FAX1, a Novel Membrane Protein Mediating Plastid Fatty Acid Export

1Fatty acids (FAs) are building blocks for the majority of cellular lipids, which are essential throughout life of organisms. Besides their role as constituents of biological membranes, plant acyl-lipids are used for diverse functions at different destinations and tissues (reviewed in). For example, triacylglycerols (TAGs) in seeds of oilseed plants represent the major form of carbon and energy storage. Cuticular waxes at the surface of plants restrict loss of water and provide protection against pathogen attack. Furthermore, the formation of pollen cell walls is strictly dependent on delivery of modified FAs from tapetum cells in anthers (reviewed in). De novo FA synthesis in plants occurs in plastids (for overview, see). Growing alkyl chains in the plastid stroma are attached as acyl moieties to acyl carrier protein (ACP), and in seed plants become available for lipid assembly mainly in the form of palmitoyl (16:0)- and oleoyl (18:1)-ACP. Part of these long-chain FAs will be integrated into lipids inside plastids (prokaryotic pathway); the majority, however, is exported to the endoplasmic reticulum (ER) for further elongation, acyl editing, and lipid synthesis (eukaryotic pathway).

2Although it is generally agreed that free FAs are shuttled across plastid envelope membranes, the mode of export still remains enigmatic since until now, no membrane-intrinsic transporter protein could be associated with a direct function in plastid FA export (for overview, see). On the one hand, a facilitated diffusion of free FAs through the lipid environment of membranes is suggested, which is supported by the recent finding that an acyl-ACP synthase in the cyanobacterium Synechocystis sp. PCC6803 is necessary and sufficient for FA transfer across membranes. On the other hand, several ATP-binding cassette (ABC) transporter proteins for lipids, FAs, or acyl-coenzyme A (CoA), and for import of FAs into peroxisomes, as well as FA-transport systems from Escherichia coli, yeast, or mammals, provide evidence for an active mode of transport in plastids. Nevertheless, before transport, acyl-ACP thioesterases at the inner plastid envelope membrane catalyze the hydrolysation of fatty acyl-ACP to free FAs. After crossing both inner and outer plastid envelope membranes (IE, OE), free FAs are re-activated to acyl-CoAs by long-chain acyl-CoA synthetases (LACSs). As demonstrated for the protein LACS9, these enzymes can attach to the cytosolic face of the plastid OE. At the ER membrane, the ABC transporter ABCA9 has recently been described to be involved in FA-uptake, most likely in the form of acyl-CoA, thereby being important for TAG synthesis during seed filling. Once arrived in the ER lumen, plastid-derived FAs are utilized for synthesis of specific lipid classes via the so-called eukaryotic pathway, where phosphatidic acid (PA) represents an important intermediate, phosphatidyl-choline (PC) is a major membrane phospholipid, and TAGs are the energy storage lipids produced. Subsequently, these eukaryotic lipids are distributed to various subcellular locations. For re-import of eukaryotic lipids into plastids, most likely in the form of ER-derived PA, an ABC transporter system (TGD1, 2, 3) at the IE and the PA-binding ß-barrel lipid transfer protein TGD4 in the OE are required. In plastids, the diacylglycerol backbone from these eukaryotic precursors is used for synthesis of the galactolipids MGDG, DGDG (monogalactosyl-, digalactosyl-diacylglycerol), and the sulfolipid SQDG (sulfoquinovosyl-diacylglycerol). In addition, however, a prokaryotic-type pathway also produces MGDG, DGDG, SQDG, and the phospholipid phosphatidyl-glycerol (PG) directly from newly synthesized FAs and thus does not require previous FA-export from plastids (for overview, see).

3Here we describe FAX1, a novel protein in the IE of plastids that belongs to the Tmemb_14 superfamily of membrane proteins with so-far unknown function. Functional studies in yeast as well as FAX1 mutant analysis in Arabidopsis thaliana clearly demonstrate that FAX1 mediates FA-export from plastids and thus, to our knowledge, represents the first membrane-intrinsic protein described to be involved in this process. In mammals, FAX1 relatives are structurally related mitochondrial membrane proteins, for which the biological task is not yet clear. Thus, FAX1 not only is a missing link to explain the mode of plastid FA-export and to improve plant lipid/biofuel production but might also propel the understanding of Tmemb_14 protein performance in general.

4The Arabidopsis protein At-FAX1 (At3g57280, for fatty acid export 1) was previously annotated as potential plastid-targeted and plant-specific solute transporter by proteomic and phylogenetic analysis. Furthermore, we identified transcripts of At-FAX1 to be up-regulated upon induction of early leaf senescence. To analyze protein function, we isolated the cDNA of FAX1 genes from Arabidopsis and pea (Pisum sativum). For both proteins, chloroplast targeting peptides and four hydrophobic α-helices are predicted (Fig. 1A). By the latter, plant FAX1 clearly groups to the so-called Tmemb_14 superfamily of proteins with so-far unknown function. The Tmemb_14 family is ubiquitous, with members in nearly all eukaryotes and some bacteria (InterPro|UPF0136). In Arabidopsis, four proteins (FAX1–FAX4) are predicted to be targeted to plastids, while three (FAX5–FAX7) most likely are directed to other, non-plastid membranes via the secretory pathway (Fig. 1B). The plastid-intrinsic FAX1 is restricted to the chlorophyll-containing plant kingdom, with representatives in mono- and dicotyledons as well as in mosses and green algae (compare InterPro|UPF0136,). Relatives of non-plastid predicted At-FAX proteins, however, can be found in eukaryotes such as mammals, insects, or yeast, and in some bacteria and cyanobacteria (e.g., Chlamydiae or Nostocales).

5For all Tmemb_14 proteins, four hydrophobic α-helical domains are predicted (Fig. 1). However, nuclear magnetic resonance (NMR) structure determination of the human Tmemb_14 proteins TMEM14A and TMEM14C showed that only three of these helices are membrane-spanning. TMEM14A contains an amphiphilic N-terminal helix, presumably located at the lipid micelle-water interface, while for TMEM14C an amphiphilic helix that orients perpendicular to the lipid bilayer, is placed between the second and third membrane domain. Amino acid sequences of the plastid FAX1 and the non-plastid At-FAX6 nicely align to both TMEM14A and 14C (Fig. 1C), but structural modeling revealed that the mature At-FAX1 and Ps-FAX1 are more similar to TMEM14C (Fig. 2A). Here, three membrane-spanning and one amphiphilic helix are likely, while the additional N-terminal amino acids of FAX1 proteins might form another α-helical domain not present in TMEM14C. In contrast, the structure of At-FAX6 resembles that of TMEM14A with an N-terminal amphiphilic helix followed by three transmembrane domains (Fig. 2B).

