Plastidic D6 Fatty-Acid Desaturases with Distinctive Substrate Specificity Regulate the Pool of C18-PUFAs in the Ancestral Picoalga Ostreococcus tauri

Charlotte Degraeve-Guilbault,a,2 Rodrigo E. Gomez,a,2 Cécile Lemoigne,a,2 Nattiwong Pankansem,b Soizic Morin,c Karine Tuphile,a Jérôme Joubès,a Juliette Jouhet,d Julien Gronnier,a Iwane Suzuki,b Denis Coulon,a Frédéric Domergue,a and Florence Corelloua,3,4 Laboratoire de Biogenèse Membranaire, Unité Mixte de Recherche 5200, Centre National de la Recherche Scientifique-Université de Bordeaux, 33883 Villenave d’Ornon, France Faculty of Life and Environmental Sciences, University of Tsukuba, Tsukuba, Ibaraki 305–0006, Japan Institut National de la Recherche Agronomique, Unité de Recherche Ecosystèmes Aquatiques et Changements Globaux, 33612 Cestas, France Laboratoire de Biologie Cellulaire et Végétale, Unité Mixte de Recherche 5168, Centre National de la Recherche Scientifique-French Alternative Energies and Atomic Energy Commission-Institut National de la Recherche Agronomique-Université Grenoble Alpes, Interdisciplinary Research Institute of Grenoble, 38054 Grenoble, France

Eukaryotic D6-desaturases are microsomal enzymes that balance the synthesis of v-3 and v-6 C18-polyunsaturated fatty acids (C18-PUFAs) according to their specificity. In several microalgae, including Ostreococcus tauri, plastidic C18-PUFAs are strictly regulated by environmental cues suggesting an autonomous control of D6-desaturation of plastidic PUFAs. Here, we identified two putative front-end D6/D8-desaturases from O. tauri that, together with putative homologs, cluster apart from other characterized D6-desaturases. Both were plastid-located and unambiguously displayed a D6-desaturation activity when overexpressed in the heterologous hosts Nicotiana benthamiana and Synechocystis sp. PCC6803, as in the native host. Detailed lipid analyses of overexpressing lines unveiled distinctive v-class specificities, and most interestingly pointed to the importance of the lipid head-group and the nonsubstrate acyl-chain for the desaturase efficiency. One desaturase displayed a broad specificity for plastidic lipids and a preference for v-3 substrates, while the other was more selective for v-6 substrates and for lipid classes including phosphatidylglycerol as well as the peculiar 16:4-galactolipid species occurring in the native host. Overexpression of both D6-desaturases in O. tauri prevented the regulation of C18-PUFA under phosphate deprivation and triggered glycerolipid fatty-acid remodeling, without causing any obvious alteration in growth or photosynthesis. Tracking fatty-acid modifications in eukaryotic hosts further suggested the export of plastidic lipids to extraplastidic compartments.
Nutrients and abiotic stresses regulate the PUFA content in cyanobacteria and microalgae (Los et al., 2013;Khozin-Goldberg et al., 2016;Kugler et al., 2019). In particular, C18-PUFAs of plastidic lipids are highly remodeled in response to abiotic stresses: in the cyanobacteria Synechocystis sp. PCC6803, ALA and SDA synthesis is triggered by chilling while in several species from the Chromista kingdom, OPA is either increased or redistributed within molecular species (Tasaka et al., 1996;Kotajima et al., 2014;Leblond et al., 2019). As plants, green microalgae display a high amount of ALA, whereas the peculiar FA, 16:4 D4,7,10,13 is specific to green microalgae (Chlorophyta; Lang et al., 2011). Several microalga classes from the Chromista kingdom, such as haptophytes and dinoflagellates, produce both OPA and DHA (Jónasdóttir, 2019). The only green microalgae species that also encompass OPA and DHA belong to the Prasinophyta phylum (Dunstan et al., 1992). Importantly, the D6-DES product SDA is abundantly present in Chromista (exception made of diatoms) and Prasinophyta, where it is a precursor for both OPA and DHA synthesis. In OPA-producing species, OPA is preferentially located in galactolipids (GaLs), whereas DHA is prevalent in extraplastidic lipids (Bell and Pond, 1996;Degraeve-Guilbault et al., 2017;Leblond et al., 2019). In diatoms, 20:5n-3, prevalent in GaLs, is thought to be imported from the endoplasmic reticulum (ER) to the chloroplast. All D6-DES characterized so far are microsomal enzymes. However, in contrast to 20:5n-3, C18-PUFAs, which are overwhelmingly present in GaLs of many Chromista species except for diatoms, do not require ER located elongases for their synthesis. Therefore, how the regulation of D6-DES and intercompartmental FA fluxes is involved in regulating the production and distribution of OPA and DHA in microalgae species remains an important and unanswered question.
Ostreococcus tauri (class Mamiellophyceae, order Mamiellales) is an ancestral green picoalga that emerged early after the divergence between Chlorophyta and Streptophyta (land plants; Leliaert et al., 2012). O. tauri has the most minimal genomic and cellular organization (Derelle et al., 2006). This coccoid cell is smaller than 2 mm (a picoeukaryote), lacks a cell wall, flagella, or an obvious sexual life (Grimsley et al., 2010). However, it displays a large panel of PUFA, including hexadecatetraenoic acid, ALA, SDA, OPA, and DHA as major PUFAs (Wagner et al., 2010;Degraeve-Guilbault et al., 2017). O. tauri glycerolipids (GLs) characterization unveiled an exceptionally clear-cut allocation of PUFA in membranes: C18-PUFAs are prevalent in plastidic lipids, with OPA being restricted to GaL, whereas VLC-PUFA are exclusively found in extraplastidic GLs (Degraeve-Guilbault et al., 2017). Many microalgae, including Chlamydomonas reinhardtii, Ostreococcus spp., and related genera are devoid of phosphatidylcholine (PC). For O. tauri, the major extraplastidic GLs identified correspond to the betaine lipid diacylglycerylhydroxymethyl-trimethyl-b-Ala (DGTA) and the peculiar phosphosulfolipid phosphatidyldimethylpropanethiol (PDPT), both of which are typical of Chromista species. Interestingly, we uncovered that the plastidic C18-PUFA pool was strictly regulated by nutrient availability: under phosphate or nitrogen deprivation, ALA content was increased at the expense of SDA in plastidic GLs, the acyl-CoA pool, and triacylglycerols (TAGs), but not in extraplastidic GLs (Degraeve-Guilbault et al., 2017). This previous result strongly suggested that nutrient starvation specifically inhibited D6-DES of plastidic lipids; however, so far, the D6-DES characterized from O. tauri and related species was demonstrated to be microsomal and to have the peculiarity of using acyl-CoA as substrates (Domergue et al., 2005;Hoffmann et al., 2008;Petrie et al., 2010aPetrie et al., , 2010bVaezi et al., 2013). We therefore reasoned that the specific regulation of plastidic C18-PUFA in Mamiellales likely involves uncharacterized plastidic D6-DES, rather than the transfer of D6-DES products from the acyl-CoA pool to the chloroplast.

