Study of Synthesis Pathways of the Essential Polyunsaturated Fatty Acid 20:5n-3 in the Diatom Chaetoceros Muelleri Using 13C-Isotope Labeling

The present study sought to characterize the synthesis pathways producing the essential polyunsaturated fatty acid (PUFA) 20:5n-3 (EPA). For this, the incorporation of 13C was experimentally monitored into 10 fatty acids (FA) during the growth of the diatom Chaetoceros muelleri for 24 h. Chaetoceros muelleri preferentially and quickly incorporated 13C into C18 PUFAs such as 18:2n-6 and 18:3n-6 as well as 16:0 and 16:1n-7, which were thus highly 13C-enriched. During the experiment, 20:5n-3 and 16:3n-4 were among the least-enriched fatty acids. The calculation of the enrichment percentage ratio of a fatty acid B over its suspected precursor A allowed us to suggest that the diatom produced 20:5n-3 (EPA) by a combination between the n-3 (via 18:4n-3) and n-6 (via 18:3n-6 and 20:4n-6) synthesis pathways as well as the alternative ω-3 desaturase pathway (via 20:4n-6). In addition, as FA from polar lipids were generally more enriched in 13C than FA from neutral lipids, particularly for 18:1n-9, 18:2n-6 and 18:3n-6, the existence of acyl-editing mechanisms and connectivity between polar and neutral lipid fatty acid pools were also hypothesized. Because 16:3n-4 and 20:5n-3 presented the same concentration and enrichment dynamics, a structural and metabolic link was proposed between these two PUFAs in C. muelleri.


Introduction
Interest and consumption of marine products have steadily increased in recent years partly due to the health benefits provided by their high content in n-3 polyunsaturated fatty acids (PUFA) such as 20:5n-3 (eicosapentaenoic acid, EPA) and 22:6n-3 (docosahexaenoic acid, DHA) [1][2][3]. n-3 PUFAs are essential to all organisms including humans, but due to a lack of or insufficient de novo synthesis capabilities in most metazoans, n-3 PUFAs have to be obtained from the diet [4]. In the environment, primary producers (phototroph and heterotroph protists) are recognized for being the main suppliers of PUFAs to trophic webs [5][6][7], while on-going climate change may affect PUFAs availability and transfer across marine food webs [8][9][10][11][12][13]. Moreover, due to the high demand for human nutrition and aquaculture of carnivore species, n-3 PUFAs may become scarce because of overfishing and marine fish stock reduction [4]. Understanding how these protists synthetize EPA and DHA appears to be of a central interest prior to evaluating the influence of global changes in their availability and transfer to higher trophic levels up to human diets. Despite the economical and ecological significance of n-3 PUFAs, biological and biochemical processes responsible for the synthesis of these compounds by phytoplankton and microzooplankton are still to be fully understood.
Production of PUFAs by marine unicellular eukaryotes occurs via different metabolic pathways [14][15][16][17][18]. The most known conventional pathway is O2-dependent and starts with the fatty acid synthase pathway (FAS pathway), allowing fatty acid synthesis from acetyl-CoA and malonyl-CoA. It is followed by elongation and desaturation processes of the n-3 and n-6 pathways to give longchain polyunsaturated fatty acids. These two pathways can be connected by the alternative ω-3 desaturase pathway [19]. Another route using Δ8 desaturase and bypassing the Δ6 desaturation of the n-3 and n-6 pathways has also been identified in Haptophytes microalgae such as Isochrysis galbana [20], Pavlova salina [21] or the coccolithophore Emiliania huxleyi [22] (Figure 1). The polyketide synthase pathway (PKS pathway) is an O2-independent pathway that can be found in heterotroph protists such as thraustochytrids [23,24] producing 20:5n-3 and 22:6n-3 [18,25,26]. Desaturases are noted with "ω" and/or "Δ" and refer to the location of carbon holding the newly formed double bond and its position within the methyl or front end of the acyl chain (2 types, methyl-end (grey or blue arrows) or front-end desaturases (yellow arrows)). Des: desaturase, Elo: elongase, FAS: fatty acid synthase.
In higher plants as in diatoms, after synthesis by FAS pathway in the chloroplast, saturated fatty acids acyl groups have two possible destinations. The prokaryotic pathway, they can be retained in the chloroplast to form plastid glycerolipids (GL) from glycerol-3-phosphate (G3P) and serve for cell Figure 1. Fatty acids synthesis in diatoms. Desaturases are noted with "ω" and/or "∆" and refer to the location of carbon holding the newly formed double bond and its position within the methyl or front end of the acyl chain (2 types, methyl-end (grey or blue arrows) or front-end desaturases (yellow arrows)). Des: desaturase, Elo: elongase, FAS: fatty acid synthase.
In higher plants as in diatoms, after synthesis by FAS pathway in the chloroplast, saturated fatty acids acyl groups have two possible destinations. The prokaryotic pathway, they can be retained in the chloroplast to form plastid glycerolipids (GL) from glycerol-3-phosphate (G3P) and serve for cell membrane structuration [31][32][33]. In the eukaryotic pathway, saturated fatty acids are transferred to the endoplasmic reticulum (ER) and acylated onto a glycerophosphate backbone to form glycerolipids and to take part in synthesis of other cell membrane lipids (such as phospholipids) or energy reserves (triacylglycerols (TAG)) [32]. According to these pathways (prokaryotic vs. eukaryotic), fatty acid composition of glycerolipids differs: GL synthesized in the chloroplast present a C 16 acyl group at the position sn-2, while those produced in the ER present C 18 or longer acyl group (C 20 , C 22 ) at the sn-2 position [17]. In diatoms, TAG synthesis occurs in the ER from DAG [17,32]. TAG are synthesized in diatoms by the Kennedy pathway, which is acyl-CoA dependent and consists in stepwise acylations of G3P to give first lysophosphatidic acid, then phosphatidic acid, which is transformed in DAG and then in TAG by the action of acyl-CoA diacylglycerol acyltransferase (DGAT) [17,32]. Another acyl-CoA-independent pathway has been hypothesized in diatoms following the identification of enzymes and genes coding for the enzymes involved in this pathway in plants [17,[34][35][36]. It would rely on the use of the phospholipid diacylglycerol acyltransferase (PDAT) to synthesize DAG from phosphatidylcholine (PC), which plays the role of an acyl donor. In plants, this pathway is supported by the PC acyl-editing machinery of the Lands cycle [32,33,37,38].
Recent studies dedicated to PUFA production have focused on studying gene expression for metabolic engineering to enhance DHA and EPA synthesis and production in microalgae or other producers initially unable to synthesize these compounds [35,39,40]. Other studies have centered on finding new genes or new enzymes implied in these FA synthesis pathways. Mülhroth et al. [35] showed that 106 genes encoding enzymes might be involved in fatty acids synthesis in Phaeodactylum tricornutum. However, our understanding is still incomplete especially regarding the routes followed and the production of these fatty acids of interest. Then, to improve our knowledge of PUFA production by microalgae, it is essential to better characterize quantitatively their synthesis pathways. For this purpose, 13 C-labeling Flux Analysis offers an interesting approach that has been applied to Escherichia coli, yeast or microalgae [41][42][43]. The principle of stable isotope labeling relies on the monitoring of an isotopically labeled substrate incorporated into targeted organic macromolecules. Different metabolic intermediates or end products like fatty acids can be identified, and their level of labeling can be quantified, giving insights into biosynthesis pathways. Recently and thanks to instruments like Gas Chromatography coupled to Mass Spectrometry (GC-MS) or Gas Chromatography coupled to Isotopic Ratio Mass Spectrometry (GC-c-IRMS) allowing compound-specific isotope analysis (CSIA), substantial advances have been made, and it is now possible to resolve with high precision the isotopic composition of organic macromolecules including fatty acids [44][45][46][47].
In this study, a stable isotope ( 13 C) labeling experiment was conducted with a monospecific culture of the diatom Chaetoceros muelleri in order to study the synthesis pathways of n-3 PUFA. This species has been chosen as the experimental model because diatoms are responsible for around 40% of the global primary production [48] and are characterized by a high 20:5n-3 content. The incorporation of the isotopic label ( 13 C enriched CO 2 ) was monitored in 10 FA, including 20:5n-3 and its potential precursors during 24 h and at a high temporal resolution (each 0.5 to 2 h). Progressive incorporation of the 13 C into FA (from precursors to PUFA) and quantification of main fatty acid production allowed us to constrain FAS, C 16 PUFAs pathways and n-3 and/or n-6 elongase/desaturase pathways and their involvement in EPA production by C. muelleri. To monitor its growth and physiological status, several cellular parameters (morphology, viability, esterase activity and lipid content) were also concomitantly measured by flow cytometry analysis (FCM).

