A 13CO2 Enrichment Experiment to Study the Synthesis Pathways of Polyunsaturated Fatty Acids of the Haptophyte Tisochrysis lutea

The production of polyunsaturated fatty acids (PUFA) in Tisochrysis lutea was studied using the gradual incorporation of a 13C-enriched isotopic marker, 13CO2, for 24 h during the exponential growth of the algae. The 13C enrichment of eleven fatty acids was followed to understand the synthetic pathways the most likely to form the essential polyunsaturated fatty acids 20:5n-3 (EPA) and 22:6n-3 (DHA) in T. lutea. The fatty acids 16:0, 18:1n-9 + 18:3n-3, 18:2n-6, and 22:5n-6 were the most enriched in 13C. On the contrary, 18:4n-3 and 18:5n-3 were the least enriched in 13C after long chain polyunsaturated fatty acids such as 20:5n-3 or 22:5n-3. The algae appeared to use different routes in parallel to form its polyunsaturated fatty acids. The use of the PKS pathway was hypothesized for polyunsaturated fatty acids with n-6 configuration (such as 22:5n-6) but might also exist for n-3 PUFA (especially 20:5n-3). With regard to the conventional n-3 PUFA pathway, Δ6 desaturation of 18:3n-3 appeared to be the most limiting step for T. lutea, “stopping” at the synthesis of 18:4n-3 and 18:5n-3. These two fatty acids were hypothesized to not undergo any further reaction of elongation and desaturation after being formed and were therefore considered “end-products”. To circumvent this limiting synthetic route, Tisochrysis lutea seemed to have developed an alternative route via Δ8 desaturation to produce longer chain fatty acids such as 20:5n-3 and 22:5n-3. 22:6n-3 presented a lower enrichment and appeared to be produced by a combination of different pathways: the conventional n-3 PUFA pathway by desaturation of 22:5n-3, the alternative route of ω-3 desaturase using 22:5n-6 as precursor, and possibly the PKS pathway. In this study, PKS synthesis looked particularly effective for producing long chain polyunsaturated fatty acids. The rate of enrichment of these compounds hypothetically synthesized by PKS is remarkably fast, making undetectable the 13C incorporation into their precursors. Finally, we identified a protein cluster gathering PKS sequences of proteins that are hypothesized allowing n-3 PUFA synthesis.


Introduction
Long chain polyunsaturated fatty acids (LC-PUFA) such as 20:5n-3 (EPA) and 22:6n-3 (DHA) are important compounds for most marine metazoans for their growth, reproduction, and development. They are not able to synthetize them in sufficient quantities and thus have to acquire them from their diet. On the basis of the food web, protists are the

Introduction
Long chain polyunsaturated fatty acids (LC-PUFA) such as 20:5n-3 (EPA) and 22:6n-3 (DHA) are important compounds for most marine metazoans for their growth, reproduction, and development. They are not able to synthetize them in sufficient quantities and thus have to acquire them from their diet. On the basis of the food web, protists are the main producers of these fatty acids and present a key role in marine ecosystem functioning. 20:5n-3 and 22:6n-3 are also particularly important in human nutrition. They are known to have beneficial effects on cardiovascular diseases or diabetes. However, due to high demand for human nutrition and aquaculture of carnivore species, a shortage of these two compounds found in fish oil is predicted to occur by 2050 [1]. Despite their economic and ecologic interests, biological and ecological processes responsible for their synthesis are still under investigation. It is, then, of first concern to understand how 20:5n-3 and 22:6n-3 are produced at the basis of the food webs, and how global changes could affect their availability at higher trophic levels.
In phytoplankton and microzooplankton, fatty acids are synthetized via different metabolic pathways [2][3][4]. The most "conventional" pathway is the fatty acid synthase (FAS) pathway, followed by the elongation and front-end desaturation steps of the n-3 and n-6 pathways. Starting with the initial formation of acetyl-CoA and then malonyl-CoA in aerobic conditions, these pathways produce more complex fatty acids by progressive addition of two atoms of carbon (elongation steps) or desaturations of precursors such as 16:0 or 18:0 [5][6][7]. These two pathways can be connected by the so-called ω-3 desaturase (or methyl end desaturase) pathway. Within the n-3 and n-6 pathways, an alternative route of Δ8 desaturation can also bypass the Δ6 desaturation step and has already been identified in Haptophyte [8]. These routes allowed the synthesis of 20:5n-3 as well as 22:6n-3 ( Figure 1).