O. tauri Fatty Acid Desaturase Sequences Retrieval and Analysis
Thirteen canonical DES sequences were retrieved from genomic and transcriptomic databases (National Center for Biotechnology Information [NCBI] databases; Supplemental Table S1). All sequences were manually checked upstream of the predicted start codon, especially to assess the completion of the protein N-terminal (Nt) part; when required, extended Nt were validated by complement DNA (cDNA) amplification (see "Materials and Methods"; Supplemental Table S2). Except for an acyl-CoA-Δ6-DES (Ot13, according to the genomic accession) and an unknown DES barely related to sphingolipid Δ3/Δ4-DES, complete sequences (using the prediction software PredAlgo; Tardif et al., 2012) were all predicted to contain a chloroplastic target-peptide (cTP). Among the seven front-end DESs, three uncharacterized D6/D8 FA-DESs retained our attention ( Fig. 1; Supplemental Table S1). All three reference sequences from the O. tauri strain RCC4221 were completed at the Nt to include a predicted cTP (see "Materials and Methods"; Supplemental Figs. S1-S3; Supplemental Table S1). One D6-DES candidate (Ot15) clustered with acyl-lipid D6/D8-DES and was closest to the diatom Thalassiosira pseudonana Δ8-sphingolipid-DES ( Fig. 1; Tonon et al., 2005). The two other candidates (Ot05 and Ot10) were closely related (49.8% identity, 71.6% similarity), and together with putative homologs, formed a cluster apart from the typical acyl-CoA D6-DES specific to Mamiellophyceae species and acyllipid D6/D8-DES from plants, fungi, and worms (Supplemental Fig. S1B). Each of these two candidates had one putative homolog in the Mamiellophyceae species, with the exception of Bathycoccus prasinos, for which the only homolog was more closely related to Ot05 (Supplemental Figs. S2 and S3). The His-boxes (His-Box), typical of the DES domain, displayed consensus motifs contrasting with those from acyl-CoA-D6-DES (Table 1; Supplemental Figs. S2 and S3;Sayanova et al., 1997;López Alonso et al., 2003). A further conserved region occurred upstream of the third His-Box and was also divergent between the D6-DES candidates and the acyl-CoA-D6-DES. More distantly related putative homologs occurred in microalgae arising from secondary endosymbiosis, i.e. species from the Chromista and Euglenozoa supergroup (Supplemental Fig. S4). All homologous sequences, which were annotated as putative proteins, displayed the typical Nt-fused-Cyt-b5 domain found in front-end DES, and distinctive His-Box consensus motifs emerged from their alignment (Table 1; Supplemental Fig. S4).
The O. tauri acyl-CoA-D6-DES (Ot13) was previously characterized in yeast, and extensively used for the reconstruction of VLC-PUFA pathways in various organisms including plants (Domergue et al., 2005;Hoffmann et al., 2008;Ruiz-López et al., 2012;Hamilton et al., 2016). However, its activity has never been assessed in the native host, so Ot13 was chosen as a supplemental reference to achieve the functional characterization of the two closely related D6-DES candidates.