Algal Culture and Isotopic Labeling
Monospecific cultures were performed with the marine diatom C. muelleri (strain Culture Collection of Algae and Protozoa (CCAP) 1010-3 obtained from the CCAP culture collection of the Scottish Marine Institute, Oban, Scotland, UK, formerly listed as Chaetoceros neogracile VanLandingham 1968) in batch (total culture volume of 2 L) and under continuous light. Cultures were prepared with 1750 mL of sterile filtered seawater, 250 mL of algal inoculum and 2 mL of nutrients medium (Conway medium) and kept sterile during the whole experiment. Three culture balloons were prepared: two replicates for 13 C labeling (balloon 1, Cm1 and balloon 2, Cm2) and one as a control (CmT) without 13 C enrichment and fed with petrochemical CO 2 (whose isotopic composition is equal to −37% ).
The three culture balloons (two labeled and one control) were subjected to a pre-culturing phase of 4 days before the start of the monitoring in order to sample the alga during the exponential growth phase and to promote label assimilation into FA. The isotopic labeling was performed using pure 13 C-CO 2 gas (Sigma-Aldrich, < 3%atom 18O, 99.0%atom 13 C). The control balloon was continuously bubbled with unlabelled CO 2 (i.e., petrochemical CO 2 ). For each of the three replicates, the CO 2 introduction in the culture was started just before the first sampling time (t 0 ) and maintained during 24 h (t 24 ). The growth of the two labeled replicates was controlled by 13 CO 2 addition using a pH-stat system, which fed the culture when pH was higher than 7.50 ± 0.05 ( Figure 2). The whole experimental system was designed in order to keep the cells in a good physiological state (controlled by flow cytometry analyses) during the experiment.
Biomolecules 2020, 10, x FOR PEER REVIEW  4 of 27 were prepared with 1750 mL of sterile filtered seawater, 250 mL of algal inoculum and 2 mL of nutrients medium (Conway medium) and kept sterile during the whole experiment. Three culture balloons were prepared: two replicates for 13 C labeling (balloon 1, Cm1 and balloon 2, Cm2) and one as a control (CmT) without 13 C enrichment and fed with petrochemical CO2 (whose isotopic composition is equal to −37‰). The three culture balloons (two labeled and one control) were subjected to a pre-culturing phase of 4 days before the start of the monitoring in order to sample the alga during the exponential growth phase and to promote label assimilation into FA. The isotopic labeling was performed using pure 13 C-CO2 gas (Sigma-Aldrich, < 3%atom 18O, 99.0%atom 13 C). The control balloon was continuously bubbled with unlabelled CO2 (i.e., petrochemical CO2). For each of the three replicates, the CO2 introduction in the culture was started just before the first sampling time (t0) and maintained during 24 h (t24). The growth of the two labeled replicates was controlled by 13 CO2 addition using a pH-stat system, which fed the culture when pH was higher than 7.50 ± 0.05 ( Figure 2). The whole experimental system was designed in order to keep the cells in a good physiological state (controlled by flow cytometry analyses) during the experiment.

Figure 2.
Experimental design of the enrichment experiment. The 13 CO2 is supplied to the culture using a pH-stat system. To prevent contamination when sampling the algae, pliers are used to close/open the tubes/ways needed to first put the balloon under pressure and then allow sampling and finally rinse the tubes after collection.

