Critical roles of mitochondrial fatty acid synthesis in tomato development and environmental response

The pathway of mitochondrial fatty acid synthesis participates in tomato morphogenesis via modulating photorespiration, redox homeostasis, and lipid metabolism.

Abstract

Plant mitochondrial fatty acid synthesis (mtFAS) appears to be important in photorespiration based on the reverse genetics research from Arabidopsis (Arabidopsis thaliana) in recent years, but its roles in plant development have not been completely explored. Here, we identified a tomato (Solanum lycopersicum) mutant, fern-like, which displays pleiotropic phenotypes including dwarfism, yellowing, curly leaves, and increased axillary buds. Positional cloning and genetic and heterozygous complementation tests revealed that the underlying gene FERN encodes a 3-hydroxyl-ACP dehydratase enzyme involved in mtFAS. FERN was causally involved in tomato morphogenesis by affecting photorespiration, energy supply, and the homeostasis of reactive oxygen species. Based on lipidome data, FERN and the mtFAS pathway may modulate tomato development by influencing mitochondrial membrane lipid composition and other lipid metabolic pathways. These findings provide important insights into the roles and importance of mtFAS in tomato development.

Introduction

The de novo biosynthesis of fatty acids in plants mainly includes plastid (ptFAS) and mitochondrial (mtFAS) fatty acid synthesis (Rawsthorne, 2002; Yasuno et al., 2004). The pathway of mtFAS has not been fully explored in plants probably because it only produces a small portion of the total lipid in plant cells. Nevertheless, mtFAS is rather underestimated regarding its roles in plant growth and development.

The mtFAS requires mitochondrial acyl carrier protein (mtACP) to carry acyl groups with different number of carbon atoms for different enzymatic reactions. There are three functionally redundant genes encoding mtACP in Arabidopsis (Arabidopsis thaliana), and simultaneously knocking out all of the three genes caused embryo lethality (Fu et al., 2020). The de novo synthesized apo-ACP needs a phosphopantetheinyl transferase (PPT) to convert it into the active form, holo-ACP (Guan et al., 2015). Holo-ACP is then used as substrate to react with malonyl-CoA to produce malonyl-ACP under the action of malonyl-CoA transacylase (MCAT), which serves as the building blocks for the carbon chain extension in mtFAS (Guan et al., 2020). Malonyl-CoA is formed by the catalysis of free malonate and CoA by malonyl-CoA synthetase (MCS) (Guan and Nikolau, 2016). Four core enzymes of mtFAS are involved in the carbon chain elongation: β-ketoacyl-ACP synthase (mtKAS) (Olsen et al., 2004; Ewald et al., 2007), β-ketoacyl-ACP reductase (mtKR) (Guan et al., 2020), 3-hydroxyl-ACP dehydratase (mtHD) (Guan et al., 2017), and enoyl-ACP reductase (mtER) (Guan et al., 2020). These enzymes coordinately add the building blocks of fatty acids (C2 units) sequentially (Hiltunen et al., 2010). In this cycling process, β-hydroxylacyl-ACP is dehydrated by mtHD to form enyl-ACP. The human mtHD was named as 3-hydroxyacyl thioester-ACP dehydratase (HTD) earlier, which belongs to the MaoC-type dehydratase family along with the FabA and FabZ from Escherichia coli (Hiltunen et al., 2009). HTD is further classified into the MaoC dehydratase-like subfamily (Hiltunen et al., 2010). All the members in this subfamily share a “hot-dog domain,” with a MaoC-type protein motif in the domain. The “hot-dog” fold consists of a long hydrophobic α-helix with an antiparallel β-sheet wrapped around it (Leesong et al., 1996; Koski et al., 2004). Functional analysis of the core genes in Arabidopsis based on reverse genetic approach suggested that mtFAS plays an essential role in photorespiration (Guan et al., 2017; Guan et al., 2020).

The metabolites derived from mtFAS are fatty acids with even numbers of carbon atom, mainly C8, C12, C14, and C16. This is because of the characteristics of adding only a C2 unit each time by KAS (Yasuno et al., 2004), which has also been confirmed in plants in the study of mtKAS (Guan et al., 2017). Among them, the C8 octanoate is the most abundant species which is converted into lipoic acid under the action of mitochondrial lipoic acid synthase (Yasuno and Wada, 1998). Lipoic acid has been the focus of mtFAS research in animals during the past two decades and in plants during the last decade (Brody et al., 1997; Ewald et al., 2007). It acts as the cofactor of various enzymes in plants, such as pyruvate dehydrogenase (PDH), α-ketoglutarate dehydrogenase (KGDH), branched-chain keto acid dehydrogenase (BCKDH), and the glycine decarboxylase complex (GDC) (Reche and Perham, 1999; Gueguen et al., 2000; Taylor et al., 2004). However, studies in animals suggest that the function of mtFAS is not limited to the synthesis of lipoic acid (Nowinski et al., 2020). The roles of mtFAS in development and metabolism are rather poorly understood and the research work on mtFAS in plants is exclusively in Arabidopsis and limited to photorespiration-associated phenotypes.

Here, we characterized a spontaneous tomato (Solanum lycopersicum) fern-like mutant displaying a variety of developmental defects which was caused by the mutation of a gene encoding the 3-hydroxyl-ACP dehydratase of mtFAS. Our results revealed that mtFAS is not only involved in photorespiration, but also contributes to tomato morphogenesis by affecting energy supply, reactive oxygen species (ROS) homeostasis, and lipid profiles.

Results

The tomato fern-like mutant shows various developmental defects

The fern-like mutant was derived from tomato variety Alisa Craig (S. lycopersicum var. AC) during tissue culture. This spontaneous mutant displays pleiotropic phenotypes, including dwarfism, yellowing, curly leaves, and increased axillary buds (Figure 1A). The new leaves of the mutant resemble a spiral fern; thus, we designated this mutant fern-like. Under normal growth conditions, the new leaves of the wild-type (AC) are explanate (Figure 1B), however, new leaves curl inward obviously in the mutant (Figure 1C). At the seedling age of 40 days, the height of fern-like plants is only about a quarter of that of AC (Figure 1D). Several times the number of axillary buds can be observed on the mutant plants (Figure 1E). More than 70% of the leaves are twisted irregularly in the mutant (Figure 1F). The phenotypic changes of the mutant remind us of diseased tomato plants caused by certain viruses, such as cucumber mosaic virus (CMV) and tomato yellow leaf curl virus (TYLCV). Therefore, we tested some physiological indicators including chlorophyll content, ascorbate peroxidase (APX) activity, and malondialdehyde (MDA) level. Compared with AC, the total chlorophyll content in the mutant decreased by about one-third (Figure 1G). A significantly higher level of APX activity (Supplemental Figure S1A) and MDA were detected in the mutant (Supplemental Figure S1B). Overall, the fern-like mutant shows developmental defects in various aspects and significant changes in stress-responsive parameters tested.

