Characterization of the Bubblegum acyl-CoA synthetase of Microchloropsis gaditana

Abstract

The metabolic pathways of glycerolipids are well described in cells containing chloroplasts limited by a two-membrane envelope but not in cells containing plastids limited by four membranes, including heterokonts. Fatty acids (FAs) produced in the plastid, palmitic and palmitoleic acids (16:0 and 16:1), are used in the cytosol for the synthesis of glycerolipids via various routes, requiring multiple acyl-Coenzyme A (CoA) synthetases (ACS). Here, we characterized an ACS of the Bubblegum subfamily in the photosynthetic eukaryote Microchloropsis gaditana, an oleaginous heterokont used for the production of lipids for multiple applications. Genome engineering with TALE-N allowed the generation of MgACSBG point mutations, but no knockout was obtained. Point mutations triggered an overall decrease of 16:1 in lipids, a specific increase of unsaturated 18-carbon acyls in phosphatidylcholine and decrease of 20-carbon acyls in the betaine lipid diacylglyceryl–trimethyl–homoserine. The profile of acyl-CoAs highlighted a decrease in 16:1-CoA and 18:3-CoA. Structural modeling supported that mutations affect accessibility of FA to the MgACSBG reaction site. Expression in yeast defective in acyl-CoA biosynthesis further confirmed that point mutations affect ACSBG activity. Altogether, this study supports a critical role of heterokont MgACSBG in the production of 16:1-CoA and 18:3-CoA. In M. gaditana mutants, the excess saturated and monounsaturated FAs were diverted to triacylglycerol, thus suggesting strategies to improve the oil content in this microalga.

A heterokont Bubblegum acyl-CoA synthetase (ACSBG), or lipidosin, is essential in Microchloropsis (Nannochloropsis) gaditana and thio-esterifies 16:1 and 18:3 fatty acids to Coenzyme A in vivo.

Introduction

In photosynthetic eukaryotes, the glycerolipid composition of subcellular membranes is specific to each compartment; it results from the combination of biosynthetic pathways, located mainly at the chloroplast and the endoplasmic reticulum (ER), and an intense trafficking of lipid intermediates within the cell (Benning, 2008, 2009; Li-Beisson et al., 2010, 2017, 2019; Boudiere et al., 2012; Horn and Benning, 2016; LaBrant et al., 2018). Glycerolipids result from the combination of three types of building blocks, i.e. a glycerol backbone originating from glycerol-3-phosphate (G3P, each carbon being numbered as sn-1, sn-2, and sn-3), fatty acids (FAs) of different carbon chain lengths and desaturation levels (indicated for instance as 18:1 for 18 carbons and 1 double bond) and polar heads of various chemical structures (Supplemental Figure S1).

The metabolic pathways are well described in cells containing chloroplasts limited by a two-membrane envelope (so-called “primary plastids” in Archaeplastida). A gap of knowledge needs to be filled in cells containing plastids limited by four membranes (“secondary plastids” in various clades, including heterokonts; Petroutsos et al., 2014; Li-Beisson et al., 2019).

In Archaeplastida, such as plants or green algae, de novo synthesis of FAs bound to acyl carrier proteins (ACPs) occurs in the stroma of chloroplasts, generating acyl-ACPs, mainly palmitoyl- and stearoyl-ACP (16:0-ACP and 18:0-ACP), which can be desaturated by a stromal acyl-ACP desaturase (AAD), producing 16:1-ACP and/or 18:1-ACP. Acyl-ACPs can be directly used to esterify G3P at the inner envelope membrane of the plastid for the production of precursors for the four major chloroplast glycerolipids, i.e. three classes of glycoglycerolipids (monogalactosyldiacylglycerol, MGDG; digalactosyldiacylglycerol, DGDG, and sulfoquinovosyldiacylglycerol, SQDG) and one phosphoglycerolipid (phosphatidylglycerol, PG; Li-Beisson et al., 2010; Boudiere et al., 2014). Chloroplasts also export FAs and import glycerolipid precursors: acyl-ACPs can be converted into free FAs (FFAs) by a chloroplast-specific thioesterase; FFAs then move to the cytosol, where they are thioesterified to Co-enzyme A (CoA), thus forming acyl-CoAs (Li et al., 2015, 2016; Li-Beisson et al., 2017). This thioesterification is catalyzed by acyl-CoA synthetases (ACSs), acting therefore in the control of FA fluxes in the cytosol, and their sorting toward different end-products (Coleman et al., 2002; Mashek et al., 2006; Mashek et al., 2007). Acyl-CoAs are acyl donors for the formation of the initial precursors for all membrane phosphoglycerolipids in the ER, like phosphatidylcholine (PC) or phosphatidylethanolamine (PE), or for triacylglycerol (TAG) stored in lipid droplets (Li-Beisson et al., 2010).

Superimposed to this glycerolipid biosynthetic network, FAs can be “removed” from a given glycerolipid by specific lipases, releasing FFAs, and be replaced by the action of specific acyl-CoA-dependent acyltransferases. This process is known as “acyl editing” (Bates et al., 2012; Hurlock et al., 2018; Menard et al., 2018). The pool of acyl-CoAs can be used for the elongation of FAs by specific elongases and the synthesis of other acylated molecules, such as sterol-esters or acyl-proteins. Eventually, acyl-CoAs can be directed to catalytic organelles, such as mitochondria or peroxisomes, to be specifically degraded by a process known as beta-oxidation (Pei et al., 2003).

It is difficult to deduce the precise cellular function of ACSs, since they can be essential in the activation of de novo synthesized FAs as well as recycled ones, for a broad diversity of “acyl-CoA-dependent” purposes (Figure 1). Some ACSs operate in more than one pathway (Coleman et al., 2002; Mashek et al., 2006, 2007). The function of each ACS depends therefore on multiple parameters, including substrate specificity, temporal expression pattern, location within the cell, interaction with other proteins, etc. Genetic impairment of ACS genes appears therefore as a method of choice to address functionality.

Figure 1.

Central role of acyl-CoAs in cell metabolism. A, In this simplified scheme, acyl-CoAs are initially generated from FAs synthesized de novo by the action of specific ACSs. B, Acyl-CoAs can be esterified to the sn-1 and sn-2 position of G3P, thus generating phosphatidic acid, which serves as a substrate for DAG and other membrane glycerolipids. Some specific acyl-CoAs, such as 18:3-CoA, can be further elongated and transferred to specific glycerolipids for additional desaturations (dashed lines), thus producing very-long chain polyunsaturated FAs (VLC-PUFAs). C, Acyl-CoAs can be esterified to the sn-3 position of DAG to form TAG. D, Furthermore, acyl-CoAs can be used for the synthesis of a variety of other acyl-molecules, such as sphingolipids, acylated proteins, sterol esters, and signaling molecules. E, In this scheme, FAs hydrolyzed from membrane glycerolipids or TAG can re-enter the pool of acyl-CoAs by the action of other specific ACSs. F, Eventually, acyl-CoAs can be degraded via the beta-oxidation pathway occurring in mitochondria and/or peroxisomes. Adapted from Coleman et al. (2002).

By contrast with Archaeplastida, Heterokonta contain a plastid with four limiting membranes (Fussy and Obornik, 2018). In models such as the diatom Phaeodactylum tricornutum or the eustigmatophytes Microchloropsis gaditana and closely related species such as Nannochloropsis oceanica, the “two innermost” membranes of the secondary plastid are considered homologous to the two-membrane envelope of plant chloroplasts (the inner and outer envelope membranes, IEM and OEM, respectively). The “outermost membrane,” the epiplastid membrane (EpM), is continuous with the outer nuclear envelope membrane connected to the ER. The periplastid membrane (PpM) located between the EpM and the OEM is of a still unresolved origin (Flori et al., 2016; Cavalier-Smith, 2018). The glycerolipid biosynthetic scheme seems to follow some rules deciphered in plants, i.e. (1) a de novo synthesis of FAs up to C16 in the stroma, generating 16:0-ACP; (2) the presence of enzymes involved in MGDG, DGDG, and SQDG biosynthesis probably located in plastid membranes; and (3) the presence of enzymes involved in phospholipid synthesis, such as PC and PE, probably located in the ER (Petroutsos et al., 2014; Abida et al., 2015; Dolch and Marechal, 2015; Dolch et al., 2017). Photosynthetic heterokonts contain an additional glycerolipid class, the betaine lipids, whose synthesis is usually associated with the ER. In P. tricornutum, the betaine lipid is diacylglyceryl-3-O-2′-(hydroxymethyl)-(N,N,N-trimethyl)-β-alanine (DGTA; Abida et al., 2015), whereas M. gaditana or N. oceanica contain diacylglyceryl-3-O-4′-(N,N,N-trimethyl)-homoserine (DGTS; Alboresi et al., 2016; Dolch et al., 2017; Murakami et al., 2018).

Besides these general features, essential differences make it extremely challenging to transfer knowledge acquired in plant or green algal models to Heterokonta:

• First, after de novo synthesis in the plastid, 16:0 is exported to the cytosol, where it is thioesterified into 16:0-CoA. This 16:0-CoA is the initial substrate for a series of elongations and desaturations forming at first 18:0-CoA and ending up with very-long-chain polyunsaturated FAs (VLC-PUFAs) in the ER, such as eicosapentaenoic acid (EPA or 20:5; Dolch et al., 2017). Since elongations operate on acyl-CoAs, whereas desaturations occur on acyl-glycerolipids, the 16:0-to-20:5 route requires five classes of enzymes, i.e. specific ACSs to supply elongases and acyl-transferases with activated FAs, desaturases acting on a “glycerolipidic platforms” harboring the acyl-group, where double bonds are added, and lipases A (Figure 1). In P. tricornutum and M. gaditana, based on the detection of 16:0-to-20:5 intermediates (mainly 18:0, 18:1, 18:2, 18:3, 20:3, and 20:4), the most likely glycerolipid platforms for 20:5 biosynthesis are PC and the betaine lipid (Abida et al., 2015; Dolch et al., 2017; Sayanova et al., 2017).