Figure 1.
Microalgae fatty acid synthesis pathways. Desaturases are noted with "ΔX" (yellow arrows) and "ωY-des (ΔX)" (blue arrows), where X refers to the location of carbon holding the newly formed double bond from the front end (or carboxyl end) and Y its position from the methyl end. Elo: elongase, FAS: fatty acid synthase.
The present study aims at investigating the synthesis pathways of essential PUFA 20:5n-3 and 22:6n-3 using stable isotope ( 13 C) labelling experiment of the haptophyte Tisochrysis lutea. T. lutea is intensively used in aquaculture (hatchery) and industry [40]. The incorporation of 13 C was monitored in 11 FA during 24 h at a high temporal resolution (each 0.5 to 2 h). Progressive accretion of the 13 C-labelled CO 2 into FA (from precursors to PUFA of interest) allowed us to constrain FAS, elongase/desaturase, and PKS involvement in 20:5n-3 and 22:6n-3 production by Tisochrysis lutea. In parallel to the monitoring of 13 C incorporation into FA, growth, physiological status, and other cellular parameters (morphology, viability, esterase activity, and lipid content) were monitored by flow cytometry analysis.

Algae Physiology and Biochemistry during the 24 h Experiment
Cell abundance for Tisochrysis lutea during the 24 h of experiment increased sharply from t 0 to t 24 . The experiment allowed cell concentration to double for the three balloons ( Figure 2A). Despite the attention given to homogenization at the time of subculture from inoculum, the second enriched balloon had a cell concentration twice higher than Tl1 and TlT. This difference remained constant during the entire experiment. Cell abundance varied from 4.3 × 10 6 cells·mL −1 to 9.5 × 10 6 cells·mL −1 for the most concentrated balloon TI2 and from, on average, 2.6 × 10 6 cells·mL −1 to 6.4 × 10 6 cells·mL −1 for the two others ( Figure 2A). Despite the concentration differences, the general slopes for the three balloons were very similar (0.22 cells·mL −1 ·h −1 for Tl2 and 0.15 cells·mL −1 ·h −1 for Tl1 + TlT) ( Figure 2A). Bacteria were also found in higher abundance Tl2 ( Figure 2B), almost five times higher than in Tl1 and TlT. However, bacteria increased only by a factor of 1.2 for Tl2 between t 0 and t 24 versus a factor 3.5 in average for Tl1 and TlT. Bacteria concentration was around 6.6 times higher than algae concentration for Tl1 and TlT on average over the 24 h of the experiment. For Tl2, bacteria concentration was 16 times higher than algae concentration at the beginning of the experiment, but this ratio decreased progressively until t 24 (8.7 times higher) ( Figure 2B).
FSC and SSC were, respectively, considered a proxy of cell size and cell complexity, and FL3 was considered a proxy of chlorophyll content. SSC and red fluorescence (FL3) did not significantly vary during the entire experiment (Bartlett tests, p > 0.05 and ANOVA p > 0.05) (Table S1). FSC increased slightly with time for the three balloons (Bartlett test, p > 0.05 and ANOVA p = 0.03) ( Table S1). The percentage of dead microalgae (as measured by SYTOX staining assay) remained below 7% for the 24h of the experiment (Table S1).
Particulate organic carbon concentration increased similarly to the cell abundance for the two labelled balloons from 3.6 to 9.0 mmolC·L −1 for Tl1, from 4.5 to 11.6 mmolC·L −1 for Tl2, and from 3.4 to 8.2 mmolC·L −1 for TlT ( Figure 3A). Increase in POC, as a function of experiment duration (R 2 = 0.81, p < 0.0001), occurred at a relatively constant rate of 0.2 mmolC·L −1 ·h −1 for Tl1, Tl2, and TlT considered together. Total fatty acid (TFA) concentration also increased for the three balloons (between 0.26 mmolC·L −1 to 0.60 mmolC·L −1 for Tl1, from 0.25 mmolC·L −1 to 0.80 mmolC·L −1 for Tl2, and from 0.17 mmolC·L −1 to 0.55 mmolC·L −1 for TlT) ( Figure 3B). The slope was 15 µmolC·L −1 ·h −1 (R 2 = 0.72, p < 0.0001). POC was significantly correlated to cell concentration (R 2 = 0.80 p < 0.0001). The slope of the relation between POC and cell abundance is a proxy of carbon content per cell for T. lutea, which was, on average for the three balloons, equal to 1.07 fmolC·cell −1 ( Figure 3C). TFA concentration was linearly and positively correlated with POC concentration (R 2 = 0.73, p < 0.0001). The slope of the regression between TFA and POC concentration indicates that TFA represent in average 7.7% of bulk POC ( Figure 3D). FSC and SSC were, respectively, considered a proxy of cell size and cell complexity, and FL3 was considered a proxy of chlorophyll content. SSC and red fluorescence (FL3) did not significantly vary during the entire experiment (Bartlett tests, p > 0.05 and ANOVA p > 0.05) (Table S1). FSC increased slightly with time for the three balloons (Bartlett test, p > 0.05 and ANOVA p = 0.03) ( Table S1). The percentage of dead microalgae (as measured by SYTOX staining assay) remained below 7% for the 24h of the experiment (Table S1).