Figure 1.

Phenotypic characteristic of the fern-like mutant. A, The fern-like mutant and wild-type (AC) at the age of 40 days. Bar = 5 cm. B and C, Morphology of new leaves for AC (B) and fern-like (C). Seedling at the age of 7 days was photographed with stereomicroscope. Bar = 1 mm. D–G, The plant height (D), number of axillary buds (E), percentage of curly leaves (F), and chlorophyll content (G) for AC and fern-like plants at the age of 40 days. Values are means ± sd of three biological replicates (n=10). Paired, two-tail Student’s t test, **P < 0.01.

FERN encodes a MaoC-dehydratase

To positionally clone FERN, we constructed an F₂ population using fern-like mutant and a wild tomato accession (LA1375, S.pimpinellifolium) as the parents. A total of 450 F₂ individuals were used for primary mapping. We identified 348 seedlings with AC phenotype and 102 seedlings with fern-like phenotype. The segregation fits a 3:1 pattern, indicating that fern-like is a single-gene recessive mutant. We applied a BSA-RNA seq analysis with a bulk of 30 F₂ plants showing fern-like and another bulk of 30 F₂ plants showing AC phenotype. This allowed us to locate FERN at the end of chromosome 6 (Figure 2A), with the physical region between 43038966 and 48276923 bp (Tomato genome, version 2.5, Figure 2B). To narrow down the location of FERN, markers were developed based on the resequencing data of AC and LA1375, and recombinants were detected in the 102 recessive plants by polymorphic markers. With this, FERN was mapped within a region of about 300 kb containing 30 genes. Later, the population was increased to 1,000 F₂ plants, and 212 recessive plants were observed and used for further mapping. Finally, FERN was fine-mapped to a 170-kb region containing 19 genes using two tightly linked makers (46.48 and 46.65) (Figure 2C and Supplemental Table S1).

Figure 2.

Positional cloning of FERN and complementation test. A and B, FERN was mapped on chromosome 6 (A) at the position between 43 and 48 Mb (B) using BSA-RNA seq (BSR) analysis. C, FERN was narrowed down to 170-kb region between 46.48 and 46.65 Mb by markers 46.48 and 46.65 (Named after their corresponding physical position). Gray bar, number above bar, number below bar, and red box represent chromosome, physical position of marker, number of recombinants, and the final mapping interval, respectively. D, Gene structure of the candidate gene FERN (Solyc06g075000). The final mapping interval contains 19 genes (represented by arrows, FERN in black arrow). A single base substitution (G–A) occurred in the third exon of FERN, resulting in a cysteine-to-tyrosine (Cys–Tyr) transition in fern-like mutant. The untranslated regions and exons are shown as colored rectangles, and the introns as black lines. E, Genetic complementation of fern-like mutant. Genomic DNA of FERN together with its 2 kb promoter was cloned from the wild-type (AC) and transformed into fern-like mutant. Bar = 3 cm. F, Protein level of FERN detected by Western blotting. Actin served as the internal control. G, Quantification of FERN protein level. Quantification was performed using ImageJ program. Values are means ± sd of three biological replicates (n=10). Paired, two-tail Student’s t test, **P < 0.01.

To find the candidate gene underlying fern-like, a comparative analysis of the resequencing data of the mutant and AC was performed, and four genes were detected with mutations in the mapped interval. Three out of the four mutations were located in non-coding regions and only one occurred in the coding region of gene Solyc06g075000 (Supplemental Table S1). This gene is annotated as MaoC dehydratase and a G–A substitution was detected in the third exon of the coding region. This substitution results in a Cys-to-Tyr conversion of amino acid at position 133 (Figure 2D). The G–A mutation was verified by DNA sequencing (Supplemental Figure S2). Therefore, Solyc06g075000 was selected as the candidate gene and designated SlFERN. Not surprisingly, the transcriptional level of SlFERN had no significant difference between AC and fern-like mutant (Supplemental Figure S3).

To verify the candidate gene, a genetic complementation experiment was carried out. The genomic DNA of SlFERN and its promoter was amplified from AC to construct the complementation vector which was introduced into fern-like mutant. Mutant phenotypes can be checked on the transgenic plantlets at the rooting stage (Supplemental Figure S4). The phenotypes of fern-like were fully reversed in complemented lines (Figure 2E), indicating SlFERN is the correct candidate.

To evaluate the influence of amino acid conversion on protein abundance, the polyclonal antibody (anti-FERN) was obtained by expressing a truncated FERN (1–132 aa) in E. coli and immunizing rabbits. The result of Western blotting showed that the endogenous FERN level in fern-like was only about 46% of that in the wild-type (Figure 2, F and G). Overall, our work revealed that a single base mutation occurs in SlFERN, leading to a Cys-to-Tyr conversion and a reduced level of FERN protein.

FERN participates in mtFAS

To functionally characterize tomato FERN, we first performed a phylogenetic analysis of FERN. It showed that FERN homologs are relatively conserved, exist in dicotyledonous and monocotyledonous plants, bacteria, and human beings (Supplemental Figure S5) and have a conserved MaoC dehydratase domain. Five members of the FERN family were identified in Arabidopsis but only two were found in tomato (Figure 3A). Among them, Arabidopsis mtHD (AT5G60335) showed the highest similarity to FERN (62%), followed by ECH2 (AT1G76150) (31%). mtHD is located in mitochondria and involved in fatty acid synthesis (Guan et al., 2017), while ECH2 is located in peroxisome and involved in the β-oxidation of fatty acids (Strader et al., 2011). Online domain prediction showed that both FERN and mtHD possess a MaoC dehydratase domain and a MaoC dehydrat_N domain with most sequences overlapping with each other. Although ECH2 also has a MaoC dehydratase domain, its motif composition and arrangement are clearly different (Supplemental Figure S6). To predict the subcellular location of FERN, Wolf PSORT and TargetP software were used. Both predictions showed that FERN contains a mitochondrial signal peptide. The Arabidopsis homolog mtHD is also predicted to be a mitochondrial protein. To confirm this, the full-length or truncated (1–40 aa) coding sequence of mtHD and FERN was fused with that of green fluorescent protein (GFP) and transiently expressed in Nicotiana benthamiana leaves. Confocal microscopic images showed that the green fluorescence signal of the fused protein overlapped with the red fluorescence signal of the mitochondrial marker MT-RK (Nelson et al., 2007; Figure 3B and Supplemental Figure S7). Therefore, FERN and mtHD are located in mitochondria and the result of mtHD localization is consistent with a previous report (Guan et al., 2017).