6With its membrane-spanning domains, FAX1 inserts into the inner envelope membrane (IE) of plastids as could be shown by in vivo GFP-targeting and immunoblot analysis. At-FAX1-GFP signals in Arabidopsis protoplasts, which can be detected as rings around chloroplasts (Fig. 2C), point to an envelope localization. This could be confirmed and specified to IE by immunoblot analysis using sub-fractionated pea chloroplasts. In pea IE membranes as well as in Arabidopsis chloroplast envelopes, FAX1 signals appear as a band of about 25kDa (Fig. 2D). In agreement, FAX1 peptides in proteomic analyses of plastid membranes were exclusively detected in IE preparations (see and references therein). To exclude ER localization, we further probed against ER-enriched Arabidopsis microsomal membranes (see), where no FAX1 signals could be detected (Fig. 2D).

7To study the in vivo function of FAX1, we analyzed loss-of-function and over-expressing mutant lines in Arabidopsis. We selected fax1–1 and fax1–2 with T-DNA insertions in the first intron and first exon of the FAX1 gene, respectively (S1A Fig.). Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis showed that both homozygous alleles represent knockouts for FAX1 (S1B Fig.). To complement this loss-of-function, At-FAX1 cDNA under control of the 35S promoter was introduced into heterozygous fax1–2 plants. Subsequently, two lines homozygous for the fax1–2 allele (Co#7 and Co#54) were selected for further analysis. To stable over-express FAX1 in wild-type plants, the 35S::FAX1 construct was introduced into Col-0, and two independent lines named ox#2 and ox#4 were identified as FAX1 over-expressors. Quantitative real time RT-PCR revealed that FAX1 transcripts in line Co#7 are at wild-type levels, whereas line Co#54 contains about 12 times more FAX1 mRNA. Over-expression in ox#2 seedlings was mild (about 2-fold), but strong in line ox#4 (about 200 times more mRNA than in wild type; S1C Fig.). Immunoblot analysis confirmed the strength of FAX1 expression in these lines and the knockout in fax1–2 on the protein level (S1D Fig.).

8Homozygous fax1–1 and fax1–2 knockout mutants both were characterized by reduced biomass at mature rosette stages (Fig. 3A, Table 1). Full flowering fax1 knockouts were significantly smaller than wild type, had thinner inflorescence stalks, and flowers producing short siliques that contained almost no seeds (Fig. 3B, C). Detailed analysis of different tissues and organs revealed that the decrease in biomass of fax1 lines was detectable throughout the entire plant body, including root, leaf, and stem tissues (Table 1). Because differences in stem dry weight were slightly more pronounced than in fresh-weight (FW) samples, most likely cell wall synthesis was affected. This could be confirmed by ultrastructural analysis of stem tissue (S2 Fig.). Here fax1 knockouts showed small vascular bundles with reduced secondary cell walls (S2A, D, G Fig.). Since the same phenotype was detected in both independent T-DNA insertion lines fax1–1 and fax1–2, and could be reverted by complementation in lines Co#7 and Co#54 (Fig. 3, Table 1), we conclude that the reduced biomass is caused by the loss of FAX1 function.

9Remarkably, FAX1 over-expressing lines ox#2 and ox#4 were slightly bigger and produced more biomass as well as thicker inflorescence stalks than wild type (Fig. 3, Table 1), thus behaving opposite to fax1 knockouts. In stems, this led to about one more hypodermal cell layer and to extended vascular strands, including an increased amount of xylem and phloem vessels, as well as a multi-layered procambuim (S2C, F, I Fig.). Because fresh weight of ox#2 and ox#4 stems was significantly higher than in wild type, but—in contrast to fax1 knockouts—stem dry weight of FAX1ox lines was similar to wild type (Table 1), the increased biomass of FAX1 over-expressors is most likely mainly due to enhanced production of cells. However, since tracheid walls of ox#2 appeared to be slightly thicker than in Col-0 (S2H, I Fig.), we can’t fully exclude an additional effect on the size of secondary cell walls. Interestingly, the rate of FAX1 overproduction—i.e., 2-fold for ox#2, 200-fold for ox#4—did not quantitatively affect the strength of biomass phenotypes, indicating a rather non-linear effect of protein function.

10To understand the peculiar loss-of-function phenotype of homozygous fax1 knockouts during flower and silique development, segregation analysis of mutant alleles was performed. Self-fertilization of heterozygous fax1–1 and fax1–2 revealed that the ratio of homozygous progeny was 7% and 4%, respectively, pointing to defect male and/or female gametophytes (Table 2). However, when stigmata from homozygous fax1–2 flowers were fertilized with wild-type fax1–2 pollen, normal seeds with 100% heterozygous fax1–2 mutant alleles were produced, indicating fertile fax1 knockout female gametophytes and sporophyte organs. In contrast, pollination of wild-type stigmata with homozygous fax1–2 anthers, produced short siliques, as observed during selfing of homozygous fax1–2 mutants (see Fig. 3C), and led to an estimated seed yield <0.1% of wild type. Furthermore, during manual crossing it became evident that pollen grains of homozygous fax1–2 flowers were improperly released from anthers. To minimize potential anther defects from the paternal sporophyte, we thus pollinated homozygous fax1–2 stigmata with heterozygous fax1–2 anthers, thereby producing 12% progeny homozygous for fax1–2 (Table 2). In summary, these results point to impaired transmission of male gametophytes (pollen) and defects of the male sporophyte (anther) in fax1 knockouts, finally leading to the observed male sterility.

11To further analyze flower development, in particular that of male parts, we examined the morphology of flower tissue from FAX1 mutant lines (Fig. 4; S3 Fig.). In FAX1 wild-type and over-expressors (ox#2, ox#4), pollen release by anther dehiscence, transfer to the stigma, and fertilization as indicated by high yield of viable seeds was normal. However, flowers of fax1 knockout mutants showed stigmata with non-pollinated papillae. In addition, fax1 anthers released only very few pollen grains (Fig. 4A, B; S3A, B, G Fig.). In flowers of complemented fax1–2 lines (Co#7, Co#54) in comparison, more free pollen grains than in fax1 knockouts but less than in wild type were visible, indicating incomplete recovery of pollen release (S3D, G Fig.). In contrast to the rest of the plant organs, where regeneration of fax1 knockout defects in Co#7, Co#54 lines was 100% (see Fig. 3; Table 1), complementation of fax1–2 pollen phenotypes was incomplete. This effect was best visualized by the colorless pollen of fax1 knockout and complementation lines (S3B, D, G Fig.), due to the absence (fax1–2) or incomplete (Co#54) assembly of a pollen coat (compare Fig. 4D–F; S3I–K Fig.) that normally includes yellow flavonoid and carotenoid deposits (for overview, see). The incomplete complementation was restricted to pollen grains and most likely is due to the 35S promoter system, which in Arabidopsis shows no or reduced activity in pollen grains and anther tissue, respectively.

