrated to have different specificity for ω6 and ω3 might be required for proper adjustment of the ω3/ω6 ratio in response to temperature (Degraeve-Guilbault et al., 2020).