Figure 3.

FERN encodes a mitochondrial 3-hydroxyl-ACP dehydratase in mtFAS. A, Phylogenetic relationships of FERN homologs between tomato and Arabidopsis. The cladogram tree was constructed using neighbor-joining method in MEGA7, with bootstrap support values (%) from 1,000 replicates. Schematics on the right show the different protein domains (colored rectangles) in sequences from the phylogenetic tree. mtHD, Arabidopsis mitochondrial 3-hydroxyacyl-acyl carrier protein dehydratase and ECH2, enoyl-CoA hydratase 2. B, Mitochondrial localization of FERN and mtHD. Plant expression vectors containing the full-length coding sequence (CDS) of SlFERN (Solyc06g075000) or mtHD (AT5G60335) fused with the GFP gene were transiently expressed in epidermal cells of N. benthamiana leaves. MT-RK (Nelson et al., 2007) fused with the RFP (red fluorescent protein) gene served as the mitochondrial tracker (Mito-tracker, colored). Bar = 25 μm. C, Arabidopsis mtHD functionally complements fern-like. The promoter of FERN was used to drive mtHD and introduced into fern-like mutant. AC = wild-type. Bar = 5 cm.

To check whether FERN functions as a 3-hydroxyl-ACP dehydratase as mtHD, the native promoter of SlFERN was used to drive mtHD and introduced into the fern-like mutant. As early as the rooting stage, it was found that mtHD could fully recover the fern-like mutant (Figure 3C). Arabidopsis mtHD has been shown to be a key enzyme for carbon chain elongation in mtFAS, which dehydrates 3-hydroxylacyl-ACP to form trans-2-enoyl-ACP. In conclusion, FERN functions as a 3-hydroxyacyl-ACP dehydratase in the mtFAS of tomato.

FERN modulates tomato morphogenesis via photorespiration

The leaves of fern-like mutant were yellow and small, and most leaflets of the compound leaf were twisted and curled inward (Figure 4A). Therefore, paraffin sectioning was used to check the cell structure of leaves. Microscopic observation revealed that the mesophyll cells of the palisade tissue in fern-like mutant were reduced or even disappeared, and most mesophyll cells looked like the irregular cells of spongy tissue (Figure 4B). When checked with transmission electron microscope (TEM), it was found that the number of chloroplasts in the mesophyll cells of fern-like was reduced obviously. Further, the chloroplasts in fern-like were flat with fewer starch grains when compared with that of AC (Figure 4C). No clear difference was observed in mitochondrial morphology between fern-like and AC (Figure 4D).

Figure 4.

CO₂ enrichment and low light alleviate fern-like phenotype. A, Representative compound leaf from wild-type (AC) and the mutant. The third true leaf from bottom was photographed at the plant age of 30 days. Bar = 1 cm. B, Leaf anatomy. Leaf pieces without veins were prepared from leaves as shown in (A). Longitudinal sections stained with toluidine blue are shown. Note reduced palisade tissues in the leaf of fern-like mutant. Bar = 75 μm. C–D, TEM images of chloroplast (C) and mitochondria (D). For easy identification, a starch grain (S) is labeled. Bar = 1 μm (C) or 0.4 μm (D). E, Representative plants grown in ambient air or with 1% CO₂ for 20 days. Ten-day-old seedlings were submitted to treatments. Bar = 5 cm. F, Plant height after CO₂ treatment. Values are means ± sd of three biological replicates (n=5). G, Leaf anatomy of plants grown under 1% CO₂ treatment. Sampling and sectioning were the same as (B). Bar = 50 μm. H, The change of glycine (Gly) and serine (Ser) content in fern-like mutant grown in ambient air. Values are means ± sd of five biological replicates (n = 15). I, Leaf chlorophyll content of plants under 1% CO₂ treatment. Values are means ± sd of three biological replicates (n = 5). J–N, Representative image (J), plant height (K), leaf chlorophyll content (L), F_v/F_m image (M), and F_v/F_m ratio (N) of plants grown under different levels of light for 20 days. Ten-day-old seedlings were submitted to treatments. Bar = 5 cm (J) or 1 cm (M). Values are means ± sd of four biological replicates (n = 6). Paired, two-tail Student’s t test, *P < 0.05, **P < 0.01.

A major function of mtFAS identified previously is to produce lipoic acid, which is a cofactor of GDC in photorespiration (Ewald et al., 2007). To check whether fern-like has alteration in photorespiration, the 10-day-old tomato seedlings of the mutant and AC were treated with 1% CO₂, using ambient air as control (∼0.03% CO₂). It was shown that fern-like can partially recover under higher CO₂ conditions (Figure 4E), especially, the plant height was close to that of the wild-type (Figure 4F). In addition, under 1% CO₂, the cell number and the morphology of the palisade tissue are close to that of the wild-type (Figure 4G). Metabolic analysis showed that the glycine and serine contents in fern-like increased by about 80 and 2 times that of the wild-type, respectively (Figure 4H). These results were similar to those in GDC-related mutants (Ewald et al., 2007; Hoffmann et al., 2013). The chlorophyll content was also increased (Figure 4I). These results indicate that the inhibition of photorespiration can alleviate the developmental defects of fern-like.

A major effect of abnormal photorespiration on plant development is photoinhibition, especially under strong light. Tomato is a photophilic plant species and the appropriate light intensity is 500–700 μmol/(m²·s) in tomato production. The light compensation and saturation points are 40 and 1,200 μmol/(m²·s), respectively (Ilic et al., 2012; Lu et al., 2017). When the 10-day-old seedlings of AC and fern-like were treated with 400 and 80 μmol/(m²·s) light for 20 days, it was found that the growth of mutant was severely inhibited under 400 μmol/(m²·s) light but largely recovered under the weak light treatment (Figure 4J). Under 80 μmol/(m²·s) light, plant height (Figure 4K) and chlorophyll content (Figure 4L) were substantially increased, but there was no substantial difference for AC under the two different levels of light intensity. Chlorophyll fluorescence imaging (Figure 4M) and F_v/F_m data (Figure 4N) showed that the maximum light energy conversion efficiency of photosystem II in fern-like mutant decreased by about 20% at 400 μmol/(m²·s) light, compared with that in AC. However, under weak light conditions, this parameter showed no substantial difference between the two genotypes. In conclusion, the mutant phenotype caused by photoinhibition can be alleviated by either increasing CO₂ concentration or decreasing irradiation.