• Second, 16:0-ACP is presumed to be desaturated into 16:1-ACP in the plastid. Palmitoleic acid is exported to the cytosol, where it is thioesterified into 16:1-CoA. Palmitoleoyl-CoA is the preferred substrate for one of the four lysophosphatidic acyltransferases occurring in N. oceanica, i.e. LPAT1, residing in the ER and adding 16:1 at position sn-2 of membrane glycerolipids, mainly PC and DGTS (Nobusawa et al., 2017).

• Eventually, DGTS, MGDG, and DGDG are marked by an extremely high content in 20:5 (Botte et al., 2011; Abida et al., 2013; Alboresi et al., 2016; Dolch et al., 2017; Murakami et al., 2018). Genetic knock out of enzymes producing DGTS in N. oceanica showed that this betaine lipid was a pool of 20:5 in the cytosol, where 20:5 is actually synthesized (Murakami et al., 2018). To accumulate into MGDG and DGDG, 20:5 needs to relocate from the ER to the plastid, a process unlikely to be homolog of the “eukaryotic pathway” described in Arabidopsis (Arabidopsis thaliana). This import of 20:5 to the plastid of heterokonts has been termed the “omega pathway” (Petroutsos et al., 2014; Dolch et al., 2017). Based on the genetic knock out of the elongase Δ0-elo1 in M. gaditana, the omega pathway seems to operate via a large-scale multi-enzymatic and multi-site channeling machinery, starting from the initial elongation of 16:0-CoA into 18:0-CoA by a Δ0-ELO isoform in the ER, and ending by the formation of 20:5 and its transfer to MGDG in the plastid (Dolch et al., 2017).

In heterokonts, ACS isoforms are therefore expected to operate in FA fluxes at all these steps. We have focused our study on M. gaditana, previously known as Nannochloropsis gaditana (Fawley et al., 2015). Microchloropsis and Nannochloropsis are oleaginous microalgae, producing high levels of TAG and representing therefore a promising feedstock for biofuel and green chemistry, motivating important efforts into optimizing strains and culture conditions (Van Vooren et al., 2012; Camacho-Rodriguez et al., 2013; Chen et al., 2013; Camacho-Rodriguez et al., 2014, 2015; Ma et al., 2014; Meng et al., 2015) and developing genetic engineering methods (Radakovits et al., 2012; Vieler et al., 2012; Iwai et al., 2015; Kang et al., 2015; Dolch et al., 2017; Poliner et al., 2018). Microchloropsis gaditana is a model where C18 desaturation occurs mainly on PC, whereas C20 desaturations occur mainly on DGTS and PE (Alboresi et al., 2016). To our knowledge, no genetic study has focused on ACS genes in heterokonts. We addressed the role of an ACS isoform in M. gaditana belonging to the Bubblegum subfamily (ACSBG; Min and Benzer, 1999; Steinberg et al., 2000; Pei et al., 2003; Mashek et al., 2007; Song et al., 2007; Lopes-Marques et al., 2018) by genome engineering with tailored TALE-N.

Results and discussion

Identification of an ACS of the “Bubblegum” subfamily (ACSBG) in M. gaditana

We sought ACS genes that could be distinct between Heterokonta and Archaeplastida. The genome of M. gaditana contains six genes annotated as ACS or Acyl-CoA ligases (ACL): Naga_100014g59; Naga_100012g66; Naga_101051g1; Naga_100047g8; Naga_100649g1; and Naga_100035g43 (http://protists.ensembl.org/Nannochloropsis_gaditana_gca_000569095/). The genome of the model diatom P. tricornutum contains five putative ACS genes: Phatr3_J20143 (annotated as ACS1), Phatr3_J12420 (ACS2), Phatr3_J54151 (ACS3), Phatr3_J45510 (ACS4), and Phatr3_J17720 (ACL1; http://protists.ensembl.org/Phaeodactylum_tricornutum). In the current state of knowledge, it is not possible to predict the specificity of a given ACS sequence for FAs.

Based on protein sequences, subfamilies of eukaryote ACSs have been defined (Steinberg et al., 2000; Watkins et al., 2007). One subfamily corresponds to the “Bubblegum” ACSs (ACSBG), named after the bubbly appearance of the first optic ganglion induced by the mutation of the corresponding gene in the fruit fly Drosophila melanogaster, leading to a neurodegeneration and elevated level of VLC-FAs (Min and Benzer, 1999). ACSBGs are also called “lipidosins” (Song et al., 2007). ACSBGs have been mainly studied in Metazoa, conserved from invertebrates to vertebrates (Lopes-Marques et al., 2018). The impairment of ACSBG in Drosophila or mice leads to an increase of VLC-PUFAs; however, the beta-oxidation of VLC-PUFAs that could partly explain this increase seems not always affected in corresponding acsbg mutants (Min and Benzer, 1999; Steinberg et al., 2000; Song et al., 2007). The human (Homo sapiens) ACSBG1 gene is expressed primarily in the brain (Steinberg et al., 2000). Brains have higher concentrations of VLC-FAs than other tissues, including VLC-PUFAs like 20:5 and 22:6; these VLC-FAs are components of complex lipids such as gangliosides, cerebrosides, sulfatides, sphingomyelin, and other phospholipids, all kinds of lipids whose synthesis could also rely on ACSBG (Steinberg et al., 2000). The specific expression of ACSBG gene in animal brains as well as biochemical studies shows that ACSBG proteins are involved in the activation of PUFAs and VLC-PUFAs, with chain lengths from 16 to at least 24 carbons (Steinberg et al., 2000). Concerning the two genes in human, although ACSBG1 was initially thought to activate VLC-FAs based on overexpression studies (Steinberg et al., 2000), investigation of the endogenous enzyme using RNA interference highlighted a high specificity for C16:0 (Pei et al., 2003). By contrast, ACSBG2 preferentially activates C18:1 and C18:2 acids (Watkins et al., 2007). In chicken, a single nucleotide polymorphism in the ACSBG2 gene is associated with abdominal fat weight and percentage (Claire D’Andre et al., 2013). ACSBG proteins contain conserved domains, among which motif II is possibly involved in FA-binding (Moriya-Sato et al., 2000; Steinberg et al., 2000; Watkins et al., 2007; Lopes-Marques et al., 2018). Based on the conservation of motif II (Figure 2), we identified ACS isoforms of the bubblegum type in M. gaditana (MgACSBG; Naga_100014g59) and P. tricornutum (annotated as ACS4 and called here PtACSBG; Phatr3_J45510). MgACSBG and PtACSBG share five conserved ACSBG motifs (Lopes-Marques et al., 2018), further confirming their structural identification as ACSBG (Supplemental Figure S2).

Figure 2.

Motif conservation in ACSs of the Bubblegum family (ACSBGs) defined by Steinberg et al. (2000). Residues conserved in more than 50% of the sequences are shown in black boxes. This motif is based on a comprehensive comparison of ACS sequences in eukaryotes (Steinberg et al., 2000). Alignment includes sequences from fruit fly (D. melanogaster; dmBG AAF44848 and dmBG-H1 AAF44850), mouse (M. musculus; mmBG AA111606), human (H. sapiens; HsACBG1 and HsACBG2), chicken (G. gallus, GgACSBG1 and GgACSBG2), Michrochloropsis gaditana (MgACSBG or Naga_100014g59), and Phaedoctylum tricornutum (Phatr3_J45510).

We compared ACS sequences found in M. gaditana and P. tricornutum with sequences of previously characterized ACSBG in animals, i.e. D. melanogaster (DmACSBGa, NP_524698; DmACSBGc, NP_001285923); H. sapiens (HsACBG1, Q96GR; HsACBG2, Q5FVE4); Mus musculus (MmACSBG1, NP_444408); Gallus gallus (GgACSBG1, F1NLD6; GgACSBG2, XP_015155301). Our Bayesian phylogenetic analysis showed a robustly supported clade (posterior probability 1.00) grouping M. gaditana and P. tricornutum ACSBG with metazoan bubblegum/lipidosin sequences. The other algal ACS sequences grouped in separate clades (Figure 3).

Figure 3.

Phylogenetic comparison of ACSs from M. gaditana and P. tricornutum, with ACSBG sequences from Metazoa. All genes annotated as ACSs or ligases from M. gaditana (Naga_100014g59, MgACSBG/Naga_100014g59; Naga_100012g66; Naga_101051g1; Naga_100047g8; Naga_100649g1; and Naga_100035g43) and P. tricornutum (Phatr3_J20143, ACS1; Phatr3_J12420, ACS2; Phatr3_J54151, ACS3; Phatr3_J45510, PtACSBG or ACS4; and Phatr3_J17720, ACL1) were compared with ACSBG sequences from D. melanogaster (DmACSBGa, NP_524698; DmACSBGc, NP_001285923); H. sapiens (HsACBG1, Q96GR; HsACBG2, Q5FVE4); M. musculus (MmACSBG1, NP_444408); G. gallus (GgACSBG1, F1NLD6; GgACSBG2, XP_015155301). The tree was calculated using a Bayesian inference method (MrBayes) as described in the “Materials and methods” section, a Poisson model for amino acid substitutions, four Markov chain Monte Carlo (MCMC) chains run for 10,000 generations, sampling every 10 generations and a 50% majority rule consensus tree.