D6-DES-Candidate Localization and Activities in Heterologous Hosts
Full-length or codon-optimized Nt-truncated (for removing the putative cTP) open reading frames (ORFs) were expressed in Saccharomyces cerevisiae (Supplemental Fig. S5A). When 18:3n-3 was supplied, none of the candidates, either full length or Nt-truncated, displayed D6activity. However, expression of the Ot13 (the acyl-CoA D6-DES) produced significant amounts of 18:4n-3. Neither the supply of D3-, D5-, or D8-substrates, nor the coexpression with the Ot13, resulted in the appearance of novel FAs; this clearly indicated that neither Ot05 nor Ot10 were active in yeast (Supplemental Fig. S5B), Overexpression of full-length C-terminal-fused yellow fluorescent protein (Ct-YFP) in Nicotiana benthamiana unveiled that both Ot05 and Ot10 exclusively localized at plastids, while Ot13 was at the ER ( Fig. 2A). Furthermore, overexpression of the full-length ORFs in N. benthamiana led to the synthesis of 18:3n-6 and 18:4n-3 from the endogenous substrates 18:3n-3 and 18:2n-6, indicating that both Ot05 and Ot10 displayed D6-DES activity in this host Figure 1. Phylogenetic tree of O. tauri front-end DES and homologs in other species. The tree was constructed from the tool COBALT (NCBI) using the accessions indicated below (Fast minimum evolution method, Maximum sequence difference 0.85, and Grishin's distance). Species are indicated in brackets; numbering refers to putative (in italics) or assessed DES D-regiospecificity. S, sphingolipid-DES; P, plastidial DES; er, microsomal DES; 6CoA, Acyl-CoA-D6-DES; put prot, putative protein. The three D6/8-DES candidates are in bold, and the labels used in this article are in brackets. Colors of nodes refer to the taxonomic groups: cyanobacteria (purple), eukaryotes (gray), green algae (deep blue), eudicots (beige), cryptomonads (light blue), haptophytes (light green), cryptomonads (yellow), euglenoids (pale pink), kinetoplastids (bright pink), fungi (deep green), and nematodes (red).  (Fig. 2B). Interestingly, disrupting the heme-binding capacity of Cyt-b5 by H . A mutation in the HPGG motif abolished the activity of Ot05 and Ot10 (Supplemental Fig. S6A; Sayanova et al., 1999). Note that Ct-YFPs were also active in N. benthamiana (Supplemental Fig. S6B). Altogether, these results unambiguously demonstrated that both Ot05 and Ot10 are plastidic D6-DES (pD6-DES).
Significant differences emerged from the bulk of the experiments (Fig. 2, B and C). First, the D6-DES product 18:3n-6 was highest in Ot13-OE and lowest in Ot05-OE. The proportion of 18:2n-6 remained stable in pD6-DES-OE, suggesting that 18:2n-6 was a nonlimiting substrate for the production of 18:3n-6 in Ot05-OE and Ot10-OE. In contrast, 18:3n-3 was significantly reduced in all overexpressors and to a greater extent in Ot05-OE (Fig. 2B). In parallel, the proportion of 18:4n-3 was significantly higher in Ot05-OE compared to Ot10-OE and Ot13-OE. These observations strongly suggested that the higher production of 18:4n-3 in Ot05-OE mostly resulted from the preference of Ot05 for v-3 compared to v-6 substrates, whereas Ot13 and Ot10 both displayed a preference for v-6 substrates. Considering the v-class substrate conversion efficiency in each sample, Ot13-OE and Ot10-OE displayed a similar trend, with the conversion of v-6 C18-PUFA being approximately three times higher than that of v-3; conversely, v-3 substrate conversion was approximately twice that of v-6 in Ot05-OE (Fig. 2C).
The relative impact on GL classes of each D6-DES-OE, expressed as a conversion percent for each lipid, was analyzed from two independent experiments (Fig. 2D, Supplemental Fig. S7). The global lipid profile appeared unchanged in overexpressor lines (Supplemental Fig.  S7A). In Ot13-OE, extraplastidic phospholipids including phosphatydylcholine (PC), phosphatydilserine together with phosphatidic acid and phosphatidylethanolamine, were more impacted compared to plastidic GLs ( Fig. 2D; Supplemental Fig. S7, B and C); the v-6 conversion was nevertheless importantly increased in monogalactosyl diacylglycerol (MGDG) and a relatively high conversion rate was observed for phosphatidylglycerol (PG). Conversely, although extraplastidic GLs were obviously impacted in the pD6-DES-OE, the most altered lipids were MGDG in Ot05-OE and PG in Ot10-OE. Note that in Ot10-0E, the PG v-3 and v-6 indexes were both the highest. Finally, the overall impact on sulfoquinovosyl diacylglycerol (SQDG) was slightly more pronounced for Ot05-OE by considering the maximal index in Ot05transgenics compared to that in Ot13 and Ot10 transgenics.
To further clarify the substrate specificity of pD6-DES, the cyanobacterium Synechocystis sp. PCC 6803 was used. This organism not only encompasses the eukaryotic classes of plastidic lipids as major GLs but also allows transgene expression from a similar genomic environment (homologous recombination; Williams, 1988). Synechocystis sp. PCC 6803 has one D6-DES (desD) and one v-3-DES (desB). In wild type, the selectivity of DesD for GaLs is reflected by the exclusive distribution of 18:3n-6 in GaLs ( Fig. 3; Wada and Murata, 1990). The transcription of desB is induced at temperatures below 30°C and results in 18:3n-3 accumulation in PG and SQDG, and of 18:4n-3 in GaLs. Note that all GL species are sn-1/sn-2 18:X/16:0 combinations.
Expressing either Ot05 or Ot10 in DdesD cells of Synechocystis at 34°C (no endogenous v-3-DES activity) resulted in 18:3n-6 production at the expense of 18:2n-6 ( Fig. 3A). As for N. benthamiana, Overexpressor of an H . A mutated version of the D6-DES resulted in the absence of D6 products, indicating that integrity of the HPGG motif in the Cyt-b5 domain was required (Fig. 3A). The Ot05-OE restored the 18:3n-6 wild-type production, whereas the Ot10-OE resulted in a much lower production, indicating a higher v-6 activity of Ot05 compared to Ot10 in this host. Most interestingly, the accumulation of 18:3n-6 occurred not only in GaLs of both pD6-DES-OE, but also in PG as well as in SQDG for Ot05-OE, contrasting with the wild-type features (Fig. 3, B-D). Noteworthy is that the amount of 18:3n-6-PG was twice as high for Ot10-OE than Ot05-OE (isomer of 18:2), likely corresponding to the D6 desaturation product of 18:1n-9, as seen in Figure 3D.