FERN contributes to energy supply and ROS homeostasis

As 1% CO₂ cannot fully recover fern-like mutant and mitochondria functions as the “energy factory” of plants, therefore, we detected the levels of nicotinamide adenine dinucleotide (NADH), NADH phosphate (NADPH), and adenosine triphosphate (ATP) in tomato leaves. It was found that the contents of NADPH and NADP⁺ in fern-like mutant were only about 40% those of AC (Figure 5A), while the ratio did not change significantly (Figure 5B). The content of NADH in mature leaves decreased by nearly 70% (Figure 5C) and the ratio of NADH/NAD⁺ decreased by 30% (Figure 5D). Regarding ATP content, the level in fern-like mutant was only 30% that of AC (Figure 5E). Taking together, these results indicate that the energy supply is impaired in the mutant.

Figure 5.

FERN contributes to energy supply and redox homeostasis. A–E, Contents of NADP⁺ and NADPH (A), the ratio of NADPH/NADP⁺ (B), contents of NAD⁺ and NADH (C), the ratio of NADH/NAD⁺(D), and the relative ATP content (E) of wild-type (AC) and fern-like plants at 40 days old. The leaflets of the fifth compound leaf from the bottom were used to measure the physiological parameters. Values are means ± sd of five biological replicates (n=5) and each biological replicate contains three technical replicates. F, NBT staining of superoxide anions in the leaves of 30-day-old plants. Bar = 1 cm. G–J, Representative plants (G), ascorbic acid peroxidase (APX) activity (H), contents of GSH and GSSG (I), the ratio of GSH/GSSG (J) for seedlings treated with malic acid or pure water. Fourteen-day-old seedlings were sprayed with either 40 mM malic acid or water once a day for four consecutive days. Photographs were taken 10 days after the treatment (G). Values are means ± sd of four biological replicates (n = 6). Paired, two-tail Student’s t test, **P < 0.01. Bar = 5 cm (G). K, Trypan blue staining of leaves from 14-day-old seedlings. Dead cells are stained. Bar = 1 cm. L, TEM images of organelles and autophagosomes in the mesophyll cells of 30-day-old seedlings. The third leaf from the bottom was sampled for TEM analysis. For easy identification, a peroxisome (p), a mitochondrion (m), and an autophagosome (a) are labeled together with different color arrows. Bar = 2 μm.

In plants, ROS is often accompanied by abnormal redox process of NAD(P)H production or consumption (Moller, 2001). In addition, the mutant phenotype looks somewhat like the symptom of tomato plants infected by certain viruses. We speculated that the phenotype of fern-like could be associated with ROS homeostasis. Firstly, superoxide anion distribution was detected by nitroblue tetrazolium (NBT) staining. The staining signal was strong in the leaves at different nodes in fern-like mutant, while it was barely detected in AC leaves (Figure 5F). Previous studies have shown that ROS can be produced and transmitted between chloroplasts and mitochondria through the malic acid shuttle (Zhao et al., 2018). Therefore, 40 mM malic acid solution was exogenously sprayed onto tomato seedlings at 14 days old and it was found that the fern-like mutant displays a more severe phenotype (Figure 5G). Further analysis revealed that the APX activity in fern-like mutant was nearly twice as high as that in AC under normal conditions, but after malic acid treatment, it was significantly reduced to a level similar to that of AC (Figure 5H). Both the reduced glutathione (GSH) and GSH disulfide (GSSG) contents in the mutant were higher than that in AC. In particular, GSH and GSSG increased significantly in the mutant after malic acid treatment (Figure 5I), while the GSH/GSSG ratio decreased (Figure 5J). Therefore, fern-like plants were in a state of oxidative stress. The oxidative stress was even intensified by malic acid treatment. Meanwhile, the malic acid treatment affected much less on AC plants, as we can see that they gradually recovered 10 days after the treatment. These results indicated that the fern-like mutant was not able to maintain redox homeostasis efficiently.

A high level of ROS can induce cell death. Therefore, we employed trypan blue staining to detect cell death in tomato leaves. The true leaves of fern-like mutant could be partially stained (Figure 5K), indicating that a certain degree of cell death occurred in the leaves of fern-like mutant. Further, the organelle structure in leaves was observed using TEM. A large number of chloroplasts were observed in AC mesophyll cells and the tight coupling of chloroplast, mitochondria, and peroxisome could be seen. Organelle coupling can promote the transmission of signals and energy during photorespiration (Oikawa et al., 2015). However, there were very few chloroplasts observed in the mesophyll cell of fern-like mutant and the close coupling of the three organelles mentioned above could hardly be observed. More importantly, a certain number of autophagosomes were found in fern-like mesophyll cells (Figure 5L). Previous studies have shown that either energy deficiency or ROS accumulation can lead to autophagy (Shibata et al., 2013). These results suggest that FERN is actively involved in tomato development by affecting energy supply and ROS homeostasis.

Multi-omics data revealed global metabolomic changes in fern-like mutant

To further understand the roles of FERN in development, transcriptome, metabolome, and lipidome profiling were performed on fern-like and AC. In the transcriptome profiling, 1,548 differentially expressed genes (DEGs) were detected between fern-like and AC (Supplemental Figure S8). Many KEGG pathways were significantly altered. Among them, photosynthesis, especially antenna protein, was affected the most, with 30 DEGs out of the 33 genes in this pathway. The secondly affected pathway is photosynthetic synthesis, with more than 40 DEGs and mainly related to either photosystem I or II. The third one is metabolic pathway and it has the largest number of DEGs (Figure 6A). The Gene Ontology (GO) enrichment analysis of DEGs identified redox as the most enriched GO biological process and chloroplast-related structures as the most enriched GO cellular component (Supplemental Figure S9).

Figure 6.

Multi-omics data revealed overall changes in fern-like mutant. A, Enriched KEGG pathways for the DEGs between fern-like and wild-type (AC). Rich factor = the number of DEGs in a pathway/number of all annotated genes in the pathway. KEGG pathways were enriched and the top nine pathways were presented. The range of P-value is [0, 1], represented by color. Dot size represents the number of DEGs in a pathway and the color indicates the −log10 (P-value). B, Scatter diagram of metabolic pathway analysis. The pathway impact (horizontal axis) value was calculated from pathway topology analysis and the range is [0, 1]. Dot size represents the number of differentially expressed metabolites in a given metabolic pathway and the color indicates the −log10 (P-value). C, Heatmap of the changes for different lipid subclasses in the mature leaf, mitochondria (Mito), and chloroplast (Chlo) of the fern-like mutant. Lipidomic profiling was conducted on the three types of samples for AC and fern-like mutant, with three biological replicates per sample. /, non-detected or abandoned due to potential traces of cross contamination. LPC, lyso-phosphatidyl choline; MG, monoglyceride; Cert, Cer(T18); and LPE, lyso-PE. D and E, The ratio changes of FFA with different carbon chain length (D) and different double bonds (E) in the mutant. Values are means ± sd of three biological replicates. F, KEGG pathways of the differential lipids in mitochondria of the mutant. The ordinate is the name of KEGG metabolic pathway and the abscissa is the number of metabolites annotated to the pathway and their proportion to the total number of metabolites annotated.

