Discovery and Characterization of the 3-Hydroxyacyl-ACP Dehydratase Component of the Plant Mitochondrial Fatty Acid Synthase System^,

Identification and characterization of the mitochondrial 3-hydroxyacyl-ACP dehydratase reveal novel functionalities associated with the mitochondrial fatty acid synthase system.

Abstract

We report the characterization of the Arabidopsis (Arabidopsis thaliana) 3-hydroxyacyl-acyl carrier protein dehydratase (mtHD) component of the mitochondrial fatty acid synthase (mtFAS) system, encoded by AT5G60335. The mitochondrial localization and catalytic capability of mtHD were demonstrated with a green fluorescent protein transgenesis experiment and by in vivo complementation and in vitro enzymatic assays. RNA interference (RNAi) knockdown lines with reduced mtHD expression exhibit traits typically associated with mtFAS mutants, namely a miniaturized morphological appearance, reduced lipoylation of lipoylated proteins, and altered metabolomes consistent with the reduced catalytic activity of lipoylated enzymes. These alterations are reversed when mthd-rnai mutant plants are grown in a 1% CO₂ atmosphere, indicating the link between mtFAS and photorespiratory deficiency due to the reduced lipoylation of glycine decarboxylase. In vivo biochemical feeding experiments illustrate that sucrose and glycolate are the metabolic modulators that mediate the alterations in morphology and lipid accumulation. In addition, both mthd-rnai and mtkas mutants exhibit reduced accumulation of 3-hydroxytetradecanoic acid (i.e. a hallmark of lipid A-like molecules) and abnormal chloroplastic starch granules; these changes are not reversible by the 1% CO₂ atmosphere, demonstrating two novel mtFAS functions that are independent of photorespiration. Finally, RNA sequencing analysis revealed that mthd-rnai and mtkas mutants are nearly equivalent to each other in altering the transcriptome, and these analyses further identified genes whose expression is affected by a functional mtFAS system but independent of photorespiratory deficiency. These data demonstrate the nonredundant nature of the mtFAS system, which contributes unique lipid components needed to support plant cell structure and metabolism.

Plant cells utilize at least three different fatty acid-forming systems, which occur in multiple subcellular compartments: plastids, the membranes of the endoplasmic reticulum, and mitochondria (Ohlrogge and Jaworski, 1997; Wada et al., 1997; Samuels et al., 2008). Plastidial (ptFAS) and mitochondrial (mtFAS) fatty acid synthase systems form fatty acids de novo, whereas the endoplasmic reticulum-localized fatty acid elongase (FAE) system utilizes preexisting acyl-CoA precursors to synthesize very-long-chain fatty acids of 20 carbons and longer (James et al., 1995). The ptFAS system generates the bulk of a plant cell’s fatty acids from acetyl-CoA, and these fatty acids serve as precursors for the assembly of acyl lipids that constitute membrane lipids (e.g. phospholipids and glycoglycerolipids), storage lipids (e.g. triacylglycerol [TAG]), and signaling lipids (e.g. sphingolipids, phosphatidylinositols, and oxylipins; Benning, 2009; Li-Beisson et al., 2013). The very-long-chain fatty acids generated by the FAE system can be incorporated into a variety of lipids, including surface cuticular lipids (Samuels et al., 2008), the ceramide moiety of sphingolipids (Markham et al., 2013), and in discrete quantities in some glycerolipids (Millar et al., 1998).

The mtFAS system appears to use free malonic acid as the substrate (Guan and Nikolau, 2016) to primarily generate octanoyl-acyl carrier protein (ACP), which is the required precursor for the biosynthesis of lipoic acid (Yasuno and Wada, 1998; Gueguen et al., 2000; Wada et al., 2001). Lipoic acid is the cofactor that is essential for pyruvate dehydrogenase (PDH), α-ketoglutarate dehydrogenase (KGDH), branched-chain α-ketoacid dehydrogenase, and the Gly decarboxylase complex (GDC; Taylor et al., 2004). To date, alternative functions for the mtFAS system have not been demonstrated, although its role in detoxifying mitochondrial malonic acid (Guan and Nikolau, 2016) and in remodeling cardiolipins (Frentzen and Griebau, 1994; Griebau and Frentzen, 1994) have been suggested.