D6-DES Overexpressor in O. tauri
To gain insight into the regulation of C18-PUFA pool by D6-DES in the native host, O. tauri overexpressors of each D6-DES were created using the pOtOXLuc vector, in which the high-affinity-phosphate-transporter promoter drives the overexpression while the luciferase reporter gene, under the CCA1 promoter, allows ready screening of the best transgenics (Moulager et al., 2010). Asterisks indicate statistically significant differences by Wilcoxon test (*P # 0.01; **P # 0.001; and ***P # 0.000001. C, Substrate conversion of v-3 and of v-6 C18-PUFA for each transformant. Conversion percent corresponds to the ratio P/(P1S), where P values (product) are either 18:4n-3 or 18:3n-6, and S (substrate) values are either 18:3n-3 or 18:2n-6 for v-3 or v-6, respectively. Samples used for lipid analyses are indicated by black (C) and blue arrows (Supplemental Fig. S7). B and C, Control lines (p19) n 5 27; Ot13-OE n 5 17; Ot10-OE n 5 21; Ot05-OE n 5 29. D, Impact of D6-DES-OE on polar GLs. Conversion percent of w-3 and of v-6 C18-PUFA is used as an index of the impact of the overexpression of each desaturase on v-3 and v-6 in each lipid class. The overall impact (Total) corresponds to averaged v-3 and v-6 conversion percent. Color scaling highlights minimal (white) and maximal values (darkest color) in polar lipids of each overexpression line. One representative experiment out of two is shown. The replicate experiment is available from Supplemental Figure S7.

Screening and Selection of D6-DES Transgenic Lines
Cultures were grown under low phosphate conditions to allow full activation of the high-affinity-phosphate-transporter promoter. These conditions were previously suspected to prevent D6-DES of C18-PUFAs, and were therefore ideal for the screening of D6-DES OE (Degraeve-Guilbault et al., 2017). Transgenics of each of the D6-DES were retrieved according to their luminescence level. Five transgenics were selected to ascertain their FA phenotype and determine the level of D6-DES gene expression ( Fig. 4; Supplemental Figs. S8 and S9). The relative content of C18-PUFA of these transgenics varied accordingly to the luminescence and transgene expression levels. For similar luminescent levels, the relative amount of v-3 C18-PUFA varied to a greater extent in Ot05-transgenics compared to Ot10 transgenics, whereas that of v-6-C18-PUFA was similarly impacted in Ot05 and Ot10 transgenics; these variations, although significant, were far more modest for most luminescent Ot13 transgenics ( Fig. 4C; Supplemental  Fig. S9). Considering the averages of the best transgenics for each desaturase, it clearly appeared that pD6-DES OE importantly impacted v3-C18-PUFA, including an increase in 18:5n-3 for the best Ot05 OE (Supplemental Fig.  S9). Although the impact of Ot05-OE was the most pronounced, it was striking that the impact of Ot10-OE was much greater than in heterologous hosts. Furthermore, considering the average values from the best transgenics, the proportion of 16:4n-3 was most significantly decreased whereas that of 16:0 was increased in Ot05 lines (Supplemental Figs. S8 and S9).
One transgenics out of five that was representative of each D6-DES OE was selected (lines Ot13-5, Ot10-5, and Ot05-5), taking care that expression levels of D6-DES were similar, and further detailed lipid analyses were achieved (Supplemental Fig. S8C). These lines are referred to as to Ot13-OE, Ot10-OE, and Ot05-OE PCC6803. The activities of desD and desB for the regulation of C18-PUFA in GLs (bold arrows) and SQDG, PG (dashed arrow) in the wild-type is schematically represented (adapted from Wada and Murata, 1990): broad arrows represent desaturation reactions that are stimulated after a shift in temperature; thin arrows minor desaturation reactions. A to D, GL FA profile at 34°C. E to H, GL FA profile at 24°C. Means and SDs of three independent experiments are shown. MGDG and DGDG displayed similar alterations and were cumulated (GaLs). The statistical Wilcoxon test was performed for C18 variations and most relevant results are shown (°P # 0.1 and *P # 0.05); some non-significant differences (ns) are further highlighted. hereafter (Fig. 4, D and E). Note that no growth defect was detected in the best D6-DES-OEs (Fig. S10).
Finally, TAG FA-profile and molecular species were importantly impacted by the overexpression of each of the three D6-DES lines ( Fig. 5H; Supplemental Figs. S12 and S13). Noteworthy, regarding Ot13-OE, is the clear alteration of the ratio of D6-DES-substrates/D6-DESproducts in TAGs, that contrasted with the minor alterations detected in structural lipids (Fig. 5H). For Ot05-OE, the level of 16:4n-3 was also significantly reduced while that of 16:0 was increased, recalling the changes occurring in DGTA and DGDG.
Despite the complexity of mass-to-charge ratio (m/z) TAG species, distinctive features emerged from MS analyses. Overall, the most pronounced changes in all D6-OE were related to m/z species containing D6substrates and products that corresponded to the major species in wild type: the 18:3-TAG species 48:3 and 50:3 were the most reduced, whereas the 18:4-TAG species 48:4, 50:4, and 50:8 were the most increased (Supplemental Figs. S12 and S13). For Ot05-OE, the species combining 22:6 and 18:4 (54:10, 56:6) were