In the non-targeted metabolome profiling, 1,728 and 5,316 metabolites were detected by negative and positive ion mode, respectively, of which 836 and 1,930 were identified/annotated and used for further analysis. The principal component analysis (PCA) demonstrated that the metabolic profiles are clearly different between AC and fern-like, whether in the positive (Supplemental Figure S10A) or negative ion mode (Supplemental Figure S10B), suggesting an overall change of metabolism in fern-like. Pathway enrichment revealed that almost a quarter of the affected pathways were related to amino acid metabolism, including glycine, serine, phenylalanine, tyrosine, alanine, aspartic acid, lysine, arginine, proline, and cysteine (Supplemental Table S2). Based on the degree of influence by differential metabolites, the following metabolic pathways were the most affected: linoleic acid metabolism, isoquinoline alkali biosynthesis, vitamin B6 metabolism, and pyruvate metabolism. Based on the variation amplitude of metabolite content, the following four metabolic pathways were the most affected: metabolism of β-alanine; degradation of valine, leucine, and isoleucine; metabolism of glycine, serine, and threonine; and GSH metabolism (Figure 6B). These results indicated that FERN plays important roles in photorespiration and various metabolic processes including redox, amino acid, and secondary metabolism.

As FERN functions in mtFAS, we compared the lipidome in more detail here. In the lipidome profiling, three different types of samples were compared: mature leaf, chloroplast (Chlo, Supplemental Figure S11), and mitochondria (Mito, Supplemental Figure S12) from the mature leaf. PCA revealed clear differences among the combinations of different genotypes (AC and the mutant) and sample types (Supplemental Figure S13). A total of 751 lipids belonging to 19 different subclasses were detected. Regarding the differential lipids (Log2 (Fold change) ≥ 1.0 or ≤0.5, VIP ≥1.0, See “Materials and methods”) between the mutant and wild-type, there were only 36 common to the three types of samples (Supplemental Figure S14). In chloroplast, the three major components of thylakoid (monogalactosyl diglyceride, MGDG; digalactosyldiacylglycerol, DGDG; and sulfoquinovosyldiacyl glycerol, SQDG) were reduced substantially in the mutant. These results were in line with the abnormal structure and malfunction of chloroplasts in the mutant. In mitochondria, the typical membrane lipid phosphatidylcholine (PC) decreased, while phosphatidyl ethanolamine (PE) increased substantially. Mitochondrial membrane lipids with low abundance were also changed substantially in the mutant. Phosphatidyl serine (PS) and phosphatidic acid (PA) were increased, but phosphatidyl inositol (PI) and phosphatidyl glycerol (PG) were opposite. All these differences suggested that the composition of membrane lipid was overall altered in the mitochondria of fern-like mutant.

Some non-typical membrane lipids of mitochondria or chloroplast were also changed to varied degrees in fern-like leaves. Ceramide (Cer) and cholesteryl ester (CE) were reduced while triglycerides (TG) and diacylglycerol (DG), an important storage lipid and its corresponding precursor, increased in the mutant (Figure 6C). All these differences suggested that many lipids were altered in the fern-like mutant.

We further analyzed free fatty acid (FFA) in fern-like mutant and AC, and found that in the mitochondria of the mutant, the ratios of C10, C12, C14, and C18 decreased to varied degrees compared with those of AC. The ratios of C16, C22, and C24 increased slightly. In chloroplast, the ratios of C12 and C14 increased remarkably. In the leaf of the mutant, C12 was reduced by 60%, but C14 was 2.5 times that of AC (Figure 6D). Compared with chloroplast and leaf, the ratio of saturated FFA in mitochondria increased in the mutant. However, the polyunsaturated FFA decreased in the mitochondria of the fern-like mutant (Figure 6E). Overall, the FFAs, especially the middle chain ones, were substantially affected.

KEGG enrichment analysis was applied on the differential lipids in three types of samples (leaf, Mito, and Chlo) and the enriched metabolic pathways were similar (Supplemental Figure S15). Besides the pathway of glycine, serine, and threonine metabolism involved in photorespiration process, pathways related to membrane remodeling and signaling were also enriched, such as glycerophospholipid metabolism, glycosylphosphatidylinositol (GPI)-anchor biosynthesis, and phosphatidylinositol signaling. In addition, the pathways related to the metabolism of important unsaturated fatty acids were also significantly affected, including linoleic acid and alpha-linolenic acid metabolism. In terms of cellular processes, autophagy was the most affected (Figure 6F).

Through multi-omics analysis, we found that FERN can affect the metabolism of glycine, serine, glyoxylic acid, and dicarboxylic acid through GDC, which was similar to the reports of Arabidopsis mtFAS. Moreover, we also found that the FFA composition and the mitochondrial membrane-associated lipids were changed remarkably in the mutant, suggesting that mtFAS is not only responsible for lipoic acid production, but also can further affect the mitochondrial membrane system.

Discussion

In this work, we applied a forward genetic approach to identify the causal mutation in the spontaneous mutant fern-like. A single base substitution occurred in the candidate gene Solyc06g075000, leading to an amino acid conversion. The candidate gene was confirmed by a genetic complement assay. FERN was found to be a mitochondrial enzyme that functions in fatty acid synthesis. FERN participates in morphogenesis via affecting photorespiration in tomato and it also affects plant development via energy supply and redox regulation. The mtFAS pathway is very important for stabilizing the lipidomic landscape both inside and outside of plant mitochondria.

Cys¹³³ plays an important role in normal function of FERN

A comparison of the amino acid sequences across different species revealed that the cysteine at position 133 (Cys¹³³) of FERN is conserved in plants, but not in animals and microorganisms, suggesting that the cysteine at this site is essential for the normal function of 3-hydroxyl-ACP dehydratase in plants. From a simulation of three-dimensional structure, it was found that Cys¹³³ is located in the β-sheet of the “hot-dog domain” in FERN. Western blotting results showed that FERN protein in the mutant decreased by about 54% (Figure 2, F and G), suggesting that the mutation probably affects the protein stability of FERN. Considering the severe phenotype of fern-like, the remaining 46% protein of FERN^C133Y does not necessarily function normally. Based on the NCBI Conserved Domain Database analysis, the mutation site is near to the Substrate-Binding Tunnel but far from the enzyme activity center, indicating that it is more likely to affect substrate binding. Overall, the Cys¹³³ of FERN may affect protein stability, substrate binding, protein tertiary structure, or protein modification, which requires further investigation.