12Subsequently, the detailed structure of anther tissue and pollen grains of fax1–2 knockout, Col-0 wild type, and the complementation line Co#54 was visualized by light- and transmission electron microscopy (TEM) at the mature, tricellular pollen stage (Fig. 4C–F; S3H–K Fig.). In general, anthers of fax1–2 were smaller than in wild type and the surface of pollen grains appeared to be wrinkled (Fig. 4C). Cross sections revealed an impaired release of fax1–2 pollen, although pollen sacks were wide open, indicating full dehiscence of anthers. Tapetum cells seemed to be degraded as expected for the developmental stage analyzed, however, the locule of fax1–2 anthers was covered by an electron-dense material, which stuck to pollen grains and thus most likely was responsible for improper pollen delivery (Fig. 4C–E). Ultrastructural resolution demonstrated that the well-defined structures of the outer pollen cell wall—i.e., the exine layer and the pollen coat, which covers the exine surface and its cavities—were absent in fax1–2 knockouts (Fig. 4E, F). The intine, representing the innermost layer of the pollen cell wall and composed of cellulose, pectin, and various proteins, secreted by the microspore (gametophytic origin, see), however, looked intact. In contrast, mature wild-type pollen showed a complete exine, consisting of a flat nexine layer and the sculpted sexine parts tectum and bacula. Furthermore, the latter were filled and covered with the tryphine pollen coat (Fig. 4E, F). Pollen cell walls of Co#54 displayed an intermediate state of biogenesis with visible nexine layers, but incomplete arrangement of tectum and bacula structures as well as pollen coat material (S3J, K Fig.). As described above, these findings point to an incomplete complementation of fax1–2 knockouts in pollen.

13In conclusion, structural analysis of anthers and mature pollen grains showed that FAX1 is essential for biogenesis of the outer pollen cell wall, in particular for the assembly of exine and pollen coat. Both layers consist of complex biopolymers, such as sporopollenin (exine) and tryphine (pollen coat), that are mainly made of FA-derivatives and lipids originating from the tapetum tissue of anthers (sporophytic origin, see). Thus, FAX1 might play a role in delivery of these compounds by mediating FA-export from tapetum cell plastids. Most likely, the electron-dense, sticky material in fax1 knockout anthers that prevents release of pollen grains represents cellular debris of degenerated tapetum cells and/or not-incorporated sporopollenin or tryphine material.

14Because during analysis of FAX1 mutants an altered surface of epidermal cells was apparent, we investigated structure as well as wax and cutin coverage of epidermis cells from primary inflorescence stalks of FAX1 mutants (Fig. 5). Microscopic analysis revealed that the width of epidermal cell walls in fax1–1 was strongly reduced when compared to wild type (Fig. 5A, B). As for cell walls in xylem vessels (S2 Fig.), a strong effect was only found for knockout and not for FAX1 over-expressing lines. However, an electron-dense cover at the extracellular side of the cell walls, most likely representing the cutin matrix of the cuticular layer, appeared to be more intense in ox#2, but reduced in fax1–1 (Fig. 5B, C). To examine the constitution of the cuticular layer, we therefore determined wax and cutin coverage from stem epidermis cells. Surprisingly, the total loads of different wax and cutin species were not altered for all lines analyzed (fax1–1, fax1–2, Col-0, WT2, Co#7, Co#54, ox#2, ox#4, see S1A Data). Furthermore, no change in composition regarding aliphatic chain length or functional groups (e.g. ketones, acids or aldehydes) could be detected. The only significant difference we found was for C29-ketone wax components, which were reduced in both fax1 knockout lines by more than 50% when compared to stems from all other plants (Fig. 5D).

15Since cutin contents were unchanged, the different strengths of the outer layer of epidermal cell walls observed by TEM most likely are due to stronger (ox#2) and weaker (fax1–1) crosslinking of the cutin matrix with cell walls. The wax composition of cuticular layers, however, is dependent on plastid FA-synthesis as well as excretion of modified FAs via the plasma membrane of epidermis cells (for overview, see). In parallel to the assembly of sporopollenin and tryphine material in pollen cell walls (see above), FAX1 might thus be necessary for plastid FA-export, previous to synthesis and release of ketone wax components.

16Because fax1 knockouts revealed a lack of FA- and lipid-derived compounds in pollen as well as stem epidermis cells (see above), we measured free FAs and polar lipid species such as phospho-, sulfo-, galacto-lipids, and triacylglycerols in leaves and flowers of mature FAX1 mutant plants (S1 Table). To spotlight changes in FAX1 mutants compared to wild type, we determined relative values and summarized representatives of significantly different levels, as well as abundant species from each molecule class in the next two figures. For comparison to the overall FA/lipid composition of each tissue, we listed contents in mol% of all significantly different species in S2–S4 Tables, and further estimated the impact of changes in mol% of each molecule class in the next table.

17In leaves of fax1 knockout plants, levels of 30 FA and polar lipid species (irrespective of TAGs) were significantly different from wild type (S2 Table). For free FAs, we observed an increase of plastid-produced FA 18:2 (Fig. 6A, S2A Table) and a decrease for FAs 20:0, 24:0, 26:0 (Fig. 6C, S2C Table), which are elongated at the ER. Whereas aggregate levels of 34:x glycolipids (MGDG, DGDG, SQDG) were only slightly elevated (Table 3, S2A Table), the highly abundant MGDG 36:6 (11.7 mol% in wt) with an ER-made diacylglycerol backbone was considerably less (2.7 mol%) than in wild type (Fig. 6C, S2C Table). The eukaryotic-type DGDG 36:6, however, increased contributing about 0.7 mol% more to the overall lipid content (Table 3, S2C Table). Strong upward changes were observed for phosphatidyl-glycerol (PG) species (2.7- to 5-fold; Fig. 6A, S2A Table), leading to an entire gain of up to 3.2 mol% (Table 3) of these mainly plastid-derived phospholipids. In contrast, the ER-produced phospholipids phosphatidyl-choline (PC) and -ethanolamine (PE) mostly decreased in fax1 knockout leaves (Fig. 6C, S2C Table). Here the effect, in particular of highly abundant PC 34:3, PC 36:6 (9.3, 7.4 mol% in wt), was especially strong and is estimated to primarily contribute to a total reduction of PCs by 8.8 mol% (Table 3). Whereas the overall decrease of PE was about 0.5 mol%, phosphatidyl-inositol (PI) contents showed a pronounced upward fold change, which, however, only very slightly contributed to the overall lipid composition, and thus leaf PI might rather be involved in signaling (Table 3, S2C Table).

18In leaves of FAX1 over-expressing lines (Fig. 6B, D; Table 3), we found an opposite distribution of free FAs and lipids as in fax1 knockouts. Here, without counting TAGs, 28 molecule species were significantly different from wild type (S2B, D Table). Contents of all differentially regulated and mainly plastid-derived FAs, 34:x glycolipids (MGDG, DGDG, SQDG) as well as PG 34:2 dropped (Fig. 6B, S2B Table). A considerable impact on the total lipid content came from reduction of highly abundant molecules such as MGDG 34:5, 34:6; DGDG 34:2, 34:3; or SQDG 34:3, all with levels higher than 2 mol% in wild type. In consequence, the estimated overall reduction was about 2.8, 1.9, and 0.9 mol% for 34:x MGDG, DGDG, and SQDG, respectively (Table 3). For lipids produced by the eukaryotic pathway at the ER, we found a mild decrease of MGDG 36:5 (0.4 mol%) and only very minor changes (0.04–0.08 mol%) for SQDG 36:4, 36:5, and PE 34:3 (Table 3, S2D Table). The effect on PC contents, however, again was quite strong (total increase of about 3.0 mol%, see Table 3), including elevated levels of the abundant PC 34:1, 34:3, 36:2, and 36:3 (Fig. 6D, S2D Table).