Samples Collection
Sampling was performed during 24 h at 30 min, 1 h, 2 h, 3 h, 4 h and then every 2 h. A total of 16 samples were thus collected during the 24 h monitoring. The sampling system ( Figure 2) was designed to collect the culture without opening the balloon and to avoid bacterial contamination and the introduction of atmospheric CO2. At each sampling time, a total volume of 52 mL per balloon was collected: (i) 2 mL for flow cytometry (FCM) analysis of cellular parameters, (ii) 25 mL for isotopic analysis of Particulate Organic Carbon ( 13 C-POC) and Dissolved Inorganic Carbon ( 13 C-DIC) by Elemental-Analyzer Isotope Ratio Mass Spectrometry (EA-IRMS), (iii) 25 mL for fatty acid analysis in neutral lipids (NL) and polar lipids (PL) by Gas-Chromatography Flame Ionisation Detection (GC-FID) and (iv) compound-specific isotope analysis (CSIA) of FA ( 13 C-FA) 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 . Experimental design of the enrichment experiment. The 13 CO 2 is supplied to the culture using a pH-stat system. To prevent contamination when sampling the algae, pliers are used to close/open the tubes/ways needed to first put the balloon under pressure and then allow sampling and finally rinse the tubes after collection.

Samples Collection
Sampling was performed during 24 h at 30 min, 1 h, 2 h, 3 h, 4 h and then every 2 h. A total of 16 samples were thus collected during the 24 h monitoring. The sampling system ( Figure 2) was designed to collect the culture without opening the balloon and to avoid bacterial contamination and the introduction of atmospheric CO 2 . At each sampling time, a total volume of 52 mL per balloon was collected: (i) 2 mL for flow cytometry (FCM) analysis of cellular parameters, (ii) 25 mL for isotopic analysis of Particulate Organic Carbon ( 13 C-POC) and Dissolved Inorganic Carbon ( 13 C-DIC) by Elemental-Analyzer Isotope Ratio Mass Spectrometry (EA-IRMS), (iii) 25 mL for fatty acid analysis in neutral lipids (NL) and polar lipids (PL) by Gas-Chromatography Flame Ionisation Detection (GC-FID) and (iv) compound-specific isotope analysis (CSIA) of FA ( 13 C-FA) 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 scatter, and three fluorescence detectors: green (525/30 nm), yellow (583/26 nm) and red (680/30 nm). Cell variables i.e., forward scatter (forward scatter, FSC), side scatter (side scatter, SSC) and red fluorescence (FL3, red emission filter long pass, 670 nm) were used to identify and select the C. muelleri cell population. FSC and SSC give, respectively, information on the relative size and complexity of cells [49][50][51]. The flow cytometry measurements were performed on fresh (living) samples.
Three fluorescent probes were used to assess viability, esterase activity, and lipid content according to Seoane et al. [52]. Briefly, the SYTOX (Molecular Probes, Invitrogen, Eugene OR, USA, final concentration of 0.05 µM), a probe that binds to the DNA of a permeable or permeabilized cell, was used to estimate the percentage of dead cells in culture sample [53]. The fluorescein diacetate (FDA, Molecular Probes, Invitrogen, Eugene OR, USA #F1303 at a final concentration of 6 µM, emission wavelength 525 nm), a substrate for intracellular esterases, was used to provide a proxy of primary metabolism. As cell membrane integrity is essential for the retention of FDA product (fluorescein) by the cell, it is also used to estimate the percentage of metabolically active cells [52]. Cells with a high level of green fluorescence (FL1) correspond to high esterase activity, while cells with a low FL1 fluorescence are considered cells with reduced esterase activity (reduced metabolism) or dead/dying cells. The BODIPY probe (BODIPY 505/515 FL; Molecular Probes, Invitrogen, Eugene OR, USA, final concentration of 10 mM), which stains lipid droplets within microalgae cells, was used as a proxy of lipid reserves [52]. The green fluorescence emitted is proportional to the quantity of lipid reserve present in the cells.
The concentration of bacteria was also monitored during the experiment according to Seoane et al. [52] by using SYBR ® Green (Molecular Probes, Invitrogen, Eugene OR, USA, #S7563), a DNA staining fluorescence probe which allows detection of DNA stained bacteria on FL1 detector (green fluorescence). Results are given as concentration of bacteria cells per mL.

POC Concentration and Bulk Carbon Isotopic Composition
Samples (25 mL) for POC concentration and stable isotopic composition were filtered through pre-combusted 0.7 µm nominal pore-size glass fiber filters (Whatman GF/F, Maidstone, UK). The filter was then dried at T = 55 • C, fumed with hydrochloric acid to remove particulate inorganic carbon, subsampled with a 13 mm punch and encapsulated into tin caps for further analysis ( [54,55] and references therein).
POC concentrations of all samples were measured using a CE Elantech NC2100 (Thermo Scientific, Lakewood, NJ, USA) according to the protocol of the United States Environmental Protection Agency [56] (with acetanilide (99.9% purity, C 8 H 19 NO CASRN 103-84-4) as a standard. Briefly, the encapsulated samples were combusted at 980 • C in the elemental analyzer. The combustion products were passed over a copper tube with chromium oxide/cobaltous acetate oxide as catalysts to aid the conversion of carbon into carbon dioxide. The mixture was released to thermal conductivity detectors to measure the levels of carbon in a sample ( [55] and references therein). The measured values are corrected with those of known standards. POC concentration is given in mmol·L −1 .
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, Bremen, Germany). Calibration was performed with certified international standards and in-house standards described in Table 1.

DIC Concentrations and Bulk Carbon Isotopic Composition
Samples for Dissolved Inorganic Carbon concentration and stable isotope composition were collected from the filtrate of POC samples (25 mL). Twelve milliliters was put into Labco Exetainer (Labco, Wales, UK) tubes, poisoned with 20 µL of mercury chloride (HgCl 2 ) and stored at 4 • C until Gas-Bench Isotope Ratio Mass Spectrometer (GB-IRMS) analysis. The sample preparation for DIC concentration and 13 C-DIC measurements was made according to Assayag et al. [57]. Briefly, 1 mL subsample was added to the mixed tube, then 23 droplets of phosphoric acid (H 3 PO 4 ) were introduced into a closed Exetainer tube (Labco, Wales, UK) that is flushed with ultra-pure helium gas to eliminate the residual air and avoid contamination. The H 3 PO 4 addition converted all DIC species into CO 2 [57], and after 15 h, equilibration at room temperature CO 2 was measured in the headspace of a Gas Bench coupled to a Delta Plus mass spectrometer from Thermo Scientific, Bremen, Germany (GB-IRMS).