Malfunction of GDC in fern-like mutant leads to photoinhibition

The first enzyme reported to be involved in mtFAS in Arabidopsis was mtKAS (Ewald et al., 2007), subsequent studies documented other enzymes including mtHD (Guan et al., 2017), leading to the establishment of the pathway in Arabidopsis. A common characteristic of these mutants is that plant development is impaired under ambient air but can be restored under high CO₂ conditions.

Studies in animals and yeast suggest that the main role of mtFAS is to produce lipoic acid (Hiltunen et al., 2010). Lipoic acid is a cofactor of the H subunit of the GDC in photorespiration. Based on the metabolome data, the contents of lipoic acid and dihydrolipoic acid in fern-like were decreased by about 20% and 30%, respectively (Supplemental Figure S16). In mtFAS-related mutants of Arabidopsis, the degree of lipid acylation for H protein was substantially reduced or even completely lost (Ewald et al., 2007; Guan et al., 2020). It appears that mtFAS-derived lipoic acid may not be as important as thought at the whole-cell level. It is possible that mitochondrial fatty acids may only account for a small portion of lipoic acid in plant cells (while plastids can also synthesize lipoic acid) (Wada et al., 1997; Yasuno and Wada, 1998; Hiltunen et al., 2010). We tried the exogenous treatment of lipoic acid on the mutant, but it cannot recover the phenotype. It could be because exogenous lipoic acid cannot be absorbed by the leaf (at least the mutated leaf), or more likely, the lipoic acid outside of mitochondria cannot be transported into the powerhouse (Guan and Nikolau, 2016; Guan et al., 2020).

In photorespiration, glycine is converted to serine under the decarboxylation activity of GDC. Our metabolome data showed that the mutant accumulated a large amount of glycine as well as a higher level of serine (Figures 4F and 6B). These results are consistent with the results of mtFAS-related mutants in Arabidopsis (Guan et al., 2015, 2017, 2020; Guan and Nikolau, 2016).

Photorespiration also contributes to plant energy metabolism. NADH generation accompanies the GDC-involving reactions. Meanwhile, NADH is consumed during the reduction of hydroxypyruvate to glycerate in peroxisome and ATP is consumed during the conversion of glycerate to 3-phosphoglycerate (3-PG). Under illumination conditions, the main source of NADH in mitochondria is from photorespiration and when photorespiration is not inhibited, the NADH produced in mitochondria can either balance or exceed the consumption in peroxisome (Bauwe et al., 2010; Lim et al., 2020). The reduction of NADH/NAD⁺ in fern-like mutant could be a side effect of photorespiration inhibition. In plant mtFAS, both steps involving mtKR and mtER consume NADPH or NADH (Venkatesan et al., 2014). The impairment of mtFAS in fern-like mutant would reduce the consumption of NADPH or NADH, but the amount should be limited as mtFAS is only a minor pathway. Considering that both NADPH and NADH were significantly reduced in the mutant (Figure 5, A–D), we thought that mtFAS affects the level of NAD(P)H/NAD(P)⁺ through photorespiration rather than through the pathway itself.

When photorespiration is impaired, photoinhibition occurs. Photoinhibition was originally thought to inhibit PSII; however, recent studies have suggested that photoinhibition affects not only PSII, but also PSI and cytochrome B6F (Adir et al., 2003; Li et al., 2018). Chlorophyll fluorescence imaging showed that F_v/F_m in fern-like mutant was significantly lower than that in AC under 400 μmol/(m²·s) light (Figure 4, M and N). Under 1% CO₂, fern-like can be partially recovered. Besides, metabolome analysis detected typic metabolites (such as glycine, glyoxylate, and dicarboxylate) involved in photorespiration are differentially accumulated in fern-like mutant. Photoinhibition causes photodamage to the mesophyll cell (Raven, 2011). In fact, the palisade tissue of fern-like leaves was severely damaged or even disappeared when plants were grown under higher light (Figure 4B). In addition, RNA-seq analysis revealed the significant changes of genes encoding antenna proteins in the photosystem (Figure 6A) and of genes associated with photosynthesis (Supplemental Figure S9), and the majority of them were up-regulated. Other than photosynthesis, the palisade tissue also plays an essential role in maintaining leaf morphology.

ROS plays an important role in shaping fern-like mutant

Our results showed that the fern-like mutant cannot be fully restored by growing in a high concentration of CO₂ environment, although photorespiration was suggested to play a leading role in the development of photorespiration-related mutants. Meanwhile, no similar phenotype as fern-like mutant was reported in other photorespiration-related mutants of tomato (Ye et al., 2020). Therefore, we speculate that ROS would also be directly or indirectly involved. Under normal growth conditions, fern-like accumulated a high level of ROS. From the results of ROS measurement, it was found that a higher level of superoxide anion accumulated in fern-like mutant than in WT, but H₂O₂ showed insignificant difference (Figure 5F and Supplemental Figure S17). A typical phenotype of fern-like mutant is that the shoot apex grows very slowly, but axillary buds appear more frequently and sooner, thus forming a bushy architecture. Recent studies have shown that ROS can determine the fate of stem cells, among which H₂O₂ and superoxide anions can regulate the proliferation and differentiation of plant stem cells in an antagonistic manner (Zeng et al., 2017; Huang et al., 2021). The changes of H₂O₂ and superoxide anions in fern-like are different from those in AC. They may affect the formation of apical and lateral buds in fern-like by regulating cell division and differentiation.

ROS functions both as a metabolite and a signaling molecule, which affects every aspect of plant development. Compared with AC, when treated with malic acid (a ROS inducer), the phenotypes of the mutant became more prominent (Zhao et al., 2018). This can be reflected by the activity of APX and GSH/GSSG ratio, whose amplitude of variation before and after treatment was obviously larger in fern-like than in AC (Figure 5, H–J). In addition, autophagy and cell death were detected in fern-like leaves (Figure 5, K and L), which may also be the consequence of excessive ROS. To our point of view, ROS plays an important role in shaping of the fern-like mutant.