19When compared to leaves, flower tissue of fax1 ko and FAX1ox lines showed a similar differential pattern of free FAs and lipids, which are presumably mainly produced via the prokaryotic pathway (Table 3, S3A, B Table). Whereas in fax1, FAs that after synthesis have to be exported from chloroplasts (i.e. 16:0, 18:0, 18:1) increased when compared to wild type (largest change for 16:0 = 0.24 mol%), the plastid internal FA 18:3 and the plastid external FA 24:0 decreased by about 0.1 mol% each (S3A, C Table). In FAX1 ox flowers only a minor increase of FA 18:0 was detected (S3B Table). As found in leaves, overall levels of 34:x MGDG, DGDG, and SQDG increased in knockouts but decreased in over-expressors (Table 3). The most prominent changes were for MGDG 34:6 (increase of 0.6 mol% in fax1, S3A Table) and for MGDG 34:5 (decrease of 0.3 mol% in FAX1ox, S3B Table).

20For several lipid species, which are assembled at the ER, however, patterns in flowers were different and more diverse than in leaves. These included a rise in 36:x MGDG levels (dominated by +0.8 mol % of MGDG 36:6) in fax1 knockouts (S3C Table); an increase and a decrease of about 0.45 mol% PE in fax1 and FAX1ox, respectively (Table 3); as well as a strong reduction of PC in FAX1ox flowers (up to 5.6. mol%, Table 3). In fax1 knockout flowers in contrast, PC (mostly PC 36:6 by-1.0 mol%; S3C Table) and also PI species (-0.26 mol%) significantly dropped (Table 3).

21The most robust differential distribution in both mutant lines and tissues, however, was found for triacylglycerol oils (Fig. 7; Table 3). Here we determined significant changes for more than half of the molecules measured (S4 Table). More important, however, was a massive decrease of high and low abundant TAGs in fax1 knockout leaves (Fig. 7A, S4A Table) and flowers (Fig. 7C, S4C Table) as well as a strong increase in FAX1ox leaves (Fig. 7B, S4B Table) and flowers (Fig. 7D, S4D Table). Fold changes were highest for low abundant TAGs (e.g., 8.3-fold decrease for TAG 56:7 in fax1 leaves, S4A Table). As expected, the biggest impact on overall TAG content was by significant changes in high abundant species, resulting in a drop of-4.3 and-7.2 mol% in leaves and flowers of fax1 knockouts and an accumulation of +3.2 and +6.6 mol% for leaf and flower tissue from FAX1ox lines, respectively (Table 3).

22In summary, determination of free FAs and lipids in FAX1 mutants clearly shows that the function of FAX1 in the IE membrane of chloroplasts impacts cellular FA and lipid homeostasis. Overall we found significant differences compared to wild type for more than 50% of all species determined (Table 3). Together with the observed lack of FA- or lipid-derived compounds in fax1 knockout pollen cell walls and cuticular waxes (see above), these findings point to a function of FAX1 in FA-export from plastids (for details, see Discussion).

23In baker’s yeast (Saccharomyces cerevisiae) import of exogenous long-chain FAs by the so-called vectorial acylation process requires a multiprotein complex, which consists of Fat1p (the membrane-spanning transport protein) and Faa1p or Faa4p, acyl-CoA synthetases for intracellular FA activation. In order to test a potential FA-transport function of FAX1, we thus analyzed growth complementation of the yeast fat1 and faa1/faa4 mutants, which represent knockouts for Fat1p and Faa1p/Faa4p, respectively. Therefore, we transformed the coding sequence of the mature At-FAX1 protein into fat1 and faa1/faa4 cells. Since previous studies revealed that the uptake of the polyunsaturated FA α-linolenic acid (C18:3) into yeast cells was toxic for wild-type but not for fat1 cells, we challenged growth of FAX1-containing yeast mutants by addition of high α-linolenic acid concentrations (3.6 mM) to the media (Fig. 8A–C). In drop tests on agar plates, all cells showed normal growth under control conditions (Fig. 8A, left). In addition, yeast mutant cells, transformed with the empty vector only, were resistant to excess α-linolenic acid (Fig. 8A, right). However, fat1 cells expressing the mature At-FAX1 protein died in the presence of α-linolenic acid overload (Fig. 8A, right), indicating that FAX1 is able to restore FA-uptake in fat1 mutants. In contrast, α-linolenic acid induced cell death was not observed in faa1/faa4 cells, neither with nor without FAX1, pointing to a FAX1 function in FA-transport and not in FA-activation.

24Furthermore, we monitored a very similar behavior for growth kinetics of the respective yeast cells in liquid media (Fig. 8B, C). Here, a FAX1-mediated toxicity of α-linolenic acid was significant after 18 h when compared to empty vector cells. While this effect was highly significant and strong in fat1 mutants as indicated by a reduction of cell density to about 54% after 29 h (Fig. 8B), only a mild growth inhibition was detected in Δfaa1/faa4 (density of FAX1 cells was about 78% of control cells after 29 h; Fig. 8C). In addition, when compared to vector-only cells grown without α-linolenic acid, we observed a slight growth reduction by addition of α-linolenic acid itself as well as for FAX1-transformed cells in absence of α-linolenic acid, independent of yeast mutant strains (Fig. 8B, C). Whereas the former observation can be explained by unspecific, background uptake of α-linolenic acid provided at excess external concentrations, the latter effect might be due to a general, but minor, toxic effect of FAX1 expression in yeast.

25To assess specificity of FAX1 for FAs, which have to be exported from chloroplasts in vivo, i.e., palmitic (C16:0), stearic (C18:0), and oleic acid (C18:1), we performed additional yeast growth complementation assays in the presence of the FA-biosynthesis inhibitor cerulenin and supply of moderate external FA concentrations (i.e., 100 μM; Fig. 8D, S4 Fig.). Results with rapidly (S4A–C Fig.) and non-exponentially growing cells (S4D Fig.) allowed definition of a potential substrate specificity of FAX1, preferring C16:0 over C18:1 and C18:0 FAs (for details see S4 Fig. and Discussion). When we tested α-linolenic acid (C18:3), which in planta is not exported from plastids (see), as a control in this assay, FAX1 specificity was in the range as for stearic/oleic acid, but significantly lower than for palmitic acid (Fig. 8D, S4B Fig.).