Isotopic Data Processing
Because we consider here 13 C-enriched samples, we used the atomic abundance of 13 C in percentage (%atom of 13 C) to express the results. Conversion between delta notation and %atom 13 C notation can be done as follows [58]: With ( 13 C/ 12 C) PDB = 0.0112372, the ratio of 13 C to 12 C in the international reference Vienna-Pee Dee Belemnite (V-PDB) standard.
Atomic enrichment (AE) is then calculated from %atom 13 C POC-corrected by POC control values (1.08% close of the natural values observed in the marine environment) and atom % 13 C DIC-corrected by control abundance values (DIC control = 1.12%) according to the following equations:

Lipid Extraction
After filtration on pre-combusted 47 mm GF/F filters (porosity 0.7 µm) of 25 mL of culture, lipases were deactivated by addition of boiling water, and lipids were extracted by diving the filter into 6 mL solvent mixture (chloroform:methanol, 2/1, v/v). Lipid extracts were flushed with nitrogen and stored at −20 • C until analysis.

Separation of Neutral and Polar Lipids
Total lipid extracts were separated into neutral and polar fractions following the method of Le Grand et al. [59]. In brief, 1 mL of extract was evaporated to dryness under nitrogen, recovered with 3 washes of 0.5 mL of chloroform:methanol (98:2 v:v; final volume 1.5 mL) and spotted at the top of a silica gel column (40 mm × 4 mm, silica gel 60A 63-200 µm rehydrated with 6% H 2 O, 70-230 mesh, Sigma-Aldrich, Darmstadt, Germany). Neutral lipids fraction (NL) was eluted using chloroform:methanol (98:2 v:v; 10 mL) and polar lipid fraction (PL) with methanol (20 mL). Both were collected in glass vials containing an internal standard (C23:0, 2.3 µg). Lipid fractions were then stored at −20 • C until further analysis.

Transesterification of FAME
Fatty acids methyl esters (FAME) transesterification was conducted according to the protocol described by Mathieu-Resuge et al. [60]. In brief, after evaporation to dryness of the neutral and polar lipid fractions, transesterification was made by adding 0.8 mL of H 2 SO 4 /methanol mixture (3.4%, v:v) to the lipid extract and heated at 100 • C for 10 min. Hexane (0.8 mL) and distilled water saturated with hexane (1.5 mL) were added. The lower MeOH-water phase is discarded after homogenization and centrifugation. Hexane fraction containing FAME was washed two more times with another 1.5 mL of distilled water.

Fatty Acid Analysis by Gas Chromatography Flame Ionisation Detector (GC-FID)
Analysis of FAME was performed on a Varian CP8400 gas chromatograph (Agilent, Santa Clara CA, USA) using simultaneously two parallel columns: a polar (DB-WAX: 30 mm × 0.25 mm ID × 0.25 µm, Agilent, Santa Clara CA, USA and apolar column (DB-5: 30 m × 0.25 mm ID × 0.25 µm, Agilent, Santa Clara CA, USA). The temperature program used by the gas chromatograph was the following: first, an initial heating to 0 from 150 • C at 50 • C·min −1 , then to 170 • C at 3.5 • C·min −1 , to 185 • C at 1.5 • C·min −1 , to 225 at 2.4 • C·min −1 and finally to 250 • C at 5.5 • C·min −1 and maintained for 15 min. The FAME were identified by comparison of their retention time with commercial and in-house standards mixtures as shown by the two columns used in Supplementary Data.
Fatty acid concentrations are expressed in µmolC·L −1 . Fatty acids are grouped according to their hypothesized synthesis pathways as described in Table 2.
The GC-c-IRMS measured values were calibrated using the F8-3 standard mixture of eight fatty acid ethyl and methyl esters (14:0, 16:0, 18:0 and 20:0 with values ranging from −26.98 ± 0.02% to −30.38 ± 0.02% ) supplied by Indiana University Stable Isotope Reference Materials (Schimmelman, Indiana University IN, USA) as described in Mathieu-Resuge et al. [60]. 2.7.6. Compound-Specific Isotope Data Processing FA atomic enrichment (AE FA ) is calculated with the same method as for AE POC and AE DIC , with %atom 13 C for each FA given by GC-c-IRMS analysis (AE FAcontrol = AE FAnat = 1.08%). AE FA is then used for estimating the specific uptake rate in FA (µ FA in h −1 ) with the following equation: µ FA is calculated between t 6 (once the DIC 13 C enrichment started) and t 24 (end of the enrichment experiment). AE FA corresponds to cell enrichment at t 24 (AE FA at t 6 is considered as negligible). The AE DIC is calculated using the time-weighted average of AE DIC over the same time period.
Finally, to evidence FA synthesis conversion from fatty acid A and fatty acid B in C. muelleri, we calculated the AE FA ratio (R) of product B over supposed precursor A. If the calculated ratio is close to 1 with a confidence interval at α = 0.1, the fatty acids A and B are supposed at equilibrium and synthesized simultaneously or very closely. If the ratio is below 1, the transformation of A into B is possible but slower. On the contrary, if the ratio is above 1, A is not considered as a precursor of B. R is defined as follows.
with A as the fatty acid that is supposed to be a precursor of fatty acid B and AE FA(A) and AE FA(B) their respective atomic enrichments at each sampling time.

Statistical Analysis
To assess the potential effect of time and difference between balloons during algae development and 13 CO 2 incorporation, PERMANOVA analysis was used on the FA mass percentage separately in NL and PL. Principal component analysis (PCA) coupled with similarity percentage analysis (SIMPER) allowed us to distinguish fatty acids that are the main ones responsible for the overall observed variability (80%). A Spearman test was also conducted on fatty acids abundance in both PL and NL to explore the relationship between fatty acids. All statistical analyses were conducted using R software.