Physiological Relevance of pD6-DES Regulation
Transcriptional regulation of desaturases is known to occur in response to environmental cues. We therefore assessed transcript levels of desaturases involved in the regulation of the C18-PUFA pool by phosphate availability, including the putative v3-DES (Kotajima et al., 2014). Consistent with our previous report, the proportion of 18:3n-3 was increased by approximately onehalf after the transfer of cells to phosphate-depleted medium (Fig. 6A ). By that time, the transcript level of Ot05 was decreased by .60%, the transcript levels of Ot10 and of the putative v-3-DES remained stable, and transcript levels of the acyl-CoA DES were increased (Fig. 6B). This result indicated that a decrease in Ot05 activity through transcriptional repression resulted in lowering of the 18:4n-3/18:3n-3 ratio under phosphate deprivation.
Thylakoid membrane PUFA composition is known to influence photosynthetic processes (Allakhverdiev et al., 2009). With the aim of assessing a possible impact of C18-PUFA alterations on photosynthesis, photosynthetic parameters and photoinhibition responses were investigated in the pD6-DES-OE Ot05 and Ot10. Photosynthetic parameters were recorded from two independent batches at stationary stage. Interexperimental variations were high and the slight differences observed in one of the experiments appeared to be not reproducible (Table 2; Supplemental Table S4). We therefore concluded that the effect of the transgene overexpression, if any, was not prevalent over the differences related to batch culture conditions, i.e. selfshading slight differences of progression into stationary phase. Photoinhibition and recovery responses of pD6-DES OE were also similar to control lines. Accordingly, we concluded that no obvious alterations regarding the photosynthesis features investigated could be detected in pD6-DES-OE in our conditions. Further detailed investigations will be required to possibly unveil photosynthesis defects and/or compensatory mechanisms in these lines.

DISCUSSION
FA remodeling of structural lipids in response to chilling in plants and cyanobacteria involves the regulation of C18-PUFA desaturation (Los et al., 2013); downstream synthesis of v-3 and v-6 VLC-PUFA in animals, fungi, and microalgae also involves the finetuning of C18-PUFA, notably by D6-DES. On the other hand, all front-end D6-DES studied so far are demonstrated to be, or assumed to be, located at the ER (Meesapyodsuk and Qiu, 2012). By unveiling the first plastidic D6-DES in the ancestral green picoalga O. tauri, our work points out the autonomous control of plastidic C18-PUFA in microalgae. The entangled substrate features instructing us on the activity of these two novel D6-DES, the possible PUFA fluxes unveiled by their overexpression, and the physiological significance of pD6-DES, are discussed here.
Substrate Specificity of O. tauri D6-DES DES specificity relies on intricate substrate features, including the acyl-chain position, length, and unsaturation level, as well as the acyl-carrier nature (Heilmann et al., 2004a;Li et al., 2016a). Although the primary importance of the His-Box and adjacent regions for front-end desaturase activity and substrate specificity could be demonstrated, neither the exact molecular features underlying DES (regio)specificity nor the hierarchical importance of these substrate features have yet been clearly identified (Song et al., 2014;Li et al., 2016a;Watanabe et al., 2016). Furthermore, assaying plant DES activity in S. cerevisiae, which lacks plastidic lipid classes, might have introduced some bias by favoring the identification of microsomal desaturases and/or by hampering the proper characterization of plastidic substrates of desaturases (Heilmann et al., 2004b).
In this work, four different hosts were used to characterize pD6-DES substrate specificity. Lipid changes occurring in each of these organisms reflect a steady state arising from both desaturation and overall FA fluxes. Nevertheless, comparison of lipid features triggered by the overexpression of each D6-DES in a given host, and of the same D6-DES in different hosts, allowed us to gain insight into pD6-DES substrate specificity. In 16:3-plants (Arabidopsis [Arabidopsis thaliana] and N. benthamiana), the Kennedy pathway contributes to the synthesis of plastidic lipid yielding sn-1/sn-2 18:3/ 18:3 species in addition to the 18:3/16:3n-3-lipid species that arise from de novo plastidic synthesis (Browse et al., 1986b). In O. tauri, as in C. reinhardtii and cyanobacteria, plastidic lipids correspond to sn-1/sn-2 18:X/16:X species while the major extraplastidic lipid species are di-VLC-PUFA in DGTA (22:6/22:6) and saturated FA/16:4, and VLC-PUFA/16:4 in both DGTA and PDPT (Degraeve-Guilbault et al., 2017). These distinctive positional signatures strongly suggest that, in O. tauri, plastidic lipid synthesis is independent of ER synthesis (Ohlrogge and Browse, 1995). On the other hand, acyl-lipid remodeling of plastidic lipids is assumed to be absent in Synechocystis species, for which no acyl-turnover was ever reported. Concerning microalgae, MGDG has been proposed, yet not clearly proven, to be a platform of FA exchange supporting the incorporation of plastidic FA into TAGs in C. reinhardtii (Li et al., 2012;Kim et al., 2018). We have based the interpretation of our results on this knowledge.