26In summary, our results show that the protein FAX1 in the IE of plastids is able to mediate FA-export as supported by the following findings: (i) FA-transport function of FAX1 in yeast; (ii) differential distribution of ER- and plastid-derived FAs/lipids in FAX1 mutant plants; (iii) male sterility of fax1 knockout lines, caused by impaired delivery of FA-derived compounds; (iv) decreased ketone wax compounds in cuticular layers of fax1 knockout stems; (v) a focus on differential expression of genes for acyl lipid as well as carbohydrate and cellular/cell wall metabolism in FAX1 mutant lines (see S1 Text).

27Complementation of the yeast fat1, but not of the faa1/faa4 mutant, indicates that FAX1 is acting only in membrane transfer of FAs and not in FA-activation. This is in contrast to, for example, yeast and human FA-transporters such as Fat1p and FATPs, which in addition have acyl-activating functional domains. FAX proteins group into the Tmemb_14 family and thus most likely contain three hydrophobic, membrane-spanning α-helical domains and one amphiphilic helix at the lipid bilayer/water interface. Thus, it is tempting to speculate that the latter might be responsible for binding and transfer of FA-chains across the IE membrane. Once loaded with a FA produced in the plastid stroma, this α-helix might become lipophilic enough to fold into the lipid bilayer and flip FAs over the IE. Furthermore, FAX1 and also FAX2 (see Fig. 1B) contain an extended N-terminal region (gray helix in Fig. 2A). Structural modeling indicates that these stretches fold into additional, most likely non-membrane associated α-helices: one for FAX1, two for FAX2, respectively. Interestingly, the two anti-parallel helices of the FAX2 N-terminus fit to sequence and structure of a ‘four-helical up-and-down bundle’ of the human apolipoprotein apoE3, which is involved in lipid transport and binding during formation of lipoprotein particles. Amphiphilic α-helices in the C-terminus of apoE3 are described to bind to lipids and thereby induce a conformational change in the N-terminal helix bundle that allows detergent-like solubilization of lipids and formation of lipoprotein particles (for overview, see). Therefore, the N-terminal helices of plastid FAX proteins might be involved in similar functions during FA-transport. The different apparent molecular weights observed for FAX1 (S1D Fig.), most likely resulting from discrete conformations and/or packing of membrane domains, support these hypotheses for a transport mode. Once at the intermembrane space, FA-handover from FAX1 to substrate binding proteins, and transport across the OE membrane via a ß-barrel protein might be possible. For plastid re-import of eukaryotic lipids for example, the latter two proteins are represented by TGD2 (substrate binding) and TGD4 (OE ß-barrel,). Furthermore, in E. coli, a similar system has been described for export of lipopolysaccharides, including an ABC transporter that flips the lipid moiety across the inner membrane, transfer proteins in the periplasm, and a ß-barrel protein in the outer membrane. For plastids, subsequently an acyl-CoA synthetase (ACS) at the cytosolic face of the OE might finally drive FA-transfer in a passive, carrier-like process. Co-expression of At-FAX1 with LACS4 (ATTED-II coexpression networks), and regulation of LACS1, 3, and 5 expression in FAX1 mutants (S5 Table, S6 Table) underline a possible cooperation with ACS.

28The close structural similarity of FAX proteins to the human TMEM14A and TMEM14C, which both localize to mitochondrial membranes, in the future might enable explanation of TMEM14 protein function in vertebrates. Whereas TMEM14C was identified to coexpress with the core machinery of heme biosynthesis and its knockdown causes anemia in zebrafish, TMEM14A was described to stabilize mitochondrial membrane potential and thereby inhibit apoptosis in a yeast system. However, their exact biological function is still unknown. Since animal mitochondria are the site for FA-degradation via ß-oxidation, a role for TMEM14 proteins in FA/lipid homeostasis, energy metabolism or disease (e.g., apoptosis) in vertebrates might be possible.

29Levels and subcellular distribution of free FAs and polar lipids in Arabidopsis FAX1 mutants mainly correlated with a FA-export function, by which FAX1 influences cellular FA and acyl lipid homeostasis (for overview, see). Most likely because of their toxicity and high metabolic fluxes for primary metabolites, changes in free FAs were not very pronounced. However, very-long–chain FAs (C20), which are elongated outside plastids and thus require previous export of C16–18 FAs, were significantly reduced in fax1 knockouts (Table 3). According to acyl-ACP synthesis rates and specificity of thioesterases in Arabidopsis chloroplasts, oleic acid (C18:1) is the major free FA exported from chloroplasts, followed by palmitic acid (C16:0) and only very little amounts of stearic acid (C18:0; compare). FAX1 in yeast assays performed best for FA 16:0 (determined specificity range: 16:0 > 18:1 ~ 18:0 ~ 18:3) and thus, most likely, mainly is involved in the plastid export of free palmitic acid but also can transport oleic acid, which at the stromal side of the plastid IE is provided at highest substrate concentrations. The fact that in yeast, FAX1 was also able to transport α-linolenic acid (C18:3), which in planta is retained inside plastids, indicates that the protein does not discriminate between different degrees of unsaturation, but in general prefers C16 over C18 FAs. In chloroplast IE membranes, FAX1 most likely functions in a passive, carrier-like mode, driven by concentration gradients of free FA substrates (see above). Interestingly, accumulation of export-directed C16–18 FAs in flower tissue of fax1 knockouts (+0.24 > +0.04 > +0.01 mol% for 16:0 > 18:0 > 18:1), reflect the substrate specificity range of FAX1 in yeast (compare S3A Table and S4 Fig.). Furthermore, non-exported FA 18:3 significantly decreased (0.14 mol%; S3A Table) in flowers of fax1 knockouts, thereby maybe pointing to stronger fluxes of FAs into plastid-intrinsic pathways (e.g., synthesis of oxilipin hormones), due to a block in FA export via FAX1.

30Besides changes in free FAs, 65% of the differential lipid patterns depicted in Table 3 underline the hypothesis of plastid FA-export via FAX1, best documented by the strong reciprocal changes in TAG oils. Here for almost 90% of all significantly distributed TAGs (compare S4 Table), the pattern in both FAX1 mutants and tissues perfectly matched to a FA-export function of FAX1. The distribution of 34:x glycolipids (MGDG, DGDG, SQDG), which increased in fax1 knockouts but decreased in FAX1 over-expressors, also corresponded to our theory. In this case, we can, however, not exclude a contribution of ER-made species, since the diacylglycerol backbone for the “34”-glycolipids can originate both, from prokaryotic (from plastids) and eukaryotic (from the ER) phospholipid precursors, respectively. Yet, Arabidopsis is a so-called 16:3 plant, which for galactolipids prefers the prokaryotic pathway with high levels of 16:3 acyl chains. In contrast, ER-derived “34” DAG-backbones contain 16:0 saturated acyl moieties (compare). Thus, we can assume that MGDG 34:x and DGDG 34:x with more than four desaturated C-bonds are completely synthesized in plastids. For the strong MGDG 34:x reductions in FAX1ox leaves and flowers (2.8 and 0.5 mol%, Table 3) and the increase in fax1 flowers (+0.7 mol%) indeed 34:4, 34:5 and 34:6 are the major contributing species (see S2 Table, S3 Table) and therefore support our hypothesis. Most likely at least the abundant forms of phosphatidyl-glycerol (PG 34:3, 34:4) are exclusively made inside plastids as well (see), and thus the pronounced overall increase of PG in fax1 leaf tissue (+3.2 mol%) also mainly is due to a block of FA-export via FAX1.