Algae Physiology During Growth
During the 24 h of the experiment, the replicability between balloons (enriched (Cm1 and Cm2) vs. control (CmT)) was satisfying with similar increasing trends despite slightly lower concentrations of the control ballon. C. muelleri grew on average from 8.3 × 10 6 cells·mL −1 at t 0 to 11.3 × 10 6 cells·mL −1 at t 24 , corresponding to an increase factor of 1.4 ( Figure 3). By considering the time point 0.5 (average of 6.6 × 10 6 cells·mL −1 ), this growth was closer to a doubling of cell density (factor 1.7). After t 16 , a sharper increase was observed for the three balloons with a change in general slope (9.2 × 10 5 cells·mL −1 ·h −1 to 19.1 × 10 5 cells·mL −1 ·h −1 ). Bacteria concentrations increased from 1.1 × 10 8 cells·mL −1 (t 0 ) on average to 1.4 × 10 8 cells·mL −1 t 24 . The strongest increase in bacteria concentrations was observed between t 0 and t 14  During the experiment, cell size (FSC), cell complexity (SSC) remained relatively constant and similar between control and enriched balloon (respectively 153.3 ± 1.7 a.u (Student test, p-value < 0.001 and ANOVA p.value = 0.3) and 52.8 ± 2.5 a.u (Student test, p-value < 0.001 and ANOVA p-value = 0.07), on average for the three balloons) ( Table 3). Chlorophyll content (FL3) decreased slightly between t0.5 and t24 from 274 a.u to 257 a.u. Percentage of dead cells as measured by SYTOX DNA staining (Green fluorescence, FL1) remained below 4% and averaged around 2%, while the percentage of metabolically active cells averaged at 90% with the lowest value at 86%. It revealed that C. muelleri cells were in good health for the whole experiment. The neutral lipid content as estimated by BODIPY staining, (green fluorescence, FL1) was more variable. It averaged at 225.3 a.u, but time points t0, t6, t10, t12, and t18 were lower (166.0 ± 16.0 a.u), with the lowest value at t0 (154.1 a.u) and the highest at t20 (308.5 a.u) ( Table 3).  During the experiment, cell size (FSC), cell complexity (SSC) remained relatively constant and similar between control and enriched balloon (respectively 153.3 ± 1.7 a.u (Student test, p-value < 0.001 and ANOVA p.value = 0.3) and 52.8 ± 2.5 a.u (Student test, p-value < 0.001 and ANOVA p-value = 0.07), on average for the three balloons) ( Table 3). Chlorophyll content (FL3) decreased slightly between t0.5 and t24 from 274 a.u to 257 a.u. Percentage of dead cells as measured by SYTOX DNA staining (Green fluorescence, FL1) remained below 4% and averaged around 2%, while the percentage of metabolically active cells averaged at 90% with the lowest value at 86%. It revealed that C. muelleri cells were in good health for the whole experiment. The neutral lipid content as estimated by BODIPY staining, (green fluorescence, FL1) was more variable. It averaged at 225.3 a.u, but time points t 0 , t 6 , t 10 , t 12 , and t 18 were lower (166.0 ± 16.0 a.u), with the lowest value at t 0 (154.1 a.u) and the highest at t 20 (308.5 a.u) ( Table 3).

Bulk POC and DIC and their 13 C-Labeling
Upon addition of 13 C-CO2 to the culture at t0, the atomic enrichment of the DIC stayed close to 0 till t4. After t4 and until t24, AEDIC increased up to 60%, with a plateau after t20 ( Figure 5). The DIC 13 C level was similar between balloons ( Figure 5). As for DIC, AE of POC remained close to zero until t4 and then sharply increased in both balloons ( Figure 5). As both AEPOC and AEDIC increased at t4-t6, the 13 C enrichment appeared thus to be almost simultaneous between DIC and POC pools. The level of enrichment of POC was higher in Cm1 than in Cm2 but remained parallel during the course of the experiment. . The blue line corresponds to increasing trend between t 0 and t 24 (POC: y = 0.27x + 14, R 2 = 76% p-value < 0.001; TFA: y = 0.08x + 2.5, R 2 = 93% p-value < 0.001). POC concentration relationships with algae concentration (c) (eq: y = 2.4x − 5.7 R 2 = 79%, p-value < 0.001) and with total fatty acid concentration (d) (eq: y = 0.23x − 0.5, R 2 = 80%, p-value < 0.001) (mean; n = 3). Points below 8 × 10 6 cells·mL −1 are not included in the linear regression and are considered as outliers (corresponding to t 0.5 and t 1 in Figure 3).

Bulk POC and DIC and their 13 C-Labeling
Upon addition of 13 C-CO 2 to the culture at t 0 , the atomic enrichment of the DIC stayed close to 0 till t 4 . After t 4 and until t 24 , AE DIC increased up to 60%, with a plateau after t 20 ( Figure 5). The DIC 13 C level was similar between balloons ( Figure 5). As for DIC, AE of POC remained close to zero until t 4 and then sharply increased in both balloons ( Figure 5). As both AE POC and AE DIC increased at t 4 -t 6, the 13 C enrichment appeared thus to be almost simultaneous between DIC and POC pools. The level of enrichment of POC was higher in Cm1 than in Cm2 but remained parallel during the course of the experiment.

Fatty Acid Composition of Neutral and Polar Lipids in C. Muelleri
Saturated fatty acids (SFA) and monounsaturated fatty acids (MUFA) accounted for, respectively, 41% and 36% of neutral lipid fatty acids (on average for the three balloons), while PUFA represented 22% (Figure 6). Polyunsaturated fatty acids (PUFA) were more abundant in polar lipids (40% of total polar lipid fatty acids on average), while SFA and MUFA were at 33% and 26% of total polar lipid fatty acids, respectively. For both polar and neutral lipids, the main FA were 14:0, 16:0, 16:1n-7 and 20:5n-3 ( Figure 6).    Table S1 of the Supplementary Data gives the concentration in µg·L −1 and µmolC·L −1 for all fatty acids identified in this study.

Variability of Fatty Acid Proportions during the Experiment
The PERMANOVA analysis on NL and PL fatty acids revealed a significant difference between sampling times for fatty acid mass percentage (p < 0.001) but no significant difference between the three balloons (p > 0.05).
The PCA of neutral and fatty acids was coupled with SIMPER and revealed a clear timedependent division for both NL and PL (Figure 7). The first sampling points were on the positive side, and the last sampling points were located on the negative side of the axis 1. In PL, axis 1 was driven by fatty acids 20:5n-3, 16:2n-7 and 16:3n-4 and in NL by 18:3n-6, 16:0, 16:1n-7 and 15:0. Following these multivariate analysis, 16:0 and 16:1n-7 were found to be correlated (Spearman test: α = 0.