The function of mtFAS beyond lipoic acid production

Curly leaf has not been documented in other photorespiration- or ROS-related mutants. However, a variety of Arabidopsis mutants related to lipid synthesis or metabolism display some sort of wrinkled or curled leaves, such as the mod1 (Mosaic death 1, encodes an enoyl-ACP reductase) mutant impaired with ptFAS (Mou et al., 2000). LCB1 (Long-chain base 1, encodes a subunit of serine palmitoyltransferase) is involved in sphingolipid synthesis and its knockdown lines also showed curly rosette and cauline leaves (Chen et al., 2006). The shine (shn) mutant and AtMYB41-overexpressing lines also display curled leaves and these genes were involved in the regulation of keratin and wax formation (Aharoni et al., 2004; Cominelli et al., 2008). In our study, a variety of sphingolipids, phospholipids, and TG were substantially changed in fern-like mutant leaf, which would contribute to the leaf phenotype in the mutant.

As mentioned before, the main products of mtFAS are C8, C10, C12, C14, and C16 saturated fatty acids. According to previous studies, C8:0 was considered to be the most important product of mtFAS, but octanoic acid was not detected in the lipidome of all the samples. It may be because the abundance of mid-chain fatty acids is low and most of them exist in binding state. For example, mtFAS needs to bind ACP in chain extension, so it could be difficult to detect free octanoic acid (Torella et al., 2013). We further analyzed the composition of FFAs in the mitochondrial lipidome of the mutants and found that the mtFAS products C10:0, C12:0, and C14:0 reduced substantially, but C16:0 increased slightly (Supplemental Figure S18). The ratio of C16:0 increased in chloroplast and leaf. Its accumulation is likely the consequence of membrane lipid destruction, as C16 is a basic component of various membrane lipids (Millar et al., 2000). The results of significantly increased MDA content in fern-like leaves (Supplemental Figure S1B) also support our speculation.

Previous reports indicate that mtFAS may affect mitochondrial function by altering the membrane. Studies in animals and yeast have shown that mtFAS can directly affect either the morphology or the structure of mitochondria (Kastaniotis et al., 2004). Although there is no substantially difference in the morphology of mitochondria between fern-like and AC, the stability of membrane may have changed in the mutant (Figure 4D). In Neurospora crassa, mtFAS is suggested to function in the repair of mitochondrial phospholipid. In the mtACP-deficient mutant of N. crassa, lyso-phosphatidyl acid (LPA) was increased as the intermediates of phospholipids following the oxidation of the active substance produced by the respiratory chain. The ACP-fatty acid produced by mtFAS may serve as a substrate for repairing mitochondrial phospholipids (Schneider et al., 1995, 1997). In our fern-like mutant, it was found that mitochondrial LPA increased by four times (Figure 6C).

The proper maintenance of the lipid composition in mitochondrial membrane is a prerequisite for its function. In the case of human congenital muscular dystrophy, the mutation in CHKB, a choline kinase beta gene associated with PC synthesis, results in the decrease of PC content and mitochondrial dysfunction (Mitsuhashi et al., 2011; Cole et al., 2012). It has also been suggested that mitochondria is likely associated with autophagy induced by starvation, because mitochondria can provide biofilm for autophagy and PE plays an important role in this process (Nebauer et al., 2007; Hailey et al., 2010). Many studies on humans and microorganisms have documented that mutations involving changes in mitochondrial membrane lipids can lead to serious developmental abnormalities, disease, and even death (Mejia and Hatch, 2016). In addition to being membrane components, lipids also function in many ways by binding to a diversity of active protein molecules. In short, the well maintenance of the mitochondrial membrane serves as the first guard for the normal operation of the plant energy factory.

MtFAS and ptFAS have very similar de novo FAS pathways, but fatty acids produced by mtFAS in plants only account for a small portion (Wada et al., 1997). The ptFAS produces C16 and C18, the most abundant fatty acid, which can be transported to other organelles for further processing by corresponding transporters such as trigalactosyldiacylglycerol 4 (TGD4) and fatty acid export 1 (FAX1) (Wada et al., 1997; Xu et al., 2008; Li et al., 2016). However, there are no reports on fatty acid transporter into or outside mitochondria, which appears to be a big obstacle in the study of mtFAS function. In addition, it is very challenging to exclude the influence of photorespiration in the functional study of mtFAS in plants. Therefore, there is still a long way to go to fully address the role of mtFAS in plants.

A proposed working model for FERN

Based on previous knowledge and our findings, we proposed a working model for FERN (Figure 7). FERN functions as a 3-hydroxylacyl-ACP dehydratase in mtFAS, which possesses four core enzymes responsible for carbon chain elongation. Studies in yeast, human beings (Hiltunen et al., 2010), and Arabidopsis (Guan et al., 2017) suggest that mutations in these enzymes affect mtFAS, leading to the reduction or even loss of lipoic acid synthesis in mitochondria. Consequently, GDC could not be acylated without lipoic acid during photorespiration, so its decarboxylase activity diminishes. As a result, glycine cannot be effectively converted to serine and NADH cannot be produced efficiently through decarboxylation in photorespiration, which inhibits the plant photorespiration cycle (Bauwe et al., 2010). When photorespiration is blocked in tomato plants, excessive reduction equivalent will be produced under the condition of either high ambient oxygen concentration or abundant illumination. Under this circumstance, the photorespiration of tomato plants is blocked and photoinhibition occurs, resulting in the overproduction of ROS in the photosynthetic electron transport chain (Pospisil, 2009). ROS can regulate cell division and differentiation (Zeng et al., 2017). Besides, NADH/NAD⁺ and ATP in fern-like mutant were significantly reduced, which would lead to the destruction of redox homeostasis, the impairment of energy supply, and even cell autophagy and death (Scherz-Shouval and Elazar, 2007; Zhao et al., 2018), thus causing irreversible damage to plants. Energy flow and ROS can also regulate tomato morphogenesis through a variety of regulatory pathways. Although C8:0 is considered as the major product of mtFAS, which is mainly used for lipoic acid synthesis, through lipidomic analysis, the overall changes of lipid profile in mitochondria and other tissue/organelle were observed. This should be partially attributed to the impairment of mtFAS.

Figure 7.

A proposed model for the roles of FERN. In mtFAS, malonate serves as the original substrate. It is converted to malonyl-CoA by malonyl-CoA synthase (mtMCS). Malonyl-CoA then reacts with holo-ACP, which is catalyzed by malonyl-CoA transacylase (mtMCAT), to produce malonyl-ACP. β-Ketoacyl-ACP synthase (mtKAS), β-ketoacyl-ACP reductase (mtKR), 3-hydroxyacyl-ACP dehydratase (FERN or mtHD), and enoyl-ACP reductase (mtER) are core enzymes for carbon chain elongation cycle. C8:0 is the most abundant product of mtFAS, which forms lipoic acid later. Lipoic acid serves as a cofactor of the GDC that functions in the conversion of glycine to serine in photorespiration. Photoinhibition caused by photorespiration affects the morphogenesis of fern-like mutant. Disruption of the redox homeostasis results in ROS accumulation and energy deficiency, leading to autophagy, cell death, and other developmental abnormalities. Besides C8:0 being used to produce lipoic acid, other fatty acid (FA) products of mtFAS could also affect plant development by participating in lipid metabolic pathways inside or outside mitochondria through yet-to-be elucidated ways.