Using the MgACSBG sequence as query, and based on (1) similarity search using BlastP, with a Blosum 62 similarity matrix, default parameters and an e-value threshold of 1 × 10⁻¹⁰, and (2) detection of a conserved motif II, we could identify ACSBG homologs in both photosynthetic (e.g. Ectocarpus, CBJ33608 or Fistulifera, GAX09496) and non-photosynthetic (e.g. Pythium, GAX99969 or Phytophthora, XP_008898610) heterokonts. By contrast, we could not identify any sequence in Arabidopsis nor in any plant (Embryophyta). Likewise, no ACSBG homolog could be found in the green alga model Chlamydomonas reinhardtii. Nevertheless, some green algae, such as Ostreococcus tauri or Micromonas pusilla, contained ACSBG genes, i.e. XP_003083453 and XP_003061097, respectively. Since Ostreococcus differs from Chlamydomonas by its ability to produce high levels of VLC-PUFAs up to 22:6 (Degraeve-Guilbault et al., 2017), the presence of an ACSBG might be related to VLC-PUFA fluxes within the cell. We addressed the function of this class of heterokont ACSBG in M. gaditana.

TALE-N generation of M. gaditana acsbg mutants

We sought to introduce mutations in the MgACSBG gene using a specifically designed transcription activator-like effector nuclease (TALE-N), a method with high target binding specificity compared with other gene-editing systems (Malzahn et al., 2017) and with the possibility to generate different types of mutations, ranging from in-frame residue substitutions to gene interruption and knock out (Ma et al., 2016). The target site for TALE-N mutation in MgACSBG was selected by comparison with the whole genome to avoid multiple binding of the nuclease. The TALE-N target site was located at the beginning of the coding sequence to allow the generation of incomplete and inactive proteins. This region encodes a hydrophobic domain of the protein, which might be important for the correct folding of the protein, but distinct from ACSBG conserved motifs. Sequences coding for the two TALE-N-binding subunits (Tal2Ng-L and Talg2Ng-R) were designed as described in the “Materials and methods” section, to position the FokI DNA cleavage domain at the target site (Figure 4, A), and they were cloned in two independent vectors, pCT61 and pCT62, respectively. Following M. gaditana co-transformation with pCT61 and pCT62 and selection on zeocin, obtained colonies were first analyzed by treatment of PCR amplified target locus with a T7 endonuclease to detect cleaved DNA at mismatched positions and assess the occurrence of genome editing. To define precisely how the TALE-N modified the target DNA site, cells from T7-positive colonies were sub-cloned onto a new selective plate to separate cells harboring the mutation from those containing wild-type (WT)-like sequences.

Figure 4.

MgACSBG mutants. A, Mutant selection. Following M. gaditana co-transformation with the two specifically designed TALE-N subunits, obtained clones were analyzed by treatment with a T7 endonuclease to detect cleaved DNA at mismatched positions and assess occurrence of genome editing at the target locus. When this test allowed detecting a mutation, the cell population was sub-cloned onto a new selective plate, and colonies arising from single cells were sequenced at the MgACSBG inside the FokI activity region (frame), resulting in the appearance of novel codons. Three mutant lines harboring mutated endogenous MgACSBG are shown, called here MgACSBG#5, MgACSBG#31, and MgACSBG#40. Only two to three amino acids were modified in the endogenous proteins. B, Point mutations in MgACSBG#5, MgACSBG#31, and MgACSBG#40. In gray boxes, WT residues at the TALE-N nucleotidic target; in black boxes, inserted mutations.

Seven T7-positive lines were obtained from multiple independent co-transformations with the pCT61 and pCT62 plasmids. No disruption of the MgACSBG full length coding sequence could be obtained, suggesting that knockout (KO) might be lethal. Only nucleotidic modifications leading to in-frame amino acid substitutions could be obtained (Figure 4, A). In some of the M. gaditana-transformed lines, e.g. MgACSBG#5, MgACSBG#31, and MgACSBG#40, TALE-N activity led to in-frame insertions/deletions and introduction of non-synonymous codons, leading to point mutant proteins (Figure 4, B).

In vivo impact of MgACSBG#5 and MgACSBG#31 in-frame mutations on glycerolipid profiles

We focused our phenotype analyses on two in-frame mutations, MgACSBG#5 and MgACSBG#31. Cells were cultivated in parallel in a nutrient-rich liquid medium, using a Multicultivator photobioreactor, supplied with CO₂ as described in the “Materials and methods” section. Control strains consisted of untransformed cells (WT) and cells transformed with an empty vector (EV). Experiments were performed in duplicate for each line and were repeated to obtain data from independent replicates (n = 4). The MgACSBG#5 and MgACSBG#31 lines showed a slower growth compared with WT and EV strains, monitored by cell counting at days 3, 7, and 12 following inoculation (D3, D7, and D12, respectively; Figure 5, A). Compared with WT and EV strains, the total FA content per cell was higher in the MgACSBG#31 mutant in the early stage of cultivation (D3; Figure 5, B), reflecting a higher content in TAG (Figure 5, C). WT and EV strains showed an accumulation of TAG at D7 and D12 induced by the shortage of nutrients in the medium, as described earlier (Simionato et al., 2013; Alboresi et al., 2016; Figure 5, C). MgACSBG#5 and MgACSBG#31 mutants also showed increased TAG content at D7 and D12, but the difference with WT and EV strains was less pronounced, probably due to the nitrogen starvation.

Figure 5.

Compared analysis of MgACSBG#5 and MgACSBG#31 with untransformed WT cells and cells transformed with an EV. A, Growth was measured by cell counting as described in the “Materials and methods” section at days 3, 7, and 12 following inoculation (D3, D7, and D12, respectively). D3 corresponds to nutrient replete cultivation, whereas D7 and D12 correspond to nitrogen-starved condition. B, FA content per million cells at D3, D7, and D12 following inoculation. C, TAG content, in mol% of total glycerolipids, at D3, D7, and D12 following inoculation. D, FA profile at D3. Total FAs were extracted and, following methanolysis, obtained FAMEs were analyzed and quantified by gas chromatography coupled with ion flame detection (GC-FID), as described in the “Materials and methods” section. E, Glycerolipid profile at D3. Each glycerolipid class was analyzed by liquid chromatography coupled to tandem mass spectrometry (LC–MSMS) and quantified as described in the Materials and methods” section. DGTS, diacylglyceryl-N,N,N-trimethylhomoserine. Data are the average of four replicates. Error bars, standard deviation. (*), P-value < 0.05; Student’s t test using WT as a reference.

We focused our comparison of MgACSBG#5 and MgACSBG#31 with WT and EV cells at D3, when no starvation of nutrients occurs. The total FA profile of mutant lines showed an increase in 16:0 proportion balanced by a slight decrease in 16:1 in MgACSBG#31 and a significant decrease of 20:5 in one of the mutants (Figure 5, D). The glycerolipid profile highlighted an increase in TAG and little change in the level of other glycerolipid classes (Figure 5, E). Since in M. gaditana TAG are 16:0-rich and 20:5-poor (Simionato et al., 2013; Alboresi et al., 2016), one could consider that the global increase in 16:0 and the decrease in 20:5 in the mutant lines might simply reflect the accumulation of TAG. We addressed this question by comparing the FA profile of each glycerolipid class at D3.

The comparison of FA profiles in glycerolipid classes in the ACSBG mutants with WT or EV strains showed a striking impact on PC, DGTS, and PE (Figure 6, A–C), synthesized in the ER of M. gaditana, and on two of the plastid lipids, MGDG and DGDG (Figure 6, F and G), whereas little or no significant change could be observed on PG and SQDG (Figure 6, D and E). In LC-MSMS analyses, m/z spectra correspond to diacyls indicated by the sum of carbons and the number of double bonds (Alboresi et al., 2016), e.g. in PC, 34-2 corresponds to the major molecular species 16:0/18:2.

Figure 6.

Comparison of the acyl profiles of glycerolipid classes in MgACSBG#5, MgACSBG#31, untransformed WT cells, and cells transformed with an EV. A, PC. B, Diacylglyceryl-N,N,N-trimethylhomoserine, DGTS. C, PE. D, PG. E, SQDG. F, MGDG. G, DGDG. H, TAG. Major molecular species corresponding to combined mass given in number of carbons and of double bonds are indicated, based on previous regioselective analysis of M. gaditana glycerolipids (Alboresi et al., 2016). Molecular species of which the proportion increases in the membrane lipids of MgACSBG#5 and MgACSBG#31 are shown in red, whereas those of which the proportion decreases are shown in green. Data are the average of four replicates and are given in mol%. Error bars, standard deviation. (*), P-value < 0.05; Student’s t test using WT as a reference.

We first noticed in the MgACSBG#5 and MgACSBG#31 mutants a decrease of molecular species containing a 16:1 at sn-2 position in PC (e.g. 16:0/16:1; 16:1/16:1) and DGTS (e.g. 14:0/16:1; 16:1/16:1; Figure 6, A–C). This could reflect a role of ACSBG in producing 16:1-CoA for the de novo synthesis of lipids at the ER, reminiscent of the effect of the KO of LPAT in the closely related species N. oceanica, an enzyme shown to transfer 16:1 at the sn-2 position of these lipids (Nobusawa et al., 2017).