31Our assumption that FAX1 mediates plastid FA-export is further confirmed by a large decrease of PC-levels in fax1 knockout tissues (up to 8.8 mol%) and a strong increase of PC in FAX1ox leaves (+3.0 mol%). However, also, considerable contrasting evidence is found for three ER-made lipids in flower tissue (i.e., +0.9–1.0 mol% MGDG 36:x, DGDG 36:x in fax1; -5.6 mol% PC in FAX1ox). The latter findings that only apply to lipid species synthesized in the cytosol/ER of flowers might be explained by the inhomogeneity of mature flowers, consisting of leaf, stalk, pollen, ovary, and seed/silique tissues, and/or by a preferential flow of FA-building blocks for lipids into TAG oils during seed development. Furthermore, for the bilayer-forming DGDG 36:x, a plastid export is described to act as surrogate lipid for the lack of PC at, e.g., phosphate-limited growth conditions (see and references therein). Thus, the observed increase of DGDG 36:x species in fax1 knockout mutants might compensate for the strong decrease of PC in the same tissues (compare Table 3).

32In summary, levels and subcellular distribution of free FAs and polar lipids in Arabidopsis FAX1 mutants mainly support a plastid FA-export function of FAX1. In addition, we can, however, not exclude plastid FA-export via different mechanisms or a bypass by other plastid FAX proteins (see below). Due to this potential functional redundancy of plastid FAX proteins, mutation of FAX1 alone does not affect all lipid species present in plants. Effects in leaf tissue, in particular of fax1 knockouts are somewhat more straightforward and stronger than in FAX1 over-expressing lines. The latter is not unexpected for mutation of a protein involved in transport, which is highly expressed in leaf tissues (see below and S5 Fig.). However, the impact of FAX1 mutation on TAG-oil levels might be of future biotechnological importance. Interestingly, already the enhanced FA-transport by FAX1 was able to significantly increase TAG contents in leaves and flower tissues. Furthermore, our finding is in line with higher TAG when FA-loading to the ER in seeds is improved by over-expression of the ABC transporter ABCA9. Thus, coupling of the bottlenecks FA-transport (e.g., FAX1, ABCA9) with those of FA-synthesis and acyl-editing processes and a seed-specific expression might boost plant oil production in future approaches.

33Transcripts of At-FAX1 are present in all developmental stages and peak in leaf tissues (cotyledons, rosette, caulinary, senescent leaves, and flower sepals) as well as in early pollen development (S5A–C Fig.). Consequentially, the strongest phenotype of fax1 knockout mutants can be observed during growth (e.g., reduced rosette leaf size and biomass) and in particular in pollen grains, leading to almost complete male sterility due to the absence of pollen cell walls and impaired pollen release by anthers. For FAX1, we propose a function in FA-export from plastids of tapetum cells in anthers, which in fax1 knockouts leads to the strongly impaired assembly of exine layers and pollen coat, most likely because of the lack of FA-precursors for sporopollenin and/or tryphine synthesis (for a detailed description, see S2 Text).

34Since FAX1 in Arabidopsis belongs to a family of seven proteins, the plastid-predicted FAX2, 3, and 4, whose expression is regulated throughout plant development as well (S5 Fig.), most likely can bypass the loss of FAX1 function in all tissues and organs, leading to the rather mild overall phenotype of fax1 knockouts. Especially in seed tissue (S5D, E Fig. and S6 Fig.), FAX2, 3, and 4 most likely play a more prominent role than FAX1. Indeed transcripts for FAX2 and FAX3—i.e., the highest plastid FAX genes in seed development and germination—showed to be significantly up-regulated (1.13- and 1.24-fold) in fax1 knockout flowers. Please note that with a relative signal of 1585 in wild type, FAX2 and FAX3 are strongly expressed in flower tissue we used for microarray analysis (among highest 9% of all genes on the chip; compare microarray dataset E-MTAB-3090 at www.ebi.ac.uk/arrayexpress).

35The function of FAX1 in FA-delivery for pollen cell wall as well as for cuticular wax assembly is further underlined by differential gene expression in FAX1 mutants (see S1 Text, S5 Table, S6 Table), and by the occurrence of phenotypes similar to fax1 when biosynthesis pathways for FA/lipid-derived precursor material are mutated in Arabidopsis. These include a plastid-intrinsic fatty-acyl-ACP reductase (AlcFAR2/MS2), involved in primary fatty alcohol synthesis for anther cuticle and pollen sporopollenin formation; as well as cytochrome P450 enzymes (CYP703A2, CYP704B1;) that hydroxylate FAs, and the ACS ACOS5 that activates FAs for sporopollenin synthesis in the cytosol of anther cells. Furthermore, several long-chain ACS (LACS1, 2, 4) are necessary to activate long-chain and very-long–chain FAs for building of cutin and wax as well as pollen exine layers. In addition, several ABC transporters in the plasma membrane are required for deposition of surface lipids, displaying fax1-like phenotypes upon mutation: ABCG26 for pollen exine formation from tapetum cells, as well as ABCG11, G12, and G13 in lipid export from epidermis cells for formation of cuticular wax layers (for overview, see). As for FAX1, pathways for synthesis of precursors of pollen cell wall and cutin/wax components often overlap. In stems of fax1 knockouts, we further identified strong regulation of two genes involved in wax biosynthesis: AlcFAR3/CER4 and CYP96A15/MAH1 (S6 Table). Because the latter enzyme is catalyzing synthesis of wax ketone components, its differential expression is in line with the observed lack of C29 ketones in fax1 knockout stems.

36Besides deranged acyl lipid homeostasis, obviously also carbohydrate and cellular/cell wall metabolism was affected in FAX1 mutants as reflected by the impact on plant biomass production and differential regulation of gene expression (see S1 Text, S7–S10 Figs.). In general, we assume that these effects are rather secondary and most likely might result from still unknown signaling events, due to changed FA/lipid homeostasis. Since fax1 knockout plants are short of energy-rich lipids, they most likely turn down anabolic carbohydrate metabolism required for polysaccharide synthesis, resulting in, e.g., reduced biomass and secondary cell walls. The opposite effect is observed in FAX1 over-expressors, in which an excess of lipids most likely leads to more biomass and the production of additional cell layers in stems. These observations clearly demonstrate regulation of energy metabolism and a close correlation between the availability of FAs/lipids and the utilization of carbohydrates in growth processes. This link is further underlined by the finding that the flow of carbon into oil can be promoted by activating plastid FA synthesis and repressing starch synthesis. In this context FAX1, to our knowledge, is not only the first membrane protein identified that mediates FA-export from plastids, but FAX1 and its relatives represent key transport proteins and thus—together with enzymes of FA/lipid-synthesis and modification—might provide powerful future tools to modulate plant lipid and bioenergy production.