Variability of Fatty Acid Proportions during the Experiment
The PERMANOVA analysis on NL and PL fatty acids revealed a significant difference between sampling times for fatty acid mass percentage (p < 0.001) but no significant difference between the three balloons (p > 0.05).
The PCA of neutral and fatty acids was coupled with SIMPER and revealed a clear time-dependent division for both NL and PL (Figure 7). The first sampling points were on the positive side, and the last sampling points were located on the negative side of the axis 1. In PL, axis 1 was driven by fatty acids 20:5n-3, 16:2n-7 and 16:3n-4 and in NL by 18:3n-6, 16:0, 16:1n-7 and 15:0. Following these multivariate analysis, 16:0 and 16:1n-7 were found to be correlated (Spearman test: α = 0.

Fatty Acid 13 C-Enrichment and Synthesis
The atomic enrichments (AE) of the 10 studied fatty acids were gathered according to hypothesized pathways (C16 PUFAs, n-3 PUFAs, and n-6 PUFAs pathways and post-FAS pathway) (Figure 9).
For the C16 pathway and connections between n-3 and n-6 pathways, all the ratios are below 1 for NL and PL (Table 4). It was also the case for 20:5n-3/18:4n-3 in PL and NL in the n-3 pathway, 20:4n-6/18:3n-6 in PL and NL and 18:3n-6/18:2n-6 in PL for the n-6 pathway. However in NL, 18:3n-6/18:2n-6 and 18:2n-6/18:1n-9 are above 1 with respective mean values at 1.25 and 2.14. Finally 18:2n-6/18:1n-9 is close to 1 for PL. The 18:0 can allow the slow production of 18:1n-9 in NL (ratios below 1) but 18:1n-9/18:0 ratio is above 1 in PL (Table 4). Figure 11 summarised the interpretation of the results obtained with ratio calculations. Table 4. Mean ratio of atomic enrichment (AE) for pairs of FA (FAA vs. FAB) in neutral lipids (NL) and polar lipids (PL) (Mean ± SD, n = 10 sampling dates) for the two enriched balloons (Cm1, Cm2; Cm = Chaetoceros muelleri). 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 the main precursor of B, which has to be synthesized by a different pathway.  Mean ratio of atomic enrichment (AE) for pairs of FA were calculated for neutral lipids (NL) and polar lipids (PL) to explore the precursor-product relationship between FA ( Figure 10).

Neutral lipids Polar lipids
For the C 16 pathway and connections between n-3 and n-6 pathways, all the ratios are below 1 for NL and PL (Table 4). It was also the case for 20:5n-3/18:4n-3 in PL and NL in the n-3 pathway, 20:4n-6/18:3n-6 in PL and NL and 18:3n-6/18:2n-6 in PL for the n-6 pathway. However in NL, 18:3n-6/18:2n-6 and 18:2n-6/18:1n-9 are above 1 with respective mean values at 1.25 and 2.14. Finally 18:2n-6/18:1n-9 is close to 1 for PL. The 18:0 can allow the slow production of 18:1n-9 in NL (ratios below 1) but 18:1n-9/18:0 ratio is above 1 in PL (Table 4). Figure 11 summarised the interpretation of the results obtained with ratio calculations. Table 4. Mean ratio of atomic enrichment (AE) for pairs of FA (FA A vs. FA B ) in neutral lipids (NL) and polar lipids (PL) (Mean ± SD, n = 10 sampling dates) for the two enriched balloons (Cm1, Cm2; Cm = Chaetoceros muelleri). 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 the main precursor of B, which has to be synthesized by a different pathway.  Figure 11. Summary of FA synthesis relationships in polar and neutral lipids pools for C. muelleri according to ratio calculation results. The solid arrows correspond to ratios of fatty acid B and suspected precursor A that are equal to or below 1, and red crosses correspond to ratios above 1. The fatty acids in grey were absent or present in trace amounts in this experiment, and consequently, the ways marked by dashed arrows were not possible to verify.

Growth and 13 C Incorporation
This study aimed at investigating the synthesis pathways of 20:5n-3 in C. muelleri using 13 CO2 labeling. The addition of 13 CO2 did not disturb the algae physiology: growth, cell size, cell complexity, chlorophyll content. Cell viability remained high throughout the experiment as revealed by constant proportions of alive (SYTOX) and active (FDA) cells. This supported the good physiological state of the algae, which was able to produce FA (around 23% of POC). Algae growth was associated with an increase of POC and TFA concentrations. The 13 C enrichments into DIC and POC were significantly detected after t4. After a lag phase, both DIC and POC pools were rapidly enriched, with slightly different kinetics. DIC enrichment increased rapidly at first and then stabilized after 14-16 h, while POC enrichment increased steadily following a linear slope. The difference of POC enrichment in 13 C between the two labeled balloons remained unclear, especially since the respective levels of 13 C found in the studied FA for both replicates stayed close. The incorporation of the 13 C into fatty acids was detectable after 4 h, but their levels of enrichment varied according to fatty acids. 13 C incorporation in some FA (16:0, 18:1n-9) was even higher than in POC.

The C16 PUFAs Pathway in C. muelleri
C. muelleri showed similar fatty acids profile as other diatoms: 14:0, 16:0, 16:1n-7 and 20:5n-3 (EPA) were the major fatty acids for polar and neutral lipids with PUFA higher in polar lipids and SFA and MUFA predominant in neutral lipids [61][62][63][64][65]. The strong concentration of 16:1n-7 and its high enrichment supported its central role as a precursor of the C16 PUFAs pathway. The C16 PUFAs pathway is initiated by 16:1n-7 which is first desaturated in 16:2n-7 or 16:2n-4 [66,67]. Then, 16:2n-4 is desaturated again in 16:3n-4 and finally in 16:4n-1. In our experiment, 16:4n-1 was only present in trace amounts, 16:2n-7 and 16:2n-4 were in low concentrations, and 16:3n-4 was the most concentrated. We assumed that the C16 PUFAs synthesis pathway led to the accumulation of 16:3n-4 at the expense of its precursors. The 16:1n-7 being much more enriched than 16:3n-4 confirmed that 16:1n-7 can be its precursor. The low concentration of 16:2n-7 and 16:2n-4 suggested that these intermediates in the C16 PUFAs pathway were rapidly converted toward 16:3n-4. Furthermore, as 13 C Figure 11. Summary of FA synthesis relationships in polar and neutral lipids pools for C. muelleri according to ratio calculation results. The solid arrows correspond to ratios of fatty acid B and suspected precursor A that are equal to or below 1, and red crosses correspond to ratios above 1. The fatty acids in grey were absent or present in trace amounts in this experiment, and consequently, the ways marked by dashed arrows were not possible to verify.