These lipids are presumed to also serve as “platforms” for the production of 20:5, i.e. PC from 16:0 to 18:3 and DGTS from 20:3 to 20:5. This role is not fully restricted to PC and DGTS: it was shown in N. oceanica that mutants lacking DGTS lead to the accumulation of 18:1, 18:2, and 20:4 in PC, consistently with an accumulation of 20:5 precursors in this lipid, and more 20:4 and 20:5 in PE, consistently with PE acting as a secondary platform for the final steps of 20:5 production (Murakami et al., 2018). Here, in MgACSBG#5 and MgACSBG#31, PC molecular species containing 18:1 and 18:2 (16:0/18:1; 16:0/18:2; 16:1/18:2) were enriched, corresponding to an accumulation of 18:3 precursors (Figure 6, A and Supplemental Table S1). We had a closer look at PC acyl profiles and could detect 18:3 only in MgACSBG#5 and MgACSBG#31 (16:1/18:3, 18:2-18:3, and 18:1/18:3; Supplemental Table S1). In these mutants, DGTS had more diacyls with 16:0 and 18:1 (16:0/16:0; 18:1/16:0; 20:4/16:0; 20:5/16:0) and fewer diacyls containing only 20-carbon FAs (20:4/20:4; 20:5/20:4; 20:5/20:5; Figure 6, B). PE was also altered, containing less 20:5 (16:1/20:5; 20:5/20:5) and higher levels of 20:3 and 20:4 (20:4/20:3; 20:5/20:5; Figure 6, C). In the MgACSBG#5 and MgACSBG#31 mutants, 18-carbon precursors of 18:3 accumulated in PC (18:0; 18:1; 18:2; and 18:3), and 20:5 proportions decreased in DGTS and PE. This could reflect a role of ACSBG in producing 18:3-CoA in the pathway producing 20:5 in DGTS and to some extent in PE.

The profile of MGDG in the MgACSBG#5 and MgACSBG#31 mutants is characterized, like in the WT, by a major molecular species with a m/z corresponding to 34 carbons and 5 double bonds (34-5 or 20:5/14:0). Little change was observed, besides a decrease of molecular species enriched in 16:1 (16:1/16:0; 16:1/16:1) and a slight increase in 20:5-rich molecular species (20:5/16:0; 20:5/20:5; Figure 6, F). DGDG showed the same trend but with a higher magnitude, i.e. a dramatic decrease in the proportion of 16:1 (16:1/16:0; 16:1/16:1) compensated by an increase in 20:5 (20:5/14:0, 20:5/16:0; 20:5/16:1; Figure 6, G). Since the amount of plastid glycerolipids was either unchanged or slightly decreased in the mutants, the observed phenotype in the MgACSBG#5 and MgACSBG#31 mutants reflected a depletion of 16-carbon FAs, mainly 16:1, partly compensated by imported 20:5.

The profile of TAG in the MgACSBG#5 and MgACSBG#31 mutants was also different from that in the WT and EV lines. It was globally characterized by a decrease in 16:1-rich molecular species (14:1/16:1/16:1; 16:1/16:1/16:0; 16:1/16:1/16:1; 16:1/18:1/16:1) and a higher level of 16:0-rich molecular species (16:0/16:0/16:1; 16:0/16:0/16:0) as well as a noticeable increase in 18:0 and 18:1 (16:1/16:0/18:0 and 16:1/16:0/18:1; Figure 6, H).

Altogether, these data indicate that a point mutation in the MgACSBG sequence leads to a modification of FA partitioning within M. gaditana cells, altering two pathways, one being the incorporation of 16:1 at the sn-2 position of ER glycerolipids via the LPAT1 pathway dissected in N. oceanica (Nobusawa et al., 2017), another being the 16:0-to-20:5 elongation/desaturation process occurring on the PC and DGTS platforms, both leading to an accumulation of 16:0 in ER lipids and TAG and an exhaustion of 16-carbon from plastid glycerolipids.

The lack of KO mutant after TALE-N genetic mutation of MgACSBG suggests that this acyl-CoA transferase operating in two key processes is critical. A possible interpretation is shown in Figure 7, following FA biosynthesis in the plastid (Figure 7, A), with a first role for ACSBG in the production of 16:1-CoA in the cytosol (Figure 7, B), prior to incorporation of 16:1 in ER lipids via the LPAT1 pathway (Nobusawa et al., 2017; Figure 7, C). 16:0-CoA also entered the 16:0-to-20:5 channeling process, following a first elongation via a Δ0-ELO producing 18:0-CoA (Figure 7, D). A second role for ACSBG in the production of 18:3-CoA was also deduced from the mutants’ phenotype in PC, DGTS, and PE (Figure 7, E and F). As a result of ACSBG impairment, an over-accumulation of 16:0-CoA was expected and observed (Figure 7, G). An important assumption for this model is that a positive feedback is exerted on the plastid, exporting most of its FAs, and explaining that the plastid is depleted of its 16-carbon FAs and that the omega pathway is activated to provide sufficient 20:5 compensating the lack of 16-carbon substrates in plastid glycerolipids (Figure 7, H). In our scheme, the omega pathway appears important for the synthesis of both MGDG and DGDG. The absence of any viable knock-out mutant indicates that no other ACS isoforms from M. gaditana could apparently compensate a lack of functional MgACSBG.

Figure 7.

Model for ACSBG role in M. gaditana. The scheme shows the major FAs in major glycerolipid classes, i.e. PC, DGTS, PE, and TAG in the ER, and MGDG in the plastid. Each membrane glycerolipid class is shown as a gray block. TAG is shown in a gray circle. A, De novo synthesis of FAs in the stroma of the plastid. The action of FA synthetase of type II (FAS II) leads to the production of 16:0-ACP, which can be desaturated by a palmitoyl ACP desaturase (PAD) into 16:1-ACP. These acyls can be exported to the cytosol, possibly as FFAs, through the four limiting membranes of the plastid. B, ACSBG phenotype is consistent with a role in the production of 16:1-CoA. Unknown ACS isoforms (ACSx) catalyze the production of 16:0-CoA. C, The 16:1 is incorporated at the sn-2 position of ER lipids via LPAT1 in N. oceanica (Nobusawa et al., 2017). Impairment of ACSBG alters this process. D, The 16:0-CoA is elongated into 18:0-CoA by the action of D0-ELO isoforms, as a committing step in the 16:0-to-20:5 synthesis pathway (Dolch et al., 2017). E, VLC-PUFA synthesis, starting with the 18:0-to-18:3 desaturation on PC by the stepwise action of an ERD9FAD, an ERD12FAD, and an ERD6FAD. Since PC accumulates 18 precursors (18:1 and 18:2), we consider that the mutants are likely impaired downstream, at the level of 18:3 elongation, leading therefore to the positioning of ACSBG at the level of 18:3-CoA production, feeding the Δ6-ELO isoform producing 20:3-CoA. Downstream processes, i.e. 20:3-to-20:5 desaturations occurring on DGTS would be consistently impaired, with a decrease in the end-product, 20:5. F, Some 20:3-to-20:5 desaturation processes occurring in DGTS operate also in PE. The slowing down of 20:5 synthesis also occurs in PE. G, As a result, following the mass action law, the slowing down occurring in the 16:0-to-20:5 pathway is expected to lead to an over-accumulation of 16:0-CoA. Excess 16:0-CoA can be stored as TAG, via the DGAT/Kennedy pathway. H, An important assumption for this model is that a positive feedback is exerted on the plastid, exporting most its FAs, and explaining that the omega pathway is activated to provide sufficient 20:5 to plastid glycerolipids to compensate the lack of 16-carbon substrates. AAS, acyl-ACP synthetase; DGAT, DAG acyltransferase; DGTS, diacylglyceryl-N,N,N-trimethylhomoserine; FAD, FA desaturase.

Impact of MgACSBG#5 and MgACSBG#31 mutations on acyl-CoA profiles

Acyl-CoAs from WT, EV, and mutant cells grown in nutrient replete condition were extracted and sent to Rothamsted Research, UK, for a targeted analysis of chain lengths from 10 to 20 carbons (using 10:0; 12:0; 14:0; 16:0; 16:1; 16:3; 18:0; 18:1; 18:2; 18:3; 18:4; 18:5; 20:0; 20:3; 20:4; and 20:5 standards), as described in the “Materials and methods” section (Figure 8). The level of only three acyl-CoAs showed a significant variation in mutants compared with WT, i.e. an increase in 16:0-CoA and a decrease in 16:1-CoA and 18:3-CoA. This analysis is consistent with a dual role of ACSBG in the activation of 16:1 and 18:3 into 16:1-CoA and 18:3-CoA. We sought to better comprehend the molecular impact of the mutation introduced in the MgACSBG protein.

Figure 8.

Comparison of the acyl-CoA profiles in MgACSBG#5, MgACSBG#31, untransformed WT cells, and cells transformed with an EV. Cells were cultivated in nutrient replete conditions as described in the “Materials and methods” section (D3). Mass-spectrometry-based analysis and data processing were performed by Rothamsted Research, UK. Data are the average of two replicates. Error bars, standard deviation. (*), P-value < 0.05; Student’s t test using WT as a reference.

Molecular impact of MgACSBG#5 and MgACSBG#31 mutations on the protein structure

The MgACSBG sequence contains 649 amino acids and is predicted to be soluble with a MW of 71,056 Da. We sought to construct a 3-D structural model that could in part provide clues to understand the effects of the mutations at the level of the IGF domain (a triad defined by isoleucine 96, glycine 97, and phenylalanine 98). It seemed essential to propose the positioning of a FA chain in the model, i.e. α-linolenic acid (18:3; C₁₈H₃₀O₂ cis-Δ9,12,15), as a substrate for MgACSBG deduced from phenotypic analyses. The search of conserved ACS motifs was refined via the online servers MOTIF (Kanehisa et al., 2002) and PROSITE (Sigrist et al., 2013), which identified an AMP binding motif, while superfamily (Wilson et al., 2009) confirmed that MgACSBG belonged to an acetyl-CoA synthetase-like family. The Porter predicting method (Mirabello and Pollastri, 2013) found 44% coil, 37% helix, and 19% extended conformation while Predictprotein (Yachdav et al., 2014) gave the secondary structure prediction shown in Supplemental Figure S3.