13 C Atomic Enrichment (AE) of Particulate Organic Carbon and Dissolved Inorganic Carbon
Dissolved inorganic carbon (DIC) progressively enriched in the two balloons with 13 CO2 ( Figure 4A). The enrichment trends were similar for the two balloons after t4, with an important increase in the DIC atomic enrichment until t20 (up to 58.1 and 61.1% for Tl1 and Tl2, respectively). The increase in AE tended to stabilize after t20. Final levels of enrichment were 61.5 and 64.6% for Tl1 and Tl2, respectively ( Figure 4A). Atomic enrichment (AEPOC) increased sharply after t4 for the two balloons until the end of the experi-

13 C Atomic Enrichment (AE) of Particulate Organic Carbon and Dissolved Inorganic Carbon
Dissolved inorganic carbon (DIC) progressively enriched in the two balloons with 13 CO 2 ( Figure 4A). The enrichment trends were similar for the two balloons after t 4 , with an important increase in the DIC atomic enrichment until t 20 (up to 58.1 and 61.1% for Tl1 and Tl2, respectively). The increase in AE tended to stabilize after t 20 . Final levels of enrichment were 61.5 and 64.6% for Tl1 and Tl2, respectively ( Figure 4A). Atomic enrichment (AE POC ) increased sharply after t 4 for the two balloons until the end of the experiment. Enrichment levels at t 24 were 25.2 and 34.7% for Tl1 and Tl2, respectively ( Figure 4B).

Fatty Acid Composition in Neutral and Polar Lipids in T. lutea
Neutral lipids and polar lipids represented, respectively, 37% and 63% of TFA on average for the three balloons ( Figure 5A). The proportions of individual fatty acid in NL and PL did not vary throughout the experiment. Total bacteria fatty acids (iso15:0, ante15:0, iso16:0, iso17:0, 15:0, 17:0, 21:0, 15:1n-5-Tables S2-S4) remained below 1% for both NL and PL fractions during the 24 h. Branched fatty acids were only present in trace amounts (Tables S2-S4). Concentrations in µg·L −1 and µmolC·L −1 as well as proportions in% of all identified and quantified FA in neutral and polar lipid fractions according to sampling time are available in the Supplementary Files (Tables S2-S4).

Fatty Acid Composition in Neutral and Polar Lipids in T. lutea
Neutral lipids and polar lipids represented, respectively, 37% and 63% of TFA on average for the three balloons ( Figure 5A). The proportions of individual fatty acid in NL and PL did not vary throughout the experiment. Total bacteria fatty acids (iso15:0, ante15:0, iso16:0, iso17:0, 15:0, 17:0, 21:0, 15:1n-5-Tables S2-S4) remained below 1% for both NL and PL fractions during the 24 h. Branched fatty acids were only present in trace amounts (Tables S2-S4). Concentrations in µg·L −1 and µmolC·L −1 as well as proportions in% of all identified and quantified FA in neutral and polar lipid fractions according to sampling time are available in the Supplementary Files (Tables S2-S4).

Identification of Candidate Proteins for PKS Synthesis in T. lutea
Thirty sequences of potential candidate proteins involved in T. lutea PUFA synthesis have been identified and are presented in Supplementary File (Table S6). Only fourteen presented the four main domains potentially coding for the enzymes used in PKS PUFA synthesis pathways: ketoacyl reductase (KR), polyketide synthase (KS), dehydrase/dehydrogenase (DH), and enoyl reductase (ER). Among these sequences, four sequences (TISO_14962, TISO_14968, TISO_14975, and TISO_14977) were part of the same cluster (group of homologous proteins) and presented multiple KS, KR, ER, and DH domains as well as phosphopantetheine (PP)-binding domains (Figure 7). TISO_14962 also possessed methyltransferases and thioesterase domains (Figure 7). TISO_14977 presented a domain acknowledged to be involved in acetyl-CoA synthesis. Within this cluster, TISO_14973 was also selected, as it contains an atypical domain, specifically recognized as being involved in n-3 PUFA synthesis. Nine other sequences (TISO_04539,  TISO_06404, TISO_06537, TISO_08047, TISO_11097 TISO_16495, TISO_27353, TISO_37260, and TISO_37631) were also found, containing the four main domains (up to 18 for KR in TISO_08047). Except TISO_37631, these sequences also have thioesterase, sulfotransferase, or peptide-synthesis-related domains, and thus they might be in charge of the synthesis of more complex lipids. volved in n-3 PUFA synthesis. Nine other sequences (TISO_04539, TISO_06404,  TISO_06537, TISO_08047, TISO_11097 TISO_16495, TISO_27353, TISO_37260, and  TISO_37631) were also found, containing the four main domains (up to 18 for KR in TISO_08047). Except TISO_37631, these sequences also have thioesterase, sulfotransferase, or peptide-synthesis-related domains, and thus they might be in charge of the synthesis of more complex lipids. lutea. The name of each protein is annotated with TISO_ (for Tisochrysis lutea) and associated number. In the legend, the text written in bold italic correspond to domain names as shown in NCBI conserved domain database, followed by its suspected role.