Growth and 13 C Incorporation
This study aimed at investigating the synthesis pathways of 20:5n-3 in C. muelleri using 13 CO 2 labeling. The addition of 13 CO 2 did not disturb the algae physiology: growth, cell size, cell complexity, chlorophyll content. Cell viability remained high throughout the experiment as revealed by constant proportions of alive (SYTOX) and active (FDA) cells. This supported the good physiological state of the algae, which was able to produce FA (around 23% of POC). Algae growth was associated with an increase of POC and TFA concentrations. The 13 C enrichments into DIC and POC were significantly detected after t 4 . After a lag phase, both DIC and POC pools were rapidly enriched, with slightly different kinetics. DIC enrichment increased rapidly at first and then stabilized after 14-16 h, while POC enrichment increased steadily following a linear slope. The difference of POC enrichment in 13 C between the two labeled balloons remained unclear, especially since the respective levels of 13 C found in the studied FA for both replicates stayed close. The incorporation of the 13 C into fatty acids was detectable after 4 h, but their levels of enrichment varied according to fatty acids. 13 C incorporation in some FA (16:0, 18:1n-9) was even higher than in POC. 16 PUFAs Pathway in C. muelleri C. muelleri showed similar fatty acids profile as other diatoms: 14:0, 16:0, 16:1n-7 and 20:5n-3 (EPA) were the major fatty acids for polar and neutral lipids with PUFA higher in polar lipids and SFA and MUFA predominant in neutral lipids [61][62][63][64][65]. The strong concentration of 16:1n-7 and its high enrichment supported its central role as a precursor of the C 16 PUFAs pathway. The C 16 PUFAs pathway is initiated by 16:1n-7 which is first desaturated in 16:2n-7 or 16:2n-4 [66,67]. Then, 16:2n-4 is desaturated again in 16:3n-4 and finally in 16:4n-1. In our experiment, 16:4n-1 was only present in trace amounts, 16:2n-7 and 16:2n-4 were in low concentrations, and 16:3n-4 was the most concentrated. We assumed that the C 16 PUFAs synthesis pathway led to the accumulation of 16:3n-4 at the expense of its precursors. The 16:1n-7 being much more enriched than 16:3n-4 confirmed that 16:1n-7 can be its precursor. The low concentration of 16:2n-7 and 16:2n-4 suggested that these intermediates in the C 16 PUFAs pathway were rapidly converted toward 16:3n-4. Furthermore, as 13 C accumulation or incorporation could not be detected within these two synthesis intermediates, we can hypothesize that the desaturation of 16:1n-7 (16:1∆9) to 16:2n-7 (16:2∆6, 9) or to 16:2n-4 (16:2∆9, 12) was the limiting step in the C 16 PUFAs pathway toward 16:3n-4 (16:3∆6, 9, 12). The fact that 16:1n-7 was among the most enriched FA in both NL and PL might reflect some storage process in "anticipation" of later C 16 PUFAs synthesis, energy shortage or stress response.