The sequence alignment of MgACSBG with four proteins used as templates in our models (see below) is shown in Supplemental Figure S4. The most clearly conserved motif YTSGTTGPPK (residues 216–225), identified as the AMP-binding motif, was requalified as playing a fundamental role in ACS catalytic activity (Gulick et al., 2003; Lopes-Marques et al., 2018). The long sequence SITGRIKELIITAGGENIPPVLIE (residues 492–515), or “motif II,” contains a part of the “FACS motif” (fatty acyl CoA synthetase signature motif; Black et al., 1997) and is supposed to be involved in acyl chain length specificity. In ACSBGs, a so-called “motif III” (F/YG-SE; residues 409–413) has been proposed as being involved in AMP binding, whereas “Motif IV” (LPLSH; residues 259–263) could be involved in CoA binding (Lopes-Marques et al., 2018).

No structure is currently available for an ACSBG in public databases. The sequence was first submitted to Blast (Altschul et al., 1990) to find sequence alignments with proteins of known 3-D structure. The first hit was referenced 1ULT in the Protein Data Bank (PDB, Burley et al., 2019) and corresponds to a long chain fatty ACS homodimer from Thermus thermophilus (Hisanaga et al., 2004), with 27% sequence identity and 40% homology. Then the sequence was submitted to several online prediction servers (Robetta [Song et al., 2013]; RaptorX [Kallberg et al., 2012]; Swiss-Model [Waterhouse et al., 2018]; PhyRe2 [Kelley et al., 2015]; I-Tasser [Zhang, 2009]; PsiPred [Buchan et al., 2013]; and PS(2) [Huang et al., 2015]). All methods failed to predict the full MgACSBG protein and most of them found 1PG4 (acetyl-CoA synthetase from Salmonella enterica; Gulick et al., 2003), and 5MSC (the A domain of carboxylic acid reductase from Nocardia iowensis; Gahloth et al., 2017) as templates. Despite its low homology with MgACSBG, another ACS with known 3-D structure and for which both AMP and CoA are resolved was chosen as template: 3EQ6, a human ACS medium-chain family member (Kochan et al., 2009). The sequences of these four template proteins were aligned with MgACSBG in Supplemental Figure S4.

Three-dimensional structural homology models of the full length MgACSBG protein were built with Modeller (Sali and Blundell, 1993) using different association of template proteins: M1pg4 from 1PG4 alone, M3eq6 from 3EQ6 alone, and two consensus models Mcons1 from 1PG4, 1ULT, and 5MSC, and Mcons2 from all four protein templates (Supplemental Models S1–S4). The positioning of CoA and AMP in the MgACSBG models was based on their neighboring residues in structure 1PG4 and 3EQ6, also identified in animal ACSBG2 (Steinberg et al., 2000; Supplemental Table S2) as described in the “Materials and methods” section. The positioning of 18:3 was placed inside the structures, with the carboxy-terminal “head” in the direction of AMP, respecting its proximity with CoA and profiting of hydrophobic holes observed in the structures.

Three models of MgACSBG, M1pg4, M3eq6, and Mcons1, have similar final energy and correctly respect the restraints. From model Mcons1 of MgACSBG (Figure 9), the proximity of the IGF triad, mutated in this study, to the inserted 18:3 FA is clear. Their relative position is globally the same in all models. Our modeling supports therefore the hypothesis that the mutation could affect the accessibility of the FA to the reaction site rather than interfere with the reaction itself through proximity with AMP or CoA.

Figure 9.

3-D model of MgACSBG bound to CoA, AMP, and α-linolenic acid (18:3). Four models of MgACSBG were obtained based on similarity with acetyl- and ACS of known 3-D structure, as described in the “Materials and methods” section. The Mcons1 model of MgACSBG is shown with the visualization program VMD (Humphrey et al., 1996). The backbone of residues around AMP and CoA, listed in Supplemental Table S2, A and B, are shown in licorice. The IGF triad is represented in pink (I96, G97, F98). The conserved Y216-K225 active site is shown in licorice mode. The position of IGF close to 18:3 is similar in all four MgACSBG models. COA, Coenzyme A; AMP, adenosine monophosphate; LIP, α-linolenic acid.

MgACSBG but not MgACSBG#5 or MgACSBG#31 partially complement a yeast defective in acyl-CoA synthase function

We sought to decipher whether MgACSBG and its two variants could be active after heterologous expression in the yeast Saccharomyces cerevisiae and complement the phenotype of a faa1Δfaa4Δ double mutation. In yeast, acyl-CoA can originate from neo-synthesis by a fatty acyl synthase (FAS) or by the activation of exogenous FAs by ACSs (Black and DiRusso, 2007). Faa1 and Faa4 are the major enzymes involved in the formation of long chain fatty acyl-CoA (Johnson et al., 1994a). In normal growth condition, the faa1Δfaa4Δ mutant does not show any growth defect. However, when yeast FAS is inactivated by the addition of cerulenin, the mutant strain cannot grow, even in the presence of exogenously supplied FAs, due to a low level of ACS activity (less than 5% of WT activity; Johnson et al., 1994a; Faergeman et al., 2001). MgACSBG, MgACSBG#5, and MgACSBG#31 were expressed in the yeast faa1Δfaa4Δ mutant from a high copy vector under the galactose inducible pGAL1 promoter. The WT and faa1Δfaa4Δ strains transformed with the EV were used as controls. Control strains will be henceforth referred to as “yeast WT” and “faa1Δfaa4Δ mutant,” respectively, for simplicity. All the strains were grown on synthetic media deprived of uracil (-U) and supplemented with galactose 2% for 6 h to induce expression of the transgenes prior to dilution to OD₆₀₀ = 0.3. Growth curves were then performed at 30°C. All the strains used in the complementation experiments were tested for expression of MgACSBG variants by immunoblot (Supplemental Figure S5). All strains presented an exponential growth in synthetic medium with galactose but did not grow in the presence of 22.5 µM cerulenin, as expected (Supplemental Figure S5).

In a first set of experiments, we tested the ability of a series of FAs, namely 14:0, 16:0, 18:0, 18:1, 18:2, and 18:3, to rescue the growth of the yeast WT in presence of cerulenin. In our conditions, 14:0 and 16:0 FAs were able to almost fully complement the growth defect induced by cerulenin, whereas 18:0 and 18:1 only partially did, as previously observed by others (Figure 10, A; Johnson et al., 1994a, 1994b). Conversely, 18:2 and 18:3 FAs were not able to rescue the growth of yeast WT in presence of cerulenin, likely because these two FAs are not naturally present in yeast.

We next investigated the ability of MgACSBG native sequence to complement the faa1Δfaa4Δ strain in presence of cerulenin with the same set of FAs previously tested. Three independent clones were analyzed. The results presented in Figure 10, B show that MgACSBG partially rescued the growth phenotype of the double mutant in presence of C14:0 and C16:0. Altogether, these results provide evidence that MgACSBG has indeed an ACS activity, is able to use at least 14:0 or 16:0 but cannot provide more information on the specificity for other FAs. MgACSBG-transformed yeast have the same order of apparent preference for 14:0 and 16:0 as that observed in the WT yeast (Figure 10, A), thus indicating that this method is not suitable to address the question of the substrate specificity of MgACSBG at the molecular level.

Figure 10.

Growth complementation of a yeast mutant defective in acyl-CoA activity by MgACSBG variants. A, WT yeast strain growth curves in galactose medium (Cer-, green circles, positive control), with the addition of 22.5 µM cerulenin only (Cer+, gray circles, negative control), or with the addition of 22.5 µM cerulenin and 100 µM of the following FFAs, 14:0, 16:0, 18:0, 18:1, 18:2, and 18:3 (curves ranging from blue to red). B, Growth curves of MgACSBG complemented faa1Δfaa4Δ double yeast mutant in galactose medium supplemented by 22.5 µM cerulenin and 100 µM of FA (curves ranging from blue to red). Medium with 22.5 µM cerulenin but FA was used as a negative control (Cer+, gray circles). Average (±standard deviation) of biological (n = 3, independent clones) as well as technical (n = 3) replicates are presented. C and D, Growth curves of the faa1Δfaa4Δ double mutant complemented with MgACSBG (green squares), MgACSBG#5 (blue squares), and MgACSBG#31 (purple squares) in presence of 22.5 µM cerulenin and 100 µM 14:0 (C) or 16:0 (D). The faa1Δfaa4Δ double mutant without complementation was used as negative control and represented by gray circles. Average (±standard deviation) of biological (n = 3, independent clones) replicates are presented.

In the light of the results presented above, 14:0 and 16:0 were retained for further investigation. The same experiments were performed with three independent clones expressing MgACSBG#5 and MgACSBG#31 in a faa1Δfaa4Δ background in presence of cerulenin and 14:0 or 16:0. Neither MgACSBG#5 nor MgACSBG#31 (Figure 10, C and D and Supplemental Figure S5, C–E) were able to complement the growth phenotype of the mutant strains in presence of 14:0 and 16:0, unlike MgACSBG. Altogether, these results indicate that the TALE-N-generated mutations interfered with the ACS activity of MgACSBG, and that this decline in activity was related with the observed phenotype in M. gaditana.