Discussion
This study investigated long-chain PUFA synthesis pathways in the haptophyte Tisochrysis lutea using the incorporation of 13 CO2. Addition of 13 CO2 did not affect T. lutea physiology. Cell viability remained above 93% during the experiment, while cell complexity and chlorophyll content did not vary significantly according to sampling time. Cell size (as attested by FSC) increased slightly during the 24 h experiment. T. lutea produced FA to a level of 7% of POC; predominantly in the form of PL (66%).

Discussion
This study investigated long-chain PUFA synthesis pathways in the haptophyte Tisochrysis lutea using the incorporation of 13 CO 2 . Addition of 13 CO 2 did not affect T. lutea physiology. Cell viability remained above 93% during the experiment, while cell complexity and chlorophyll content did not vary significantly according to sampling time. Cell size (as attested by FSC) increased slightly during the 24 h experiment. T. lutea produced FA to a level of 7% of POC; predominantly in the form of PL (66%).
The final level of atomic enrichment (AE) into the different FA witnessed active synthesis, as most fatty acids had a higher AE than that of POC (30% on average for the two balloons). 22:5n-6 was the most enriched long chain PUFA (LC-PUFA) in the PL fraction. 22:5n-6 and 18:2n-6 were the only 13 C labelled n-6 fatty acids detectable by GC-c-IRMS. None of the known synthesis intermediates (18:3n-6, 20:3n-6, 20:4n-6, and 22:4n-6) between 18:2n-6 and 22:5n-6 [4] had measurable 13 C-labelling and were below 1% in the FA profile during our experiment. It is then difficult to hypothesize the pathway used to create 22:5n-6 with this missing information. However, even though the different intermediates were undetectable, 18:2n-6 and 22:5n-6 atomic enrichments being very close cannot exclude them to be related to each other. While studying the existence of an alternative ∆8 desaturase in Haptophyte, Qi et al. (2002) [8] noticed the absence of intermediates of the n-6 ∆8 desaturase pathway (20:2n-6, 20:3n-6 and 20:4n-6)) in Isochrysis galbana. It was attributed to relatively high active enzymes that could form the end-product 22:5n-6 with a rapid flow through these n-6 intermediates. Our results agree with this, as 13 C enrichment of n-6 intermediates could not be detected by compound specific isotope analysis. To demonstrate the existence of these pathways, it would be interesting to combine functional analysis of desaturases by expression in yeast and GC-c-IRMS monitoring of the intermediates after 13 C labelling of their precursors. However, it is also possible that another pathway not involving "classical" n-6 FA intermediates exist in T. lutea. Previous studies showed the existence of PKS genes in various species of the prymnesiophytes including Isochrysis galbana [47], closely phylogenetically related to Tisochrysis lutea. We identified five candidates; proteins potentially involved in PKS synthesis pathway in T. lutea. Even if their function has not been verified, it is possible that at least one of the proteins presented in Figure 7 was responsible for the formation of n-6 PUFA in the haptophyte. Thus, our hypothesis is that an n-6 PKS pathway might also exist in T. lutea ( Figure 8). Finally, PKS and "classical" n-6 routes might not be completely independent and could interact in the synthesis of 22:5n-6 in T. lutea. balloons). 22:5n-6 was the most enriched long chain PUFA (LC-PUFA) in the PL fraction. 22:5n-6 and 18:2n-6 were the only 13 C labelled n-6 fatty acids detectable by GC-c-IRMS. None of the known synthesis intermediates (18:3n-6, 20:3n-6, 20:4n-6, and 22:4n-6) between 18:2n-6 and 22:5n-6 [4] had measurable 13 C-labelling and were below 1% in the FA profile during our experiment. It is then difficult to hypothesize the pathway used to create 22:5n-6 with this missing information. However, even though the different intermediates were undetectable, 18:2n-6 and 22:5n-6 atomic enrichments being very close cannot exclude them to be related to each other. While studying the existence of an alternative Δ8 desaturase in Haptophyte, Qi et al. (2002) [8] noticed the absence of intermediates of the n-6 Δ8 desaturase pathway (20:2n-6, 20:3n-6 and 20:4n-6)) in Isochrysis galbana. It was attributed to relatively high active enzymes that could form the end-product 22:5n-6 with a rapid flow through these n-6 intermediates. Our results agree with this, as 13 C enrichment of n-6 intermediates could not be detected by compound specific isotope analysis. To demonstrate the existence of these pathways, it would be interesting to combine functional analysis of desaturases by expression in yeast and GC-c-IRMS monitoring of the intermediates after 13 C labelling of their precursors. However, it is also possible that another pathway not involving "classical" n-6 FA intermediates exist in T. lutea. Previous studies showed the existence of PKS genes in various species of the prymnesiophytes including Isochrysis galbana [47], closely phylogenetically related to Tisochrysis lutea. We identified five candidates; proteins potentially involved in PKS synthesis pathway in T. lutea. Even if their function has not been verified, it is possible that at least one of the proteins