Interconnection between Polar and Neutral Pools in C. muelleri
The successions of fatty acid enrichments differed between NL and PL. While enrichments of 18:1n-9, 18:2n-6, 18:3n-6 and then 20:4n-6 were in agreement with the conventional n-6 pathway used by C. muelleri to synthesize its PUFAs in the polar fraction [14][15][16]19], the order varied for neutral lipids. Furthermore, the higher enrichments of 18:1n-9, 18:2n-6 and 18:3n-6 in PL as compared to NL strongly suggested that their synthesis and more specifically the insertion of the new double bound (desaturation) occurred in the form of glycolipid or phospholipid and are transferred to neutral lipids once synthesized. This can be linked to triacylglycerol (TAG) synthesis.
In diatoms, TAG are synthesized in the endoplasmic reticulum (ER) from diacylglycerol (DAG) [17,32]. They are used as energy or carbon storage. The synthesis of TAG occurs through the acyl-CoA pathway, called the Kennedy pathway. Some authors also hypothesized the possibility of TAG synthesis through an acyl-CoA-independent pathway [17,34]. The de novo Kennedy pathway consists in three consecutive acylations: (i) acylation of the sn-1 position of G3P by G3P acyltransferase (GPAT), producing lysophosphatidic acid (LPA), (ii) following esterification of LPA sn-2 position by lysophosphatidic acid acyltransferase (LPAAT) to give PA and (iii) finally, after release of the phosphate group, DAG sn-3 position acylation by DAG acyltransferase (DGAT) to produce TAG ( Figure 12).  In the acyl-CoA-independent pathway, the phospholipid:DAG acyltransferase (PDAT) needs PC as an acyl donor to form DAG. This pathway relies on the Lands cycle, which consists of an acylediting mechanism that progressively changes the fatty acids composition of PC to a more unsaturated one [33,37] and thus modifies the final composition of the newly formed DAG (and consequently TAG) (Figure 12). The enzyme responsible for the combination of PC and DAG, the phospholipid:diacylglycerol acyltransferase (PDAT), has been identified in plants [74,75]. Enzymes involved in this Lands cycle pathway linking PC with DAG to form TAG and genes coding for PDAT In the acyl-CoA-independent pathway, the phospholipid:DAG acyltransferase (PDAT) needs PC as an acyl donor to form DAG. This pathway relies on the Lands cycle, which consists of an acyl-editing mechanism that progressively changes the fatty acids composition of PC to a more unsaturated one [33,37] and thus modifies the final composition of the newly formed DAG (and consequently TAG) ( Figure 12). The enzyme responsible for the combination of PC and DAG, the phospholipid:diacylglycerol acyltransferase (PDAT), has been identified in plants [74,75]. Enzymes involved in this Lands cycle pathway linking PC with DAG to form TAG and genes coding for PDAT have been identified in the diatom Phaeodactylum tricornutum [35,36]. This could support the existence of this pathway in diatoms as well.
If we apply these putative pathways to the observed enrichment dynamics, it seemed likely that 18:1n-9 (PL) was first synthesized in the chloroplast or in the cytosol where its suspected precursors can be found and entered PC after being transferred to the ER to be rapidly desaturated in 18:2n-6 (PL) and 18:3n-6 (PL) by the acyl-editing machinery. Then, these newly formed PUFAs can enter the acyl-CoA pool and be used for de novo synthesis by Kennedy pathway in the ER or C 18 PUFAs can directly be exchanged with a low unsaturated FA in DAG to create a more highly unsaturated DAG, which was then used for TAG synthesis. Such interaction between phospholipids pool and DAG/TAG pool was suggested by Mus et al. [36] and Mülhroth et al. [35] following the discovery of gene coding for putative enzyme involved in this process. This would mainly concern 18:2 and 18:3 fatty acids [32]. Combined with our results, this supports the implication of PC-associated-PUFA in the TAG synthesis via DAG in diatoms ( Figure 12).
Based on cloning and functional characterization of P. tricornutum desaturases, Domergue et al. [66] and Mülhroth et al. [35] discussed the existence of connected pathways in diatoms between C 16 -PUFAs and 20:5n-3. Their idea came from the existence of a glycolipid presenting 16:3n-4 and 20:5n-3 at analogous positions (respectively position sn-2 and sn-1) of typical glycolipids (GL) produced via the prokaryotic pathway in P. tricornutum. Indeed, depending on the location of the synthesis of glycolipids (i.e., chloroplast "prokaryotic" or endoplasmic reticulum (ER) "eukaryotic"), the fatty acid composition varies: GL produced by the prokaryotic pathway will have C 16 fatty acids at the sn-2 position, while those formed in the ER will present longer chain fatty acids (likely C 18 /C 20 or C 22 ) at sn-2 position [80,81]. Our results suggested the plastid origin of the glycolipids linking 20:5n-3 (sn-1) and 16:3n-4 (sn-2). However, 20:5n-3 synthesis and the responsible desaturase in P. tricornutum seemed exclusively located out of the chloroplast, contrasting with higher plant metabolism [66]. Then, to be added to chloroplast glycolipids, 20:5n-3 had to be transferred from the ER to the plastid. Transfers of FA precursors from the ER to the plastid have been reported to exist in plants and microalgae [17,82].
Thus, as for other diatoms and following the results introduced here, a structural and metabolic link could exist between 16:3n-4 and 20:5n-3 in C. muelleri, even if they were not synthesized in the same cell compartments (Figure 13). 20:5n-3 synthesized in the ER was assumed to be used to form prokaryotic glycolipids in association with 16:3n-4 produced in the chloroplast. Along the course of its synthesis, the molecule passed through polar and neutral pools, for example, PA to DAG (PL to NL pool) or DAG to glycolipids (NL to PL pool), which would explain the opposite dynamics observed in this experiment. Similar dynamics observed with 18:4n-3 could also be related to eukaryotic DAG synthesis in the ER (Figure 13).

Conclusions
C. muelleri synthetized 20:5n-3 via the combined use of n-3 and n-6 pathways. The conventional route via 18:3n-3 seemed inefficient and/or slow in the diatom: apart from 18:4n-3 and 20:5n-3, all the other n-3 PUFAs were present in low concentrations. As an alternative, C. muelleri appeared to produce its n-3 PUFAs from the n-6 substrate via the ω-3 desaturase pathway. This was especially the case of 18:3n-6, transformed into 18:4n-3 and 20:4n-6 converted into 20:5n-3. It was not possible to discriminate which PUFA between 20:4n-6 and 18:4n-3 was predominant in EPA synthesis. The use of 13 C-label allowed us as well to assume the existence of an acyl-editing mechanism in C. muelleri responsible for the progressive desaturation of 18:1n-9. Initially produced in the plastid, this fatty acid would be transmitted to the ER combined with phosphatidylcholine (PC) to be desaturated into 18:2n-6 and 18:3n-6. This would allow the formation of a more unsaturated diacylglycerol, serving in turn to format of triacylglycerol or membrane lipids. C. muelleri relied on a tight connection between the neutral and polar pools. In the diatom, synthesis of galactolipids in the chloroplast might be dependent on acyl-CoA produced in the ER. This has been assumed with the opposite abundance dynamic of 16:3n-4 and 20:5n-3, which were both suspected to be located on the same glycerol backbone of glycolipids. 16:3n-4 would be produced in the plastid while 20:5n-3 by successive desaturation and elongation occurring in the ER.

Conclusions
C. muelleri synthetized 20:5n-3 via the combined use of n-3 and n-6 pathways. The conventional route via 18:3n-3 seemed inefficient and/or slow in the diatom: apart from 18:4n-3 and 20:5n-3, all the other n-3 PUFAs were present in low concentrations. As an alternative, C. muelleri appeared to produce its n-3 PUFAs from the n-6 substrate via the ω-3 desaturase pathway. This was especially the case of 18:3n-6, transformed into 18:4n-3 and 20:4n-6 converted into 20:5n-3. It was not possible to discriminate which PUFA between 20:4n-6 and 18:4n-3 was predominant in EPA synthesis. The use of 13 C-label allowed us as well to assume the existence of an acyl-editing mechanism in C. muelleri responsible for the progressive desaturation of 18:1n-9. Initially produced in the plastid, this fatty acid would be transmitted to the ER combined with phosphatidylcholine (PC) to be desaturated into 18:2n-6 and 18:3n-6. This would allow the formation of a more unsaturated diacylglycerol, serving in turn to format of triacylglycerol or membrane lipids. C. muelleri relied on a tight connection between the neutral and polar pools. In the diatom, synthesis of galactolipids in the chloroplast might be dependent on acyl-CoA produced in the ER. This has been assumed with the opposite abundance dynamic of 16:3n-4 and 20:5n-3, which were both suspected to be located on the same glycerol backbone of glycolipids. 16:3n-4 would be produced in the plastid while 20:5n-3 by successive desaturation and elongation occurring in the ER.