Conclusion and perspectives

ACS enzymes generate acyl-CoA pools at different locations of the cell, of different FA profiles and for specific acylation reactions. They are therefore key actors of FA partitioning in lipid metabolism. Unraveling an ACS specific role is made difficult by the possibility of multiple isoforms to compensate a defective enzyme. In photosynthetic eukaryotes, ACS is therefore expected to operate in important acyl fluxes, most importantly from the plastid to the cytosol and vice versa. Heterokonta are marked by the production of VLC-PUFAs in the cytosol, rapidly transferred to the secondary plastid by a channeled processed called the omega pathway. ACS isoforms are expected to operate at “strategic” levels in this metabolon, in particular after release of 18:3-FFA following acyl desaturation on PC, to generate 18:3-CoA for elongation into 20:3-CoA, precursor for further desaturations on DGTS and/or PE to generate EPA, 20:5. Our functional analyses of the M. gaditana Bubblegum ACS showed that this enzyme was likely vital, since no KO could be obtained, and that a functional defect impaired two critical steps of acyl-lipid metabolism, the activation of palmitoleic acid, generating 16:1-CoA used to acylate the sn-2 position of ER lipid, and of linolenic acid, generating 18:3-CoA used as intermediate in the 16:0-to-20:5 synthetic process. The obtained mutants using TALE-N editing still contain 20:5, probably because the level of this VLC-PUFA cannot fall below a critical threshold necessary for M. gaditana survival. Our results do not exclude that the heterokont ACSBG also acts in other cellular locations, for other purposes. In a recent study, the P. tricornutum homolog, PtACSBG/ACS4, was found in the proteome of the lipid droplet (Lupette et al., 2019), indicating a possible role in the acyl-CoA pool closely associated to this cytosolic organelle. Future works will include the study of this enzyme’s location within M. gaditana, which is a challenging task given the small size of this microalga. The search of protein partners is also an important perspective, to evaluate the possible association with LPAT1 acting on ER lyso-lipids, or other components of the machinery converting 16:0 into 20:5 (such as Δ0-ELOs, ERΔFAD9, ERΔFAD12, a PC-specific phospholipase A, Δ5-ELOs, ERΔ5FAD, ERω3FAD, a DGTS/PE-specific lipase A, and subunits of a still-to be identified carrier/transporter importing a 20:5-precursor inside the secondary plastid), or any other protein component that could further highlight the full range of function of this type ACS in a photosynthetic eukaryote.

Materials and methods

MgACSBG phylogenetic analysis

All genes annotated as ACSs or ligases from M. gaditana (Naga_100014g59, MgACSBG; Naga_100012g66; Naga_101051g1; Naga_100047g8; Naga_100649g1; and Naga_100035g43) and P. tricornutum (Phatr3_J20143, ACS1; Phatr3_J12420, ACS2; Phatr3_J54151, ACS3; Phatr3_J45510, PtACSBG or ACS4; and Phatr3_J17720, ACL1) were retrieved from public databases, http://protists.ensembl.org/Nannochloropsis_gaditana_gca_000569095/ and http://protists.ensembl.org/Phaeodactylum_tricornutum, respectively. These sequences were compared with ACSBG sequences from Metazoa, i.e. D. melanogaster (DmACSBGa, NP_524698; DmACSBGc, NP_001285923); H. sapiens (HsACBG1, Q96GR2; HsACBG2, Q5FVE4); M. musculus (MmACSBG1, NP_444408); and G. gallus (GgACSBG1, F1NLD6; GgACSBG2, XP_015155301). A phylogenetic tree was reconstructed after a multiple alignment of sequences using the MUSCLE method (v3.8.31) configured for high accuracy with default settings (Edgar, 2004), removal of gaps and poorly aligned sequences using Gblock (Castresana, 2000). The tree was calculated using the phylogeny.fr internet platform (Dereeper et al., 2008) using a Bayesian inference method implemented in the MrBayes program (v3.2.6), with a number of substitution types fixed to 6, a Poisson model for amino acid substitutions, four Markov chain Monte Carlo chains run for 10,000 generations, sampling every 10 generations, discard of the first 250 trees sampled, and a 50% majority rule consensus tree (Huelsenbeck and Ronquist, 2001; Ronquist et al., 2012).

Cultivation of M. gaditana cells

Microchloropsis gaditana strain CCMP526 was used in all experiments. For nuclear transformation, cells were grown under constant light in f/2 medium (Dolch et al., 2017) until they reached the late exponential phase. All cultures were maintained on f/2 plates solidified with 1% agar, w/v, under a 12:12 light/dark regime in presence (transformed strains) or absence (WT strain) of the selective antibiotic zeocin (7 μg·mL⁻¹). For phenotypic analyses, M. gaditana was cultivated in artificial seawater (ESAW) using 10 times enriched nitrogen and phosphate sources (5.49 × 10⁻³ M NaNO₃ and 2.24 × 10⁻⁴ NaH₃PO₄; called “10× ESAW”; Dolch et al., 2017). When grown in batch in 250 mL flasks, a volume of 50 mL liquid medium was inoculated at 2.5 × 10⁶ cells·mL⁻¹ and cultivation was achieved under gentle agitation (100 RPM) at 20°C on a 12:12 light (60 µmol photons·m⁻²·s⁻¹)/dark cycle. When grown with a controlled CO₂ supply, a volume of 80 mL liquid medium was inoculated at 2.5 × 10⁶ cells mL⁻¹ and cultivation was achieved in small photobioreactors (Multi-Cultivator MC 1000, Photon Systems Instruments, Czech Republic) at 24°C, in continuous light (illumination at 60 µmol photons·m⁻²·s⁻¹). Culture mixing was achieved by gas bubbling in which CO₂ levels were maintained constant at 0.5% as in air-lift photobioreactors. Precise and constant level of CO₂ was supplied by a Gas Mixing System GMS 150 (Photon Systems Instruments, Czech Republic) following manufacturer’s instructions. In a multicutivator system, growth was conducted in nutrient replete medium until day 3, then transferred to same medium without nitrogen to investigate possible impact on low nitrogen response. When needed, cells were counted using a LUNA Automated Cell Counter following manufacturer’s instructions. For lipidomic analyses, cells were harvested by centrifugation at 3,500 × g for 10 min. Cells were then immediately frozen in liquid nitrogen and stored at −80°C, until use.

Design of vectors containing MgACSBG-specific TALE-N subunits

TALE-Ns consisted of restriction enzymes made of two subunits, called here Tal2Ng-R and Talg2Ng-L, engineered by fusing a TAL effector DNA-binding domain to a FokI DNA cleavage domain. The DNA binding domain contained a repeated highly conserved amino acid sequence with divergent amino acids, referred to as the repeat variable diresidue (RVD). The Tal2Ng-L and Tal2Ng-R RVDs were purchased from ThermoFisher GeneArt-TALs and synthetized into their commercial vector. Tal2Ng-L subunit was designed following established guidelines (Sanjana et al., 2012) to bind to the TCCTGCGCATGGTGCCATT sequence in the MgACSB gene (Naga_100014g59), corresponding to RVD (T)-HD-HD-NG-NN-HD-NNHD-NI-NG-NN-NN-NG-NN-HD-HD-NI-NG-NG, whereas the Tal2Ng-R subunit was designed to bind specifically to TGGCAATGAGCCATTCGGG, corresponding to RVD (T)-NN-NN-HD-NI-NI-NGNN-NI-NN-HD-HD-NI-NG-NG-HD-NN-NN-NN. In these RVD sequences, “(T)” indicates that the first binding repeat is provided by the vector. Part of the Tal2Ng-L and Tal2Ng-R commercial sequences (comprising RVDs) were subcloned in two distinct home-designed vectors pCT5Ng and pCT6Ng to give pCT61 and pCT62, respectively. pCT5Ng is a pET15b backbone which contains a first part with the bleomycin/zeomycin-resistance protein from Streptoalloteichus hindustanus (ShBle) under the control of the UEP and with the fcpA terminator and a second part with a codon-adapted nuclear localization sequence (NLS) DYKDHDGDYKDHDIDYKDDDDKMAPKKKRKVGIHGVPAA (Sanjana et al., 2012), a codon-adapted HA-tag, N-ter sequence of commercial TAL up to AflII site, C-ter sequence of commercial TAL from XhoI site under the control of the endogenous ubiquitin extension promoter (UEP) and with the fcpA terminator. pCT6Ng is a pET28b backbone which contains a first part with the bleomycin/zeomycin-resistance protein from S. hindustanus (ShBle) under the control of the UEP and with the fcpA terminator and a second part with a codon-adapted NLS DYKDHDGDYKDHDIDYKDDDDKMAPKKKRKVGIHGVPAA (Sanjana et al., 2012), a codon-adapted HA-tag, N-ter sequence of commercial TAL up to AflII site, C-ter sequence of commercial TAL from XhoI site under the control of the endogenous UEP and with the fcpA terminator. Codon optimization for expression in M. gaditana was done using the Kazusa webtool (http://www.kazusa.or.jp/codon/).

Nuclear transformation of M. gaditana

Plasmids pCT61 and pCT62 were linearized by digestion with ScaI and column-purified using a NucleoSpin gel and polymerase chain reaction (PCR) clean-up kit (Macherey-Nagel) following manufacturer’s instructions. Two micrograms (1 μg TALE-N left subunit plasmid + 1 μg TALE-N right subunit plasmid) linearized plasmids were electroporated into M. gaditana as previously described (Dolch et al., 2017). Transformed lines were selected on f/2 plates containing 7 µg mL⁻¹ zeocin. Integration of both TALE-N subunits was assessed by colony PCR with primers TAL-Nterm-Rev (GCAGGTCGCTAAAAGAATCG) and TAL-HA Fw (CCCCGACTACGCTAGCG) for TALEN left subunit and TAL-Nterm-Rev (GCAGGTCGCTAAAAGAATCG) and TAL-His Fw (CACCACCACCACCACAGC) for TALEN right subunit.