37Experiments were performed on Arabidopsis thaliana ecotype Columbia (Col-0, Lehle Seeds; Round Rock, United States). The T-DNA insertion lines SAIL_66_B09 (fax1–1) and GABI_599E01 (fax1–2) were purchased from NASC (Nottingham Arabidopsis Stock Center, Nottingham, United Kingdom) and GABI-Kat (MPI for Plant Breeding Research, Köln, Germany), respectively. To generate complementation lines of fax1–2 and over-expressing At-FAX1 under the control of the 35S promoter, the coding sequence of At-FAX1 was subcloned into pH2GW7. At-FAX1/pH2GW7 was transformed into Agrobacterium tumefaciens GV3101, which was used to transfect heterozygous fax1–2 and Col-0 plants as described. Arabidopsis seeds were sown on soil, vernalized at 4°C in the dark for two days, and grown in a 16 h light (22°C; 100 μmol photons ⋅ m–2 ⋅ s–1) and 8 h dark (18°C) cycle.

38At-FAX1 cDNA was purchased as SSP pUNI51 clone U12755. The corresponding mRNA (NCBI reference sequence NM_15588) is predicted to be 1,030 bp long, including 180 bp 5´- and 169 bp 3´-untranslated regions (UTRs; S1A Fig.). For amplification of FAX1 from pea, RT-PCR was performed using pea seedling cDNA as template and oligonucleotide primers designed according to a pea EST contig sequence. The corresponding mRNA molecule was 1,115 bp long, with 143 bp 5´UTR, 699 bp coding region, and 273 bp 3´UTR (GenBank accession no. KF981436). For primer sequences, see S7 Table; for amino acid sequences, see Fig. 1A.

39To generate a fusion of GFP to the preprotein At-FAX1, the coding sequence was subcloned into the p2GWF7 plasmid. p2GWF7 provides a fusion of GFP to the C-terminal end of the respective proteins, which are expressed under the control of the constitutive 35S promoter. Transformation and analysis of Arabidopsis mesophyll protoplasts was performed as described. GFP fluorescence was detected at 672 to 750 nm and chlorophyll autofluorescence was monitored at 503 to 542 nm by confocal laser scanning microscopy (Leica TCS SP5/DM 6000B, argon laser, excitation wavelength of 488 nm).

40Pea chloroplasts isolated from leaf tissue of 10-day-old pea plantlets were sub-fractioned into OE and IE membranes, stroma and thylakoids as described. Chloroplast envelopes, total protein extracts, and microsomal membranes from Arabidopsis plants were prepared as specified in and, respectively. FAX1 antisera were raised in rabbit (Pineda Antibody Service, Berlin, Germany) against N-terminal peptide sequences of At-FAX1 (17 aa) and Ps-FAX1 (18 aa), respectively (see Fig. 1A). Antisera for marker proteins were produced as described previously. Appropriate amounts of organellar or total cellular proteins were separated by SDS-PAGE, transferred to PVDF membranes and subjected to immunoblot analysis using primary antisera in 1:250 to 1:5000 dilutions in TTBS buffer (100 mM Tris-HCl pH 7.5, 150 mM NaCl; 0.2% Tween-20; 0.1% BSA). Non-specific signals were blocked by 3% skim milk powder and 0.1% BSA. Secondary anti-rabbit IgG alkaline phosphatase antibodies (Sigma-Aldrich) were diluted 1:30,000. Blots were stained using the alkaline phosphatase reaction with 0.3 mg/ml nitroblue tetrazolium (NBT) and 0.16 mg/ml bromochloroindolyl phosphate (BCIP) in 100 mM Tris pH 9.5, 100 mM NaCl, 5 mM MgCl2.

41Genomic DNA of the T-DNA insertion lines fax1–1 and fax1–2 was screened by PCR genotyping. To identify plants with T-DNA insertion in both At-FAX1 alleles (homozygous), combinations of gene-specific primers that flank the predicted insertion sites with each other and with T-DNA-specific left border (LB) primers (S7 Table) were used. Positions and orientations of T-DNA inserts and oligonucleotide primers in fax1–1 and fax1–2 are shown in S1A Fig. To verify insertion sites, PCR-genotyping products were sequenced. T1 generations of generated FAX1 over-expression and complementaion lines were selected by hygromycin (30 μg/ml). Stable insertion of 35S::FAX1 was controlled by PCR-genotyping in all subsequent generations. Therefore, a vector-specific primer in combination with a FAX1 cDNA specific primer was used (S7 Table). In the T2 generation, complementation lines were selected for homozygous alleles of the original T-DNA insertion in fax1–2 (see above), resulting in lines Co#7 and Co#54. For FAX1 over-expression in Col-0 background, we selected the lines ox#2 and ox#4 in the T2 generation.

42For microscopic analysis we used 5-week-old plants and dissected anthers from mature flowers or cut 1–2 mm2 stem segments 1 cm above the bottom of the second internode of the primary inflorescence stalk. We analyzed four individual fax1–2 knockouts, two of each Co#7, Co#54 complementation lines, and five Col-0 wild-type plants for anthers/pollen grains, and pictured stem tissue of independent fax1–1, fax1–2 knockouts, three ox#2, four ox#4 over-expressors, as well as seven individual Col-0 wild-type plants, respectively. Tissue was fixed immediately after harvest with 2.5% (w/v) glutaraldehyde (4°C, at least 24 h) in 75 mM cacodylate buffer (2 mM MgCl2, pH 7.0), rinsed several times with fixative buffer, and subsequently post-fixed with 1% (w/v) osmium tetroxide for at least 2.5 h in fixative buffer at 20°C. After two washing steps in distilled water, samples were stained with 1% (w/v) uranyl acetate in 20% acetone, dehydrated with a graded acetone series and embedded in Spurr’s low viscosity epoxy resin. For light microscopy, semithin-sections (1–2 μm) were cut with a glass knife (Pyramitome 11800, LKB). Ultrathin-sections (50–70 nm) for transmission electron microscopy were prepared with an ultramicrotome (EM UC6, Leica) and post-stained with aqueous lead citrate (100 mM, pH 13.0). Micrographs were taken at 80 kV with a 268 electron microscope (Fei Morgagni).

43The second to fourth internode region of primary inflorescence stalks from 7-week-old plants was used for wax and cutin analyses. For each replicate, stem segments from three to four individual plants were pooled, and samples were provided from two independent harvests of each FAX1 mutant line and respective wild-type controls. Determination of wax and cutin coverage of stems was essentially carried out as described previously. Wax was extracted in chloroform and C24 alkane was added as internal standard. For cutin analysis, exhaustively extracted stems (1:1; methanol:chloroform) were transesterified using methanolic HCl, and cutin monomers were extracted in hexane containing C32 alkane as internal standard. Gas chromatographic and mass spectrometric analysis was carried out after derivatization of extracted wax and cutin with pyridine and BSTFA.