Head-Group Specificity
As major changes occurred in GaLs independently of the host, it can reasonably be concluded that at least MGDG is a substrate of both Ot05 and Ot10. Most interestingly, Ot05 OE in Synechocystis PCC 6803 triggered the accumulation of D6-DES products in SQDG; SQDG was also unambiguously more impacted in O. tauri Ot05-OE compared to Ot10-OE and slightly more in N. benthamiana Ot05-OE. Ot10 OE in Synechocystis sp. PCC 6803 and N. benthamiana impacted PG more than other lipid classes, including the production of 18:2-PG (likely 18:2 D6,9 ) in Synechocystis sp. Altogether, these results suggest that Ot05 displayed a broad specificity for plastidic substrates, while Ot10 appears to be selective for PG, at least in heterologous hosts, as well as for some GaL species in the native host (see below). Interestingly, in Arabidopsis, the plant chloroplastidic v-3-DES FAD8 was shown to display a preference for PG compared to the closely related FAD7 (Román et al., 2015).
The impact of pD6-DES OE on extraplastidic structural GLs from N. benthamiana and O. tauri could be explained either by the access of pD6-DES to extraplastidic substrates or by the export of overproduced PUFA to extraplastidic membranes. Because overexpression of the pD6-DES without the cTP part did not yield any D6-DES products in S. cerevisiae whereas they did in Synechocystis sp. PCC 6803, it is reasonable to assume that PC is not a suitable substrate for O. tauri pD6-DES and, accordingly, that the occurrence of D6-DES products in the PC of N. benthamiana likely arose from their export from the chloroplast (Domergue et al., 2003). In support of this assumption, one should recall that Ostreococcus D5-DES and D4-DES OE in S. cerevisiae resulted in the synthesis of a barely detectable amount of product (Tavares et al., 2011;Ahmann et al., 2011;Vaezi et al., 2013). Because we showed that PC is absent from O. tauri (Degraeve-Guilbault et al., 2017), the low activity of these two DESs might be related to the lack of native substrates (DGTA and possibly PDPT) in S. cerevisiae. Altogether, our results and previous reports therefore point to the importance of the structural lipid head-group for front-end DES activity. Consequently, we propose that the first level of substrate recognition of O. tauri pD6-DES relies on the head-group of plastidic lipid. Recently, cocrystallization of the Stearoyl-CoA DES with its substrate revealed that interaction  between the DES and the acyl-carrier was indeed fundamental to orient the acyl-chain in the catalytic tunnel of the enzyme (Wang et al., 2015).
The O. tauri acyl-CoA-D6-DES has been reported to efficiently produce D6-DES products in the heterologous host, but it has never been characterized in the native host. We confirmed that O. tauri acyl-CoA-D6-DES OE in N. benthamiana resulted in the accumulation of D6-DES products in all phospholipids; its overexpression in O. tauri moderately impacted structural GLs, but led to important accumulation of D6-DES products in TAGs. These results suggested that oversupplied D6-DES products from the acyl-CoA pool are preferentially incorporated into TAGs in the native host, whereas in N. benthamiana leaves, they are preferentially incorporated into PC, and then in part, likely transferred to MGDG (see below).

v-3/v-6 and 16:4-GL Species Selectivity
For all host organisms in this study, the impact of Ot05-OE on v-3-C18-PUFA was more pronounced compared to that of Ot10, indicating a preference of Ot05 for v-3-substrates. Conversely, Ot10 appeared globally more selective for v-6 substrates. Nevertheless, compared to other organisms, the overexpression of Ot10 in O. tauri impacted the amount of v-3-C18-PUFA to a greater extent, especially for 16:4-GaL species. Considering that Ot10 could display a higher v-6 desaturation activity in the native host, the v-3-desaturation of the overproduced 18:3n-6 in Ot10-OE might be involved in the rise of 18:4n-3. However, one would expect a higher accumulation of 18:3n-6 in Ot10-OE compared to Ot05-OE if only this route was used, which is not the case. Indeed, the endogenous v-3-desaturase and the overexpressed D6-DES compete for 18:2n-6; consequently, if 18:3n-6 production prevailed over that of 18:3n-3 (i.e. D6-DES activity higher than v-3 activity), one would expect that the conversion of 18:3n-6 to 18:4n-3 would be limited and would result in the higher proportion of 18:3n-6 in Ot10-OE compared to Ot05-OE. Accordingly, the most reasonable assumption is that Ot10 has a preference for highly unsaturated v-3 16:4-GaLs.

Export of Plastidic PUFAs
Overexpression of each of the three D6-DES in O. tauri led to a similar and significant increase of 20:4n-6 in DGTA and TAGs and a specific and significant increase of 18:4n-3 in DGTA for Ot05-OE. We previously reported that the acyl-CoA pool was enriched in 18:3n-3 and 18:4n-3, whose amount varied according to the plastidic C18-PUFA content (Degraeve-Guilbault et al., 2017). Together with the discovery of pD6-DES, these previous observations support that D6-DES products are exported from the plastid to the acyl-CoA pool in O. tauri. Though it cannot be excluded that pD6-DES have (limited) access to extraplastidic substrates, as suggested for the plastidic v-3-DES C. reinhardtii, it should be emphasized that the specificity of front-end DES is known to be more restricted than that of v-3-DES (Meesapyodsuk and Qiu, 2012;Nguyen et al., 2013;Wang et al., 2013).
Unexpectedly, pD6-DES-OE impacted extraplastidic GLs to a much greater extent in N. benthamiana. As mentioned above, these changes are likely arising from reallocation of xenobiotic plastidic D6-DES products to other membranes, possibly to circumvent deleterious effects, as well as from the low capacity of leaves to synthesize TAGs. A two-way exchange of lipids between the chloroplast and the extrachloroplastic compartments was proposed back in the mid-1980s to explain the altered 18:3 content of extraplastidic lipids observed in Arabidopsis plastidic v-3-DES (FAD7)-deficient mutants (Browse et al., 1986a). Most of the following work focused on the transfer of lipids and PUFAs from the ER to the chloroplast, establishing the idea that the reverse transport was negligible (Miquel and Browse, 1992;Li et al., 2016b). Nevertheless, the recent characterization of two Arabidopsis plastid lipase mutants highlighted that plastidic PUFA indeed contributed to PUFA remodeling of extraplastidic GLs (Wang et al., 2017;Higashi et al., 2018).