presented in Figure 7 was responsible for the formation of n-6 PUFA in the haptophyte. Thus, our hypothesis is that an n-6 PKS pathway might also exist in T. lutea ( Figure 8). Finally, PKS and "classical" n-6 routes might not be completely independent and could interact in the synthesis of 22:5n-6 in T. lutea. Despite being one of the most abundant FA, 18:4n-3 showed a low 13 C-enrichment (23%). The synthesis of 18:4n-3 from 18:3n-3 by ∆6 desaturase had already been described by Isochrysis sp. [48]. We assume that such activity also exists in Tisochrysis, phylogenetically close to Isochrysis. However, as 18:3n-3 co-elute with 18:1n-9, it was not possible to measure its AE and to assess whether this could be a limiting step in n-3 pathway (Figure 9). The 18:5n-3 had the lowest enrichment, and the ratio 18:5n-3/18:4n-3 was below the threshold value (R = 0.78), indicating a feasible transformation of 18:4n-3 into 18:5n-3. The existence of ∆3 desaturase that could support the production of 18:5n-3 (18:5∆3,6,9,12,15) from 18:4n-3 (18:4∆6,9,12,15) had been suggested by Joseph (1975) [49] to explain the presence of this unusual FA in dinophytes. A more recent study by Ahman et al. (2011) [23] showed in Ostreococcus lucimarinus that a ∆4 desaturase was surprisingly able to add a double bond in 18:4n-3 at the ∆3 position leading to the formation of 18:5n-3 when the gene was expressed in yeast cell and supplemented by 18:4n-3 as substrate. With our results and the discovery of Ahman et al. (2011) [23], we proposed that a ∆4 desaturase of T. lutea might be able to act as a ∆3 desaturase on 18:4n-3 to produce 18:5n-3 ( Figure 9). Desaturation of 18:4n-3 into 18:5n-3 had been previously hypothesized by Kotajima et al. (2014) [50] in the prymnesiophyte Emiliania huxleyi. from 18:4n-3 (18:4Δ6,9,12,15) had been suggested by Joseph (1975) [49] to explain the presence of this unusual FA in dinophytes. A more recent study by Ahman et al. (2011) [23] showed in Ostreococcus lucimarinus that a Δ4 desaturase was surprisingly able to add a double bond in 18:4n-3 at the Δ3 position leading to the formation of 18:5n-3 when the gene was expressed in yeast cell and supplemented by 18:4n-3 as substrate. With our results and the discovery of Ahman et al. (2011) [23], we proposed that a Δ4 desaturase of T. lutea might be able to act as a Δ3 desaturase on 18:4n-3 to produce 18:5n-3 ( Figure 9). Desaturation of 18:4n-3 into 18:5n-3 had been previously hypothesized by Kotajima et al. (2014) [50] in the prymnesiophyte Emiliania huxleyi.
The 18:5n-3 was also described as an intermediate of 22:6n-3 synthesis by PKS pathway [4]. However, its low enrichment, as compared to 22:6n-3, appeared not compatible with a hypothetical production through this pathway. Nevertheless, one may speculate that there are two separated PKS pathways, one for the 22:6n-3 and one for the 18:5n-3, as these two PUFA are localized in different cell compartments. The 18:5n-3 is generally associated with chloroplastic glycolipids, while the 22:6n-3 is predominant in the other cellular compartments [51][52][53].  The 18:5n-3 was also described as an intermediate of 22:6n-3 synthesis by PKS pathway [4]. However, its low enrichment, as compared to 22:6n-3, appeared not compatible with a hypothetical production through this pathway. Nevertheless, one may speculate that there are two separated PKS pathways, one for the 22:6n-3 and one for the 18:5n-3, as these two PUFA are localized in different cell compartments. The 18:5n-3 is generally associated with chloroplastic glycolipids, while the 22:6n-3 is predominant in the other cellular compartments [51][52][53].
Considering the diversity of PKS gene in haptophytes [47], the possibility of production of 22:6n-3 directly by PKS PUFA synthesis pathway might be possible, as previously shown with thraustochytrids [10,58]. Synthesis of 22:6n-3 by PKS pathway might be at play in parallel with the n-3 pathway. Indeed, we identified a protein cluster gathering the four main domains potentially coding for the enzymes used in PKS PUFA synthesis pathways: ketoacyl reductase (KR), polyketide synthase (KS), dehydrase/dehydrogenase (DH), and enoyl reductase (ER). Protein clusters are groups of similar proteins that most likely shared the same or similar functions [59]. By considering this cluster (candidate proteins TISO_14962, TISO_14968, TISO_14968, TISO_14973, TISO_14975, and TISO_14977, Figure 8), it could be possible that these proteins act together and allow n-3 PUFA synthesis via PKS pathway. Interestingly, protein TISO_14973, while possessing only two of the four domains of interest (KS and DH), presented a specific n-3 domain. This protein might act concomitantly with the other proteins of the same cluster and allow the access to the missing reductase activities (KR and ER). Finally, the ratio 22:5n-6/22:6n-3 was below the threshold value making possible the conversion of 22:5n-6 into 22:6n-3 if we assumed that ω3-desaturase might exist in haptophyte. Synthesis of 22:6n-3 by both n-3 and n-6 pathway might be feasible in Tisochrysis lutea ( Figure 10). These different ways to produce 22:6n-3 might contribute to betaine lipids synthesis. Indeed, betaine lipids are generally highly unsaturated in C 20