T7 endonuclease test and sequencing of TALE-N genetic targets

Following M. gaditana co-transformation with pCT61 and pCT62 and selection on zeocin, obtained clones harboring both TALE-N subunits were first analyzed by treatment with a T7 endonuclease (T7E1, New England Biolabs) to cleave DNA at mismatched positions and assess occurrence of genome editing at the target locus. To this aim, a genomic fragment of around 396 bp in the MgACSB gene (Naga_100014g59), where the FokI activity region is placed asymmetrically from the 5′ and 3′-ends, was PCR-amplified from positive colonies, using a proofreading polymerase. After electrophoresis of amplified DNA on a 1% agarose gel, bands corresponding to the desired products were purified and quantified. For each clone, 1.3 µg of PCR product was treated with T7E1 (+) or left untreated (−). A non-transformed WT colony was also included as a negative control. T7E1 reaction was performed following manufacturer’s instruction, with the following minor modifications. One microliter of T7E1 enzyme was used in each reaction and incubation performed at 37°C for 1 h. Separation of obtained fragments on 2% agarose gels allowed the detection of mismatched DNA and occurrence of mutations at target sites in some of the clones tested. Positive M. gaditana colonies were transferred onto a new selective plate to separate cells harboring the mutation from those that presented a WT-like sequence. Mutation at MgACSBG target sites was confirmed by sequencing of the DNA at the target locus (Eurofins, France).

Glycerolipid analyses

Glycerolipids were extracted from freeze-dried M. gaditana cells grown in 50 mL of indicated medium. About 50–100 × 10⁶ cells were required for each triplicate analysis. A freeze-dried pellet was suspended in 4 mL of boiling ethanol for 5 min to prevent lipid degradation, and lipids were extracted as described earlier (Simionato et al., 2013) by addition of 2 mL methanol and 8 mL chloroform at room temperature. The mixture was saturated with argon and stirred for 1 h at room temperature. After filtration through glass wool, cell debris was rinsed with 3 mL chloroform/methanol 2:1, v/v, and 5 mL of NaCl 1%, w/v, was then added to the filtrate to initiate biphase formation. The chloroform phase was collected and dried under argon before solubilizing the lipid extract in 1 mL of chloroform. Total glycerolipids were quantified based on their FA content: in a 10-µL aliquot fraction, a known quantity of 15:0 was added and FAs were converted into FA methyl esters (FAMEs) by a 1-h incubation in 3 mL 2.5% H₂SO₄ in pure methanol, v/v, at 100°C (Jouhet et al., 2003). The reaction was stopped by addition of 3 mL water and 3 mL hexane was added for phase separation. After 20 min incubation, the hexane phase was transferred to a new tube. FAMEs were extracted a second time via the addition, incubation, and extraction of another 3 mL hexane. The combined 6 mL were argon-dried and re-suspended in 30 µL hexane for gas chromatography-flame ionization detector (GC-FID; Perkin Elmer) analysis on a BPX70 (SGE) column. FAMEs were identified by comparison of their retention times with standards (Sigma) and quantified by the surface peak method using 15:0 for calibration. Extraction and quantification were performed with at least three biological replicates. Glycerolipids were then analyzed and quantified by high pressure liquid chromatography–tandem mass spectrometry (HPLC-MS/MS), with appropriate standard lipids. The lipid extracts corresponding to 25 nmol of total FAs were dissolved in 100 µL of chloroform/methanol [2/1, (v/v)] containing 125 pmol of each internal standard. Internal standards used were PE 18:0-18:0 and diacylglycerol (DAG) 18:0-22:6 from Avanti Polar Lipid, and SQDG 16:0-18:0 extracted from spinach (Spinacia oleracea) thylakoid (Deme et al., 2014) and hydrogenated (Buseman et al., 2006). Lipids were then separated by HPLC and quantified by MS/MS. The HPLC separation method was adapted from previously described procedure (Rainteau et al., 2012). Lipid classes were separated using an Agilent 1200 HPLC system using a 150 mm × 3 mm (length × internal diameter) 5 µm diol column (Macherey-Nagel), at 40°C. The mobile phases consisted of hexane/isopropanol/water/1 M ammonium acetate, pH 5.3 [625/350/24/1, (v/v/v/v)] (A) and isopropanol/water/1 M ammonium acetate, pH 5.3 [850/149/1, (v/v/v)] (B). The injection volume was 20 µL. After 5 min, the percentage of B was increased linearly from 0% to 100% in 30 min and kept at 100% for 15 min. This elution sequence was followed by a return to 100% A in 5 min and an equilibration for 20 min with 100% A before the next injection, leading to a total runtime of 70 min. The flow rate of the mobile phase was 200 µL/min. The distinct glycerophospholipid classes were eluted successively as a function of the polar head group. Mass spectrometric analysis was performed on a 6,460 triple quadrupole mass spectrometer (Agilent) equipped with a Jet stream electrospray ion source under following settings: drying gas heater at 260°C, drying gas flow at 13 L·min⁻¹, sheath gas heater at 300°C, sheath gas flow at 11 L·min⁻¹, nebulizer pressure at 25 psi, capillary voltage at ±5,000 V, and nozzle voltage at ±1,000 V. Nitrogen was used as collision gas. The quadrupoles Q1 and Q3 were operated at widest and unit resolution, respectively. PC and DGTS analyses were carried out in positive ion mode by scanning for precursors of m/z 184 and 236, respectively, at a collision energy (CE) of 34 and 52 eV. SQDG analysis was carried out in negative ion mode by scanning for precursors of m/z −225 at a CE of −56 eV. PE, phosphatidylinositol (PI), PG, MGDG, and DGDG measurements were performed in positive ion mode by scanning for neutral losses of 141, 277, 189, 179, and 341 Da at CEs of 20, 12, 16, 8, and 8 eV, respectively. DAG and TAG species were identified and quantified by multiple reaction monitoring (MRM) as singly charged ions [M+NH4]⁺ at a CE of 16 and 22 eV, respectively. Quantification was done for each lipid species by MRM with 50 ms dwell time with the various transitions previously recorded. Mass spectra were processed using the MassHunter Workstation software (Agilent) for identification and quantification of lipids. Lipid amounts (pmol) were corrected for response differences between internal standards and endogenous lipids as described previously (Jouhet et al., 2017).

Acyl-CoA analysis

Freshly harvested cells cultivated in 10× ESAW (nutrient replete) medium were frozen in liquid nitrogen, lyophilized, and transferred to Rothamsted Research, UK, for acyl-CoA extraction and analysis (15–30 mg dry weight per sample). In brief, acyl-CoAs were extracted using a bead beating method and analyzed using liquid chromatography coupled to tandem mass spectrometry (LC–MS/MC) plus MRM in positive ion mode. Acyl-CoA extraction efficiency was normalized to a 17:0 CoA (Sigma–Aldrich) internal standard. All samples were immediately analyzed after extraction and were maintained on a cooled stage prior to LC–MS/MS plus MRM analysis (using an ABSciex 4000 QTRAP device), as described earlier (Haynes et al., 2008) using an Agilent 1200 LC system with a Agilent Eclipse XDB-C18 (3 × 100 mm; 3.5-μm particle size) column. For the purposes of identification and calibration, standard acyl-CoA esters with acyl chain lengths from C10 to C20 (10:0; 12:0; 14:0; 16:0; 16:1; 16:3; 18:0; 18:1; 18:2; 18:3; 18:4; 18:5; 20:0; 20:3; 20:4; 20:5) were purchased from Sigma–Aldrich as free acids or lithium salts.

Modeling of MgACSBG structure bound to Co-enzyme A, alpha-linolenic acid (18:3), and AMP

Structure modeling was based on the identification of ACS conserved motifs, secondary structure predictions, and alignments with homologous sequences, for which protein 3-D structures have been resolved and stored in the PDB (Burley et al., 2019). Motif predictions were performed using the online servers MOTIF (Kanehisa et al., 2002), PROSITE (Sigrist et al., 2013), and Superfamily (Wilson et al., 2009). Secondary structures were predicted using the Porter (Mirabello and Pollastri, 2013) and Predictprotein (Yachdav et al., 2014) online tools. Search for homologous sequences was performed using Blast (Altschul et al., 1990). Automated structure prediction was attempted using Robetta (Song et al., 2013), RaptorX (Kallberg et al., 2012), Swiss-Model (Waterhouse et al., 2018), PhyRe2 (Kelley et al., 2015), I-Tasser (Zhang, 2009), PsiPred (Buchan et al., 2013), and PS(2) (Huang et al., 2015) servers. Four templates were selected for MgACSB modeling: 1ULT, (a long chain fatty ACS homodimer from Thermus thermophiles; Hisanaga et al., 2004); 1PG4 (acetyl-CoA synthetase from S. enterica; Gulick et al., 2003); 5MSC (the A domain of carboxylic acid reductase from N. iowensis; Gahloth et al., 2017); 3EQ6, a human ACS medium-chain family member (Kochan et al., 2009). The sequences of these four template proteins were aligned with MgACSBG in Supplemental Figure S4. 3-D structural homology models were built with Modeller (Sali and Blundell, 1993) for the full length MgACSBG protein, from residue M1 to A649. Modeler being able to use multiple structure alignments, four models were built using different association of template proteins: M1pg4 from 1PG4 alone; M3eq6 from 3EQ6 alone; and two consensus models Mcons1 from 1PG4, 1ULT, and 5MSC; and Mcons2 from all four protein templates (Supplemental Data S1–S4). Positioning of CoA and AMP in the models was based on their neighboring residues in structures 1PG4 and 3EQ6 (Supplemental Table S2, A and B) using a transformation matrix to orient these residues in the X-ray structure of 1PG4 toward their homologs in the models of MgACSBG. In brief, to position CoA, backbone atoms of residues 305, 356, 357, 360, etc. in 1PG4 were translated and rotated to minimize the root-mean-square (rms) deviation with residues 259, 310, 311, 314, etc. in the model of MgACSBG. The same transformation was applied on the CoA molecule present in 1PG4 to properly insert it in the models. An AMP molecule, also present in 1PG4, was positioned in the two models M1pg4 and Mcons1, in the same way. Since 3EQ6 also contains both CoA and AMP, the positioning of CoA and AMP in models M3eq6 and Mcons2 was done using the matrix transformations pertaining to 3EQ6. As an example of a FA, α-linolenic acid (18:3; C₁₈H₃₀O₂ cis-Δ9,12,15) was used. The lipid was initially placed manually inside the structures with the head in the direction of AMP, respecting its proximity with CoA and profiting of hydrophobic holes seen in the structures. The final models were finally energy minimized with the molecular dynamics (MD) program CHARMM (Brooks et al., 2009) using distance restraints between residues listed in Supplemental Table S2, A and B and AMP or CoA. All missing parameters in the CHARMM force field were obtained from the SwissParam server (Zoete et al., 2011). All models were checked for allowed residues in the Ramachandran plot (Ramachandran et al., 1963) and corrected in case of presence of D-amino acids or cis-peptide bonds with Procheck (Laskowski et al., 1993). Finally, all models were subject to 1 ns Langevin MD simulation at 300 K to regularize the structures before a final minimization. A summary of model energies and number of misplaced residues is shown in Supplemental Table S3. Three models M1pg4, M3eq6, and Mcons1, have similar final energy and correctly respect the restraints. Mcons2 shows a larger energy due mainly to the more difficult respect of distance restraints. All of them present in regions distant from the active site, long unstructured loops, and some amino acids in disallowed regions of the Ramachandran plot (Ramachandran et al., 1963). All models are provided as Supplemental Data (Model) S1–S4.