Significance of Plastidic D6-DES
The absence of alteration in growth and photosynthetic processes in O. tauri D6-DES-OE might be related to certain compensatory mechanisms. These may include the increase of 16:0 and the decrease of 16:4, possibly alleviating any important changes in membrane unsaturation from the 18:3n-3 and 18:4n-3 variations that occur in Ot05-OE. On the other hand, photosynthesis defects were not detected in Synechocystis mutants devoid of 18:3n-3 and 18:4n-3, and such defects could only be unveiled in very specific conditions in the Arabidopsis mutant lacking trienoic PUFA (Gombos et al., 1992;Vijayan and Browse, 2002). There is, overall, very little evidence that plastidic PUFAs directly support photosynthetic processes (Mironov et al., 2012;Kugler et al., 2019).
The only plastidial front-end desaturase so far described was a D4-DES from C. reinhardtii, and its Cyt-b5 domain was shown to be active in vitro (Zäuner et al., 2012). Our data further show that a functional Cyt-b5 is absolutely required for O. tauri pD6-DES activity in both N. benthamiana and Synechocystis sp. PCC 6803. These results illustrate the tight coevolution of Cyt-b5 and desaturase domain in Cytb5 fused front-end DES (Napier et al., 2003). It further indicates that another redox partner, different from the eukaryotic Cyt-b5 oxidoreductase, is involved (Napier et al., 2003;Kumar et al., 2012;Meesapyodsuk and Qiu, 2012). One possible candidate is the Ferredoxin NADP 1 -reductase (Yang et al., 2015).
The requirement of three D6-DES in a photosynthetic organism that displayed the most reduced set of genes points out the necessity of distinctly regulating the chloroplast and the cytosolic C18-PUFA pools. The existence of putative homologs of pD6-DES in other microalgae species, such as haptophytes and dinoflagellates, might also be related to the co-occurrence of 18:4n3 and 18:5n-3 in GaLs, and the prevalence of VLC-PUFA in extraplastidic lipids. For the diatom Phaeodactylum tricornutum, the putative pD6-DES homolog might be involved in the desaturation of plastidic C16-PUFA, as was suspected long ago (Domergue et al., 2002). For most Mamiellophyceae, with the exception of B. prasinos, the two pD6-DES most likely arose from gene duplication. Ot10 would have evolved to restrict its specificity to a particular set of substrates. Transcriptional regulation of Ot05 by phosphate deprivation indicates that it is a physiological target under these conditions. The selectivity of Ot05 for SQDG, which is considered as a surrogate for PG under phosphate limitation, supports the use of Ot05 for conditions in which PG is metabolized to provide phosphorus. Ot10 and the putative v-3-DES transcriptional regulation might be targeted by other environmental cues, which still need to be uncovered.

MATERIALS AND METHODS
Chemicals were purchased from Sigma-Aldrich, unless otherwise stated.
Annotated ORFs were manually checked for the completion of Nt sequences in species from the class Mamiellophyceae; cTPs were predicted with the tool PredAlgo (Tardif et al., 2012); alignment of Mamiellalophyceae homologous sequences, identified by BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi) search, was used to further determine putative cTPs assumed to correspond to the nonconserved Nt region (SnapGene trial version; Clustal Omega). These nonconserved regions were discarded for expression in Saccharomyces cerevisiae and Synechocystis sp. Codon-optimized sequences were obtained from Genewiz (Europe).

Cloning Strategy
PCR amplifications of DES ORFs were achieved using Q5 Polymerase by twostep PCR on a cDNA matrix (primers are indicated in Supplemental Table S2). A Monarch DNA Gel Extraction kit was used as needed (New England BioLabs). Overexpression vectors were OtoxLuc for O. tauri (Moulager et al., 2010), pTHT2031S for Synechocystis sp. PCC 6803 (Kotajima et al., 2014), GATEWAY destination vector pVT102-U-GW for S. cerevisiae (Domergue et al., 2010), and PK7W2G2D for N. benthamiana (Karimi et al., 2002). For subcellular localization, the final destination vector was pK7YWG2 (Karimi et al., 2002) Nt-YFP-fusion was used. Subcloning was performed in pGEMT vector by restriction enzymes (Promega), in pUC57 for the codon-optimized sequence used in Synechocystis and for S. cerevisiae (GenScript Biotech), and/or in pDONR 221 for GATEWAY cloning. Restriction cloning was used for cloning in pOtLux, and ligation was used to introduce the synthetic gene with pTHT2031S In-Fusion HD Cloning Kit (Takara Bio). Sequencing was achieved by Genewiz.
Site-directed mutagenesis H . A of the HPGG motif of the Cyt-b5 domain was either performed by GenScript (GenScript Biotech; N. benthamiana) or using an In-Fusion HD Cloning Kit (Takara Bio) for Synechocystis after amplification using two mutagenic complementary primers for amplifying pTHT2031-Ot5H46A-S and pTHT2031-Ot10H20A-S from pTHT2031-Ot5-S and pTHT2031-Ot10-S, respectively. The mutated DNA sequence was validated for the correct modification using a BigDye Terminator v.3.1 Cycle Sequencing Kit (Life Technologies/Thermo Fisher Scientific).