Algal Culture and 13 C Labelling
This study was conducted following the experimental design described by Remize et al. (2020) [34]. The marine prymnesiophyte Tisochrysis lutea (T-iso, CCAP 927/14) was cultured in 2 L batch condition in balloons under continuous light (24 h light cycle, 100

Algal Culture and 13 C Labelling
This study was conducted following the experimental design described by Remize et al. (2020) [34]. The marine prymnesiophyte Tisochrysis lutea (T-iso, CCAP 927/14) was cultured in 2 L batch condition in balloons under continuous light (24 h light cycle, 100 µmoles photons m −2 ·s −1 ) at 20 • C and with pH regulation at 7.50 ± 0.05 by CO 2 injection. Filtered seawater was previously enriched with Conway medium [64] and inoculated with T. lutea preculture in a growing stage (exponential phase, four days old). The experimental setup was composed of two cultures (Tl1 and Tl2) receiving the labelling 13 C-CO 2 gas (Sigma-Aldrich, <3%atom 18 O, 99.0%atom 13 C) and one culture (TlT) receiving petrochemical CO 2 gas. 13 C-incorporation in Tl1 and Tl2 began after inoculation (t 0 ) and was maintained for 24 h (t 24 ).
During the first hours of the experiment, the 13 CO 2 injection tube of balloon TI1 had been temporarily disconnected from the system. Consequently, balloon Tl2 received earlier the 13 CO 2 and thus started to incorporate 13 C before balloon Tl1.

Samples Collection
Sampling was performed as described in Remize et al. (2020) [34], i.e., at 30 min, 1 h, 2 h, 3 h, 4 h, and then every 2 h. A total of 16 samples was collected during the 24 h monitoring. At each sampling time, a total volume of 30 to 70 mL was collected for (i) flow cytometry analysis of cellular parameters, (ii) bulk isotopic analysis of particulate organic carbon ( 13 C-POC) and dissolved inorganic carbon ( 13 C-DIC) by EA-IRMS, (iii) fatty acid (FA) analysis in neutral lipids (NL) and polar lipids (PL) by GC-FID, and (iv) compound specific isotope analysis (CSIA) of FA ( 13 C-FA) by GC-IRMS, as described in the following paragraphs.

Flow Cytometry Analysis
Algae growth cellular variables were measured using an Easy-Cyte Plus 6HT flow cytometer (Guava Merck Millipore ® , Darmstadt, Germany) equipped with a 488 nm blue laser, detectors of forward (FSC) and side (SSC) light scatters, and three fluorescence detectors: green (525/30 nm), yellow (583/26 nm), and red (680/30 nm). The protocol, the variables studied, and the probes used for this cytometry following are described in Remize et al. (2020) [34]. Briefly, forward scatter (FSC), side scatter (SSC), and red fluorescence (FL3, red emission filter long pass, 670 nm) are used to study, respectively, cell size, complexity, and chlorophyll content. The fluorescent probe (SYTOX, Molecular Probes, Invitrogen, Eugene OR, USA, final concentration of 0.05 µM) was used to assess cell viability on FL1 detector (green fluorescence). The concentration of bacteria was also monitored by using SYBR ® Green (Molecular Probes, Invitrogen, Eugene, OR, USA, #S7563) on FL1 detector. Concentrations of algae and bacteria were given cells per mL, and cellular variables were expressed in arbitrary units (a.u).