44For each independent harvest (2-times for fax1 knockout, 4-times for FAX1 over-expressing lines) cauline leaves and flowers (stage 10–15, according to) were sampled from at least ten individual, 7-week-old plants and grinded in liquid nitrogen. To be able to work on tissue of identical sample pools (i.e., from 7-week-old plants) for wax/cutin analysis, FA/lipid determination, and transcript profiling, as well as because FAX1 is highly expressed in cauline leaves (see S5A Fig.), we chose the latter instead of old rosette leaves. Tissue powder of each harvest was portioned into three aliquots of 50 mg, which were used to determine polar lipid and free FA contents. For details on data analysis, see S1 Table. Lipids/FAs were extracted from six (fax1 k.o) to 12 (FAX1 over-expressors) biological replicates using 1 ml of a pre-cooled (−20°C) methanol:methyl-tert-butyl-ether (1:3) mixture, spiked with 0.1 μg/ml PC 34:0 (17:0, 17:0) as internal standard. The samples were incubated for 10 min at 4˚C, followed by another 10 min incubation in an ice-cooled ultrasonication bath. After adding 650 μl of UPLC grade water:methanol (3:1), the homogenate was vortexed and centrifuged for 5 min in a table top centrifuge. The addition of water:methanol leads to a phase separation producing an upper organic phase, containing the lipids, and a lower phase containing the polar and semi-polar metabolites. The upper organic phase was removed, dried in a speed-vac concentrator, and re-suspended in 500 μl buffer B (see below) and transferred to a glass vial. 2 μl of this sample were injected onto a C8 reversed phase column (100 mm × 2.1 mm × 1.7 μm particles BEH C8, Waters), using a Waters Acquity UPLC system. The two mobile phases were water (UPLC MS grade, BioSolve) with 1% 1 M NH4Ac, 0.1% acetic acid (buffer A), and acetonitrile:isopropanol (7:3, UPLC grade BioSolve) containing 1% 1 M NH4Ac, 0.1% acetic acid (buffer B). The gradient separation, which was performed at a flow rate of 400 μl/min, was 1 min 45% A, 3 min linear gradient from 45% A to 35% A, 8 min linear gradient from 25 to 11% A, 3 min linear gradient from 11% A to 1% A. After washing the column for 3 min with 1% A the buffer was set back to 45% A and the column was re-equilibrated for 4 min (22 min total run time). Mass spectra were acquired as described using either an Orbitrap Exactive mass spectrometer (Thermo-Fisher) for fax1 knockout lines or a Waters Synapt G1 (Waters) for FAX1 over-expressors, and corresponding wild types, respectively. The spectra were recorded using altering full scan and all-ion fragmentation scan mode, covering a mass range from 100–1,500 m/z. The resolution was set to 10,000 with 10 scans per second. Spectra were recorded from min 0 to min 20 of the UPLC gradients. The analysis of the spectra (alignment, peal picking, normalization and peak integration) was performed with the software package CoMet 2.0 (Nonlinear Dynamics) according to the instructions of the vendor.

45For growth assays in yeast, the coding sequence of the mature At-FAX1 protein was subcloned into the yeast expression plasmid pDR195 (XhoI/BamHI). Therefore, we fused the open reading frame of the predicted mature At-FAX1, starting with aa 34 of the preprotein, behind an “ATG” base triplet by PCR amplification. The yeast mutant strains fat1 (LS2020-YB332) and faa1/faa4 (LS1849-YB525) are specified in. Both strains were transformed with mature At-FAX1/pDR195 and the vector control pDR195 as described. If not denoted elsewhere, liquid cultures of the respective yeast cells were grown to exponential phase in synthetic defined medium (SD-ura), containing 0.1% (w/v) glucose, 0.7% (w/v) yeast nitrogen base without amino acids, and necessary auxotrophic amino acids without uracil. Subsequently, 2 μl drops of the cultures were spotted in different dilutions onto SD-ura plates (2% agar), supplemented with 3.6 mM α-linolenic acid (0.1%, w/v in ethanol), and 1% tergitol (to increase α-linolenic acid solubility). For control plates, an equal amount of the solvent ethanol was added instead of α-linolenic acid. Assays in the presence of cerulenin were performed according to in SD-ura media supplemented with 2% (w/v) glucose, 0.5% Brij 58, 0.7% KH2PO4, 10μM cerulenin and either 100μM of palmitic, stearic, oleic or α-linolenic acid. Growth of yeast cells on solid media was documented between 2 to 6 days at 30°C. OD600 measurements were performed in identical liquid media, inoculated to a starting OD600 of 0.05 or 0.06/0.03 for cerulenin experiments with the respective yeast cells. Cultures were continuously shaken at 30°C and the OD at 600nm was determined at indicated time points.

46Tissue from flowers (stage 10–15 according to, compare S3 Fig.) and from the second to fourth internode of primary inflorescent stalks for each harvest was pooled from more than ten individual, 7-week-old plants (identical sample pool for lipid analysis) and used for preparation of RNA samples by the Plant RNeasy Extraction kit (Qiagen). RNA (200 ng) of four or five independently harvested samples (n = 4–5) from wild type (Col-0 and WT2, segregated from heterozygous fax1–2), fax1 knockout (fax1–1 and fax1–2 lines) as well as FAX1 over-expressors (ox#2 and ox#4 lines) was processed and hybridized to Affymetrix GeneChip Arabidopsis ATH1 Genome Arrays using the Affymetrix 3´-IVT Express and Hybridisation Wash and Stain kits (Affymetrix, High Wycombe, UK) according to the manufacturer’s instructions. Raw signal intensity values (CEL files) were computed from the scanned array images using the Affymetrix GeneChip Command Console 3.0. For quality check and normalization, the raw intensity values were processed with Robin software default settings as described. Specifically, for background correction, the robust multiarray average normalization method was performed across all arrays (between-array method). Statistical analysis of differential gene expression of mutant versus wild-type samples was carried out using the linear model-based approach developed by. In total, we analyzed the following comparisons (see S7 Fig.): (A) flowers: fax1 knockout (n = 5) versus wild type (n = 5); (B) flowers: FAX1 over-expressors (n = 8, four times each ox#2, ox#4) versus wild type (n = 5); (C) stems: fax1 knockout (n = 4) versus wild type (n = 4). The obtained p values were corrected for multiple testing using the nestedF procedure, applying a significance threshold of 0.05 in combination with the Benjamini and Hochberg false-discovery rate control. All microarray data are available in the ArrayExpress database (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-3090.

47Structural models of At-FAX1 and At-FAX6 were generated by Phyre2, based on alignments with the PDB entries for human TMEM14C (c2losA) and TMEM14A (c2lopA), respectively. Identity of At-FAX1 with its template TMEM14C was 21% and for At-FAX6 with TMEM14A 36%, while confidence of both models was 99.9%, thereby indicating a high confidence and accuracy of the core models. Structural alignments were created with PyMOL.