Yeast expression of MgACSBG variants

Strains and growth media

The isogenic S. cerevisiae strains WT (YB332: MATa ura3 leu2 his3A200 ade2 lys2-801) and faa1Δfaa4Δ (YB525: MATa ura3 leu2 his3Δ200 ade2 lys2–801 faa1Δ::HIS3; faa4Δ::LYS2) used in this study were kindly provided by Dr. Paul Black (University of Nebraska at Lincoln, USA) and have been described earlier (Johnson et al., 1994a, 1994b). After transformation, all the strains described in the present work were grown on synthetic media containing 6.7 g L⁻¹ of Yeast Nitrogen Base (YNB, MP biomedicals), 0.77 g L⁻¹ of CSM-URA (MP Biomedicals) and g L⁻¹ of either raffinose or galactose, as described. In complementation experiments, the media were supplemented with 22.5 µM cerulenin (Sigma–Aldrich), 100 µM FFA, and 0.2% (v/v) final concentration TWEEN 80 (Sigma–Aldrich). The different FFAs were dissolved in 100% TWEEN 80 at a concentration of 50 mM. Cerulenin was dissolved in acetone at 45 mM. As controls, media without FFAs and/or cerulenin were also supplemented with 0.2% (v/v) TWEEN 80 and/or acetone.

Cloning of the coding sequences of variants of MgACSBG

The three variants of MgACSBG (MgACSBG, MgACSBG#5, and MgACSBG#31) were cloned into the PstI-BamHI cloning site of the YEplac195 backbone (X75459.1) under the control of a galactose inducible promoter (pGAL1) and with the ADH1 terminator (t-ADH1). A 6× His tag was cloned in frame with the genes of interest in 3′, upstream and in frame with the stop codon. Cloning was performed using the NEB Gibson Assembly Master Mix following manufacturer’s instructions. Fragments of the pGAL1 promoter, the t-ADH1 terminator, and WT or mutated MgACSBG coding sequences were amplified by PCR performed with Thermo Scientific Phusion High-Fidelity DNA Polymerase, following the manufacturer’s instructions. The assembled plasmids were transformed in MAX Efficiency DH5α Competent Cells. Ampicillin-resistant Escherichia coli colonies were PCR screened using pGAL1 and t-ADH1 flanking primers to validate the insert size. Two PCR positive colonies per construct were sequenced. The sequence-validated plasmids were transformed in the faa1Δfaa4Δ S. cerevisiae double mutant by the lithium acetate method (Gietz and Schiestl, 2007) and selected on CSM-Ura. As controls, WT and faa1Δfaa4Δ strains were transformed with the empty YEplac195 vector.

Three clones per transformation were used for complementation tests. For each experiment, one colony was inoculated in 2.7 mL of CSM-URA + raffinose and grown over-night at 30°C at 250 rpm. Transgene expression was induced by the addition of galactose 2% (p/v) final concentration and incubation for 6 h before measurement of the optical density at 600 nm. Time course was performed in 96-well microplates (Thermo Scientific Nunc MicroWell 96-Well, Nunclon Delta-Treated, Flat-Bottom Microplate) filled with 200 µL of culture in galactose medium supplemented or not by FA or cerulenin and set at an OD₆₀₀ = 0.3. The microplate was incubated at 30 ± 0.5°C in an Infinite M1000 PRO microplate reader (Tecan) set for plate orbital shake for 15 s every 20 min and absorbance measurements (OD₆₀₀) every 60 min, with four measurements per well. To the OD₆₀₀ reads from experimental wells, the OD₆₀₀ measured in blank wells containing the culture media only were subtracted to obtain adjusted OD₆₀₀ values. Such values were transformed using an equation experimentally calculated from a calibration curve using an Eppendorf BioPhotometer UV/Vis Spectrophotometer.

Protein expression levels were evaluated by immunoblot analyses. After induction, an amount of cells corresponding to an OD₆₀₀ of 50 was retrieved by a centrifugation of 3 min at 600 × g and washed with 1 mL of sterile water. Cells were resuspended in 200 µL of protein extraction buffer (150 mM Tris–HCl pH 6.8, 8 M urea, 1 mM EDTA, 1 mM DTT), and 100 µL of acid-washed 0.5 µm glass beads were added. Cells were heated for 5 min at 70°C and ground using the PreCellys Evolution (Bertin Instruments) by three cycles of 20 s of agitation at 10,000 rpm with pauses of 30 s between each cycle. The heating and grinding steps were performed twice and samples were centrifuged 5 min at 5,000 × g at 4°C. The supernatant was retrieved and proteins were quantified using Bradford Assay (BioRad) according to manufacturer’s instructions. Twenty micrograms of total proteins was used for immunoblot analysis. Proteins were separated on Bolt Bis-Tris Plus 4%–12% (Life Technologies) polyacrylamide gels in MES SDS Novex Bolt migration buffer (Life Technologies) and then transferred in liquid in Novex Bolt Transfer buffer (Life Technologies) according to manufacturer’s instructions. Immunoblots were performed with anti-His C-Term antibodies (Invitrogen) to detect MgACSBG expression and anti-Kar2 antibodies (Santa Cruz Biotechnology) for loading control. Revelation was performed with Clarity Western Blot ECL substrate (BioRad) using an ImageQuant800 imager (Amersham).

Accession numbers

Naga_100014g59; Naga_100012g66; Naga_101051g1; Naga_100047g8; Naga_100649g1; Naga_100035g43; Phatr3_J20143; Phatr3_J12420; Phatr3_J54151; Phatr3_J45510; Phatr3_J17720; NP_524698; NP_001285923; NP_444408; XP_015155301; Q96GR2; Q5FVE4; F1NLD6.

Supplemental data

Supplemental Figure S1. Overview of glycerolipid biosynthesis.

Supplemental Figure S2. Multiple alignment of ACSBG proteins sequences from animals and heterokonts.

Supplemental Figure S3. Secondary structure prediction of MgACSBG (Naga_100014g59).

Supplemental Figure S4. Sequence alignment of MgACSBG (Naga_100014g59 or NAGA) with proteins of known 3-D structure.

Supplemental Figure S5. Expression of MgACSBG, MgACSBG#5, and MgACSBG#31 in the S. cerevisiae faa1Δfaa4Δ double mutant.

Supplemental Table S1. In depth analysis of PC and betaine lipid (DGTS) FA profiling by mass spectrometry

Supplemental Table S2. Amino acids in MgACSBG sequence predicted to be in the vicinity of Coenzyme A (CoA) and adenosine monophosphate (AMP)

Supplemental Table S3. Evaluation of 3-D models of MgACSBG

Supplemental Model S1. Model (1) acs4_3eq6_ac_3eq6_lipr_d1_min, in pdb format.

Supplemental Model S2. Model (2) acs4_al2_ac_3eq6_lipr_d1_min, in pdb format.

Supplemental Model S3. Model (3) acs4_all_ac_1pg4_lipr_min, in pdb format.

Supplemental Model S4. Model (4) acs4_1pg4_ac_1pg4_lipr_min, in pdb format.

 

E.B., L.M., S.L., M.B., L.P., A.A., and S.Co. performed the molecular biology experiments and genetic and phenotypic studies. A.A.-R. contributed to the growth experiments in multiple devices and contributed to phenotypic analyses. M.M. and G.F. provided the specific expertise in metabolic and physiologic analyses. F.R. and J.J. provided the specific expertise in lipidomic analyses. E.B., R.D., and S.Cr. performed the bioinformatic and modeling analyses. E.B., L.F., F.L., S.Co., J.J., and E.M. participated to the conception of research plans. J.J. and E.M. conceived the project. All the authors contributed to the writing of the manuscript.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plcell) is: Eric Maréchal (eric.marechal@cea.fr).