POC Concentration and Bulk Carbon Isotopic Composition
For particulate organic carbon (POC) and stable isotopic composition measurements, 30-70 mL of samples were filtered through pre-combusted 0.7 µm nominal pore-size glass fiber filters (Whatman GF/F, Maidstone, UK). The filter was processed, subsampled, and encapsulated as described in Remize et al. (2020) [34]. POC concentrations of all samples were measured using a CE Elantech NC2100 (ThermoScientific, Lakewood, NJ, USA) according to protocol by Remize et al. (2020) [34]. Bulk carbon isotopic composition ( 13 C-POC) was analyzed by continuous flow on an Elemental Analyzer (EA, Flash 2000; Thermo Scientific, Bremen, Germany) coupled to a Delta V+ isotope ratio mass spectrometer (Thermo Scientific). Calibration was performed with international standards and in-house standard described in Table 2. Table 2. List of international and in-house standards used for EA-IRMS and GB-IRMS analysis.

Description
Nature Analysis δ 13 C ( were measured using a CE Elantech NC2100 (ThermoScientific, Lakewood, NJ, USA) according to protocol by Remize et al. (2020) [34]. Bulk carbon isotopic composition ( 13 C-POC) was analyzed by continuous flow on an Elemental Analyzer (EA, Flash 2000; Thermo Scientific, Bremen, Germany) coupled to a Delta V+ isotope ratio mass spectrometer (Thermo Scientific). Calibration was performed with international standards and inhouse standard described in Table 2.

DIC Concentration and Bulk Carbon Isotopic Composition
Samples for dissolved inorganic carbon (DIC) concentration and stable isotopic composition were collected from the filtrate of POC samples and processed as described in Remize et al. (2020) [34]. Analyses were conducted in a gas bench coupled to a Delta Plus mass spectrometer from Thermo Fisher Scientific, Bremen, Germany (GB-IRMS).

Isotopic Data Processing
We used the atomic proportion of 13 C in percent (%atom of 13 C) to express the results instead of the δ notation due to 13 C-labelling. Conversion between δ notation and%atom 13 C notation can be done as follow [65]:

DIC Concentration and Bulk Carbon Isotopic Composition
Samples for dissolved inorganic carbon (DIC) concentration and stable isotopic composition were collected from the filtrate of POC samples and processed as described in Remize et al. (2020) [34]. Analyses were conducted in a gas bench coupled to a Delta Plus mass spectrometer from Thermo Fisher Scientific, Bremen, Germany (GB-IRMS).

Isotopic Data Processing
We used the atomic proportion of 13 C in percent (%atom of 13 C) to express the results instead of the δ notation due to 13 C-labelling. Conversion between δ notation and%atom 13 C notation can be done as follow [65]: where ( 13 C/ 12 C) PDB = 0.0112372, the ratio of 13 C to 12 C in the international reference VPDB standard. Atomic enrichment (AE) of POC and DIC is then calculated from atom% 13 C-POC correction by POC control values (i.e., corrected by 1.08%) and from atom% 13  Lipid extraction, separation of neutral and polar lipid fractions, and transesterification processes are described elsewhere [34]. Fatty acids methyl esters (FAME) samples were analyzed by gas chromatography on a Varian CP8400 gas chromatograph (Agilent, Santa Clara, CA, USA) and separated concomitantly on two columns: one polar (ZB-WAX:
To evidence FA conversion of fatty acid A into fatty acid B in T. lutea, we calculated the AE FA ratio (R) of product B over expected precursor A. R was defined with a confidence interval calculated at α = 0.1 (defined arbitrarily) as follows: where A is the fatty acid hypothesized to be a precursor to fatty acid B, and AE FA(A) and AE FA(B) are their respective atomic enrichments at each sampling time.
If the AE of the product (B) exceeds the AE of the reactant (A), ratio > 1, then it is necessary to consider another formation process for B, since any molecule formed from A would have the same AE as A or below. If the ratio is <1, transformation of A into B is considered possible. If the ratio is close to 1, the fatty acids A and B are at equilibrium in terms of label incorporated, implying B is then synthesized simultaneously or very rapidly from A.

Identification within the in Silico Proteome of T. lutea of PKS Enzymes Involved PUFA Synthesis
The in silico proteome generated from last annotated version of the genome of T. lutea was used to identify putative proteins involved in n-3 PUFA PKS pathways [66]. We used the PKS previously identified in the haptophyte Chrysochromulina tobin as query for a BLASTp analysis, using e-value <10 −3 as threshold [67]. Analysis of conserved domain was performed using the NCBI CD database V3,18 with e-value <10 −2 as threshold. The genome location of genes encoding selected proteins was identified to evaluate genes' proximity and occurrence of gene clusters. Proteins and cluster of proteins containing the four domains ketoacyl reductase (KR), dehydrase/dehydrogenase (DH), enoyl reductase (ER), and polyketide synthase (KS) were selected as candidates.