RNA and cDNA Preparation and Real-Time Quantitative PCR Analysis
An RNeasy-Plus Mini Kit (Qiagen) was used for RNA purification. rDNase I was used to remove contaminating DNA (DNA-Free Kit; Invitrogen), and cDNA was obtained using an iScript Reverse Transcription Supermix kit (Bio-Rad). Real-time quantitative PCR reactions were performed in a CFX96 Real-Time System (Bio-Rad) using the GoTaq qPCR Master mix (Promega; primers are indicated in Supplemental Table S3). The software CFX Manager (v.3.1; Bio-Rad) was used for data acquisition and analysis. The Ct method was used to normalize transcript abundance with the references mRNA EF1a (elongation factor), CAL (calmodulin), and Actin protein-related2. PCR efficiency ranged from 95% to 105%. A technical triplicate was used, and at least two independent experiments were performed.

Genetic Transformation
O. tauri electroporation was adapted from Corellou et al. (2009). Transgenics were obtained by electroporation and prescreened according to their luminescent level (Moulager et al., 2010). S. cerevisiae was transformed using a PEG/lithium acetate protocol, and FA supplementation was achieved. Control lines are transgenics of empty vectors. N. benthamiana leaves from 5-week-old plants were infiltrated with A. tumefaciens previously transformed by electroporation; p19 protein, to minimize plant post-transcriptional gene silencing, was used in all experiments. Briefly, A. tumefaciens transformants were selected with antibiotics (25 mg mL 21 of gentamycin with 100 mg mL 21 of spectinomycin or 50 mg mL 21 of kanamycin). A. tumefaciens transformants were grown overnight, diluted to an OD 600 to 0.1 and grown to an OD 600 of 0.6 to 0.8. Cells were resuspended in 5 mL of sterilized water for a final OD of 0.4 and 0.2 for overexpression and subcellular localization experiments, respectively, and 1 mL was agroinfiltrated. Plants were analyzed 2 and 5 d after A. tumefaciens infiltration for subcellular localization experiments and for overexpression, respectively.
Synechocystis sp. PCC 6803 transformation was achieved by homologous recombination (Williams, 1988). Briefly, the plasmid was transformed into 103 concentrated cells of the DdesD strain collected at midlog phase. Subsequently, the cell was incubated at 30°C under white fluorescent lamps for 16 to 18 h and selected by 25 mg mL 21 of chloramphenicol and 5 mg mL 21 of spectinomycin on BG-11 solid media (1.5% [w/v] Bacto-agar; BD Diagnostics/VWR).
MS analyses of O. tauri GL species was performed as in Abida et al. (2015). Purified lipids were introduced by direct infusion (electrospray ionization MS) into a trap-type mass spectrometer (LTQ-XL; Thermo Fisher Scientific) and identified by comparison with standards. Lipids were identified by tandem mass spectrometry analysis with their precursor ion or by neutral loss analyses. Positional analysis of FA in GLs was performed as in Degraeve-Guilbault et al. (2017).

Confocal Microscopy
Live cell imaging was performed using a SP5 confocal laser scanning microscopy system (Leica) equipped with Argon, DPSS, He-Ne lasers, and hybrid detectors, and a 633 oil-immersion objective. N. benthamiana leaf samples were transferred between a glass slide and coverslip in a drop of water. Fluorescence was collected using excitation/emission wavelengths of 488/490 to 540 nm for chlorophyll, 488/575 to 610 nm for YFP, and 561/710 to 740 nm for mCherry. Colocalization 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.

Photosynthesis Measurement
Measurements were made using a PHYTO-PAM analyzer (Heinz Walz).

Light Response of PSII Activity
Rapid light-response curves (RLCs) of chlorophyll fluorescence of the cultures were achieved according to the method in Serôdio et al. (2006). Briefly, the cultures were exposed to 12 increasing actinic light levels (10-s light steps of 100 mE increase from 64 to 2064 mE), and the electron transport rates (ETRs) were calculated on each step to draw RLCs. The following parameters were extracted from the ETR-irradiance curve fitted to the experimental data: the initial slope of the curve (a), the light-saturation parameters (I k ), and the maximum relative ETR.

Photoinhibition Experiment
Optimal conditions for PSII inhibition and recovery were adapted from Campbell and Tyystjärvi (2012). Cultures (50 mL, 33) were maintained under fluorescent white light (low light: 30.4 6 1.0 mE, white light) without agitation at 20.2 6 0.2°C and moved to high light (117.6 6 4.9 mE, blue LED) for 45 min (photoinhibition); photo recovery under initial condition was monitored for over 2 h. One-mL sampling was used to assess photosynthetic efficiency (quantum yield of photochemical energy conversion in PSII; Y corresponds to Yield 5 dF/Fm).

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. Sequence features of O. tauri D6-DES candidates and of Mamiellales homologs.
Supplemental Figure S4. Alignment of D6-DES candidate and closest homologs from phylogenetically distant species.
Supplemental Figure S5. Heterologous expression in S. cerevisiae.
Supplemental Figure S7. GL analyses of N. benthamiana leaves transformed with O. tauri D6-DES candidates.
Supplemental Figure S9. FA-profile of best transgenics for each D6-DES.
Supplemental Figure S13. TAG molecular species analysis.
Supplemental Table S1. O. tauri protein sequence features of FA desaturases.