Materials and methods

Plant materials and growth conditions

The fern-like mutant is a spontaneous mutant derived from the tissue culture of tomato variety AC (S.lycopersicum). A wild tomato accession LA1375 (S. pimpinellifolium) was used to construct the population for linkage analysis. Nicotiana benthamiana plants were used in the subcellular localization analysis of proteins. All materials, unless otherwise specified, were grown in growth chambers with day/night temperature of 25°C/20°C and 16-h/8-h photoperiod cycle.

Positional cloning of FERN

To map the gene underlying the fern-like mutant, a bi-parental population was constructed. The fern-like mutant was crossed with LA1375 to get F₁ and then selfed to obtain F₂ generation. For the first round of mapping, 450 F₂ plants were grown. Phenotypes were distinct in the young seedlings. Based on phenotypic results, 30 plants were selected in respective to construct the wild-type and fern-like bulks. The second true leaf from the top was taken from 30-day-old seedlings. Leaf samples were ground in liquid nitrogen and then an equal amount of tissue powder was mixed for RNA extraction using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to the supplier’s instruction. Total RNA samples were sent to Novogene Technology Co., Ltd. (Tianjin, China) for library construction and PE150 sequencing using Hiseq4000 (Illumina, USA). Bulked segregant RNA-Seq (BSR) mapping was performed to get the initial region according to a published method (Hill et al., 2013). InDel (Insertion/Deletion) and SNP (Single Nucleotide Polymorphism) sites with high reliability were identified using the resequencing data of AC and LA1375 from Solanaceae Genome Database (https://solgenomics.net) based on a published method (Li, 2011). The selected InDels and SNPs were mapped to the tomato genome (Version SL2.5) and the 150 bp upstream and downstream sequences flanking the InDel or SNP sites were derived for primer design and markers were developed. Polymorphic markers were used to narrow down the mapping region obtained by BSR analysis. Then, the population was increased to 1,000 F₂ plants and recessive plants were selected for the second round of mapping with markers developed in the mapped region. The final interval was determined based on recombination event analysis.

To identify the causal mutation, fern-like was re-sequenced and compared with the wild-type AC within the fine mapped region. Genomic DNA from fern-like seedlings was extracted by a cetyl trimethylammonium bromide (CTAB) method and sent to BGI (Shenzhen, China) for quality control, library construction, and sequencing. Resequencing was performed by Illumina/HiSeq2500, with a sequencing depth of 20×. Similar to marker development, a comparative analysis was carried out between the resequencing data of fern-like mutant and AC (derived from Solanaceae Genome Network database), based on published methods (Hillier et al., 2008; Nordstrom et al., 2013). Potential mutation sites were further verified using Sanger sequencing. The genomic DNA of fern-like and AC plants was amplified with primers FERN-coseg-F and FERN-coseg-R (Supplemental Table S3) and Phanta Max Super-Fidelity DNA Polymerase (Vazyme, Nanjing, P505-D1) before sequencing.

Other methods

Details of the methods for genetic complement assay, RT-qPCR, Western blotting, gene family analysis and construction of phylogenetic tree, subcellular localization, Paraffin section and microscopic observation, TEM analysis, CO₂ enrichment culture of tomato seedlings, light intensity treatment, transcriptome analysis, non-targeted metabolome analysis, staining of superoxide anions, malic acid treatment, trypan blue staining, preparation of mitochondrial and chloroplast samples for lipidomic analysis, and lipidomic analysis are available in Supplemental Materials and Methods.

Statistical analysis

Data of plant growth and physiological and biochemical parameters were analyzed using Microsoft Excel 2019 or Sigma Plot 14.0. The two-tailed Student’s t test was applied, with the significant levels at P < 0.05 and P < 0.01.

Accession numbers

Accession numbers for the genes mentioned in this work are as follows: SlFERN (Solyc06g075000), mtHD (AT5G60335), ECH2 (AT1G76150), and ACTIN (Solyc04g011500).

Supplemental data

Supplemental Figure S1. APX activity and MDA content.

Supplemental Figure S2. Validation of the SNP variation between AC and fern-like mutant by Sanger sequencing.

Supplemental Figure S3. Relative expression level of SlFERN in 15-day-old AC and fern-like plants.

Supplemental Figure S4. SlFERN genetically complements the fern-like mutant.

Supplemental Figure S5. A phylogenetic tree of SlFERN.

Supplemental Figure S6. The motif structure of MaoC-dehydratase family.

Supplemental Figure S7. Mitochondrial localization of the truncated FERN and mtHD.

Supplemental Figure S8. Volcano map of the DEGs between AC and fern-like mutant.

Supplemental Figure S9. The histogram of GO for DEGs between the wild-type AC and fern-like mutant.

Supplemental Figure S10. PCA of AC and fern-like metabolomes.

Supplemental Figure S11. Quality check of purified chloroplasts.

Supplemental Figure S12. Quality check of purified mitochondria.

Supplemental Figure S13. PCA of lipidome data.

Supplemental Figure S14. Venn diagram of the differential lipids from mature leaf, mitochondria, and chloroplast of AC and the fern-like mutant.

Supplemental Figure S15. KEGG enrichment for the differential lipids detected in mature leaf, and chloroplast of AC and fern-like mutant.

Supplemental Figure S16. The change of lipoic acid (LA) and hydrolipidic acid (DHLA) contents in the fern-like mutant grown in ambient air.

Supplemental Figure S17. H₂O₂ content.

Supplemental Figure S18. Changes of some free saturated fatty acids from mature leaf, mitochondria, and chloroplast of AC and the fern-like mutant.

Supplemental Table S1. Predicted genes in the mapped region of FERN.

Supplemental Table S2. Differentially expressed metabolic pathways between AC and fern-like mutant based on metabolome analysis.

Supplemental Table S3. Primer list.

Supplemental Materials and Methods .

 

B.O. and Y.Zhou designed the project and wrote the manuscript. Y.Zhou, H.Y., Y.T., Ro.C., J.L., C.S., S.T., X.L., and X.S. performed the wet lab and data analysis. Rongf.C. delivered the metabolome data. Y.Zhang helped perform the CO₂ enrichment experiment. Z.Y., L.G., and Y.L. advised on experiments related to mechanism study and revised 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/plphys/pages/general-instructions) is: Bo Ouyang (bouy@mail.hzau.edu.cn).