Statistical Analysis
To assess the potential effect of time and difference between balloons during algae development and of 13 CO 2 incorporation, Bartlett tests and ANOVA were performed on physiological and biochemical parameters, as well as PERMANOVA analysis on FA percentage separately in NL and PL. All statistical analyses were performed using R software.

Conclusions
The synthesis of long-chain PUFA in Tisochrysis lutea appeared to involve multiple pathways (Figures 8-10). First, the assumption of the use of PKS pathway for 22:5n-6 (DPA-6) was attested regarding the fast enrichment observed for this FA as well as the absence of detectable intermediates more or equally enriched. PKS pathway appeared to be particularly efficient in T. lutea and induced a strong incorporation of the 13 C-marker. However, the possibility of use of the conventional n-6 PUFA pathway should not be excluded, as 18:2n-6 presented a similar level of enrichment as 22:5n-6. It would only endorse that following desaturation and elongation steps to form 22:5n-6 were particularly dynamic and thus did not allow the accumulation of the 13 C-label into n-6 intermediates. Within n-3 PUFA pathway, the ∆6-desaturase route seemed slower than the n-6 pathway in T. lutea in producing the two C 18 polyunsaturated fatty acids 18:4n-3 and 18:5n-3. We assumed 18:4n-3 and 18:5n-3 were unlikely synthesis intermediates of 20:5n-3 and 22:6n-3, as their enrichments were lower than the latter. Although 22:6n-3 was present in higher proportion than 22:5n-6, it was not enriched as fast, possibly because its synthesis may be more complex. Indeed, 22:6n-3 could be synthesized by Tisochrysis lutea via a combination multiples pathway: from 22:5n-6 via ω-3 desaturase pathway, from desaturation and elongation of 20:5n-3 and 22:5n-3, and via PKS pathway. Further studies are needed to better constrain the plausible routes taken by this prymnesiophyte to produce long chain PUFA.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/md20010022/s1, Table S1: Cellular parameters of Tisochrysis lutea (morphology (FSC and SSC), viability (FL1-SYTOX), and chlorophyll content (FL3) using flow cytometry analysis (mean ± SD of the 3 balloons). Values for SYTOX are in%, values for FL3/SSC/FSC in arbitrary unit (a.u), Table S2: Concentrations in µg·L −1 of all identified and quantified FA in neutral and polar lipid fractions according to sampling time, Table S3: Concentrations in µmolC·L −1 of all identified and quantified FA in neutral and polar lipid fractions according to sampling time, Table S4: Concentrations in% of all identified and quantified FA in neutral and polar lipid fractions according to sampling time. Table S5: Mean ratio of atomic enrichment (AE) for pairs of FA (FAA vs. FAB) in the neutral lipids (NL) (mean ± SD, n = 9 sampling dates t8 to t24) for the two enriched balloons (Tl1, Tl2, Tl = Tisochrysis lutea). If the ratio is equal to or close to 1, A and B are assumed at equilibrium, and B is synthesized quickly from A; if the ratio is below 1, the transformation of B from A is possible but slow. Finally, if the ratio is above 1, A is not a main precursor of B, which has to be synthesized by a different pathway. Table S6: List of the potential candidate protein sequences involved in Tisochrysis lutea PKS synthesis pathway. The suspected function of each protein has been assumed using the NCBI conserved domain database (CDD) (Marchler-Bauer et al., 2017) by identifying the role of each domain recognized in the sequence. In columns KS/KR/DH/ER are written the number of domain corresponding to these functions in the studied sequences. ACS: acetyl-CoA synthetase, A_NRPS: adenylation domain of the non-ribosomal peptide synthetase (NRPS), Croto: crotonase/enoyl-CoA hydratase, EntF: enterobactin non-ribosomal peptide synthetase or thioesterase domain of Type I PKS, FAAL: fatty acyl-AMP ligase, GrsT: alpha/beta hydrolase, HM: hydroxymethylglutaryl-CoA synthase, MT: methyltransferase, PP: phosphopantetheine-binding (="swinging arm"), Sulf: sulfotransferase, Thio: thioesterase. Figure S1: Proportions (%) of NL vs. PL (A) and proportions (%) of fifteen fatty acids in the NL fraction in average over the 24 h (B). Figure S2: Atomic enrichment of 11 main fatty acids in the polar lipid (NL) fraction during a 24h 13 C labelling experiment. Tl: Tisochrysis lutea. Reference [68] is cited in the supplementary materials.