Based on the characterization of three Arabidopsis (Arabidopsis thaliana) mtFAS enzymatic components, mitochondrial β-ketoacyl-ACP synthase (mtKAS; Olsen et al., 2004; Yasuno et al., 2004; Ewald et al., 2007), phosphopantetheinyl transferase (mtPPT; Guan et al., 2015), and malonyl-CoA synthetase (mtMCS; Guan and Nikolau, 2016), it appears that the plant mtFAS system resembles the type II FAS system that occurs in bacteria and plant plastids (Ohlrogge and Jaworski, 1997; White et al., 2005). Type II FAS systems recruit ACP as the carrier of the intermediates of the process and utilize dissociated, monofunctional enzymes to catalyze the iterative reactions that produce fatty acids (Hiltunen et al., 2010). This contrasts with the type I FAS that occurs in the cytosol of fungi, mammals, and some bacteria, where a multifunctional protein that contains all of the catalytic centers required for fatty acid biosynthesis iteratively catalyzes the formation of fatty acids from acetyl-CoA and malonyl-CoA (Smith et al., 2003).

Here, we report the identification and characterization of a gene encoding the mitochondrial 3-hydroxyacyl-ACP dehydratase (mtHD) that catalyzes the third of the four iterative reactions that constitute the mtFAS cycle. Systematic investigations (i.e. biochemical, morphological, metabolomic, and transcriptomic analyses) confirm its role in mtFAS and the important role this process has in supporting photorespiration. Moreover, these characterizations identify additional novel mtFAS functions in supporting the assembly of lipid A-like molecules and an unexpected function in maintaining chloroplastic starch granule morphology.

RESULTS

Biochemical Identification of AT5G60335 as the Arabidopsis mtHD Component

Sequence-based identification of a candidate plant mtHD gene is somewhat complex, because the two well-characterized mtHD homologs from yeast (Kastaniotis et al., 2004) and humans (Autio et al., 2008b) share heterogenous and low sequence similarity with many genes from Plantae. For example, BLAST analysis with the yeast HTD2 sequence failed to identify any significant Arabidopsis homolog (e > 3.9), whereas parallel analysis with the human HsHTD2 identified a single mtHD candidate AT5G60335 (e = 3e-17; Fig. 1).

Figure 1.

Comparison of the amino acid sequences of the Arabidopsis mtHD (encoded by AT5G60335) and the human HsHTD2 proteins. Residues shaded in black are identical, and those shaded in gray share similarity in the chemistry of the side chains. Alignment was performed using ClustalΩ (AlignX, Vector NTI 10), with a gap opening penalty of 10 and a gap extension penalty of 0.1.

The AT5G60335 protein coding sequence (CDS) was cloned with a reverse transcription (RT)-PCR strategy using an RNA template isolated from aerial organs of young Arabidopsis seedlings. This CDS encodes a protein of 166 amino acids that contains an N-terminal 25-residue segment that is not homologous with human HsHTD. This N-terminal sequence is rich in basic amino acids and lacks acidic residues (Fig. 1), these being characteristics typical of mitochondrial targeting presequence elements. The AT5G60335-encoded protein is predicted to be mitochondrially localized by MitoProt II (24 N-terminal residues with a score of 0.9957; Claros and Vincens, 1996), PSORT (with a score of 0.751; Nakai and Horton, 1999), and Target P (17 N-terminal residues with a score of 0.68; Nielsen et al., 1997; Emanuelsson et al., 2000).

The AT5G60335 CDS was expressed in the yeast htd2 mutant strain that lacks a functional mtHD enzyme. This strain cannot grow on glycerol as the sole carbon source due to a respiratory deficiency (Kastaniotis et al., 2004). In this experiment, the putative mitochondrial targeting presequence (24 residues) of the AT5G60335 protein was replaced by the mitochondrial presequence of the yeast COQ3 protein (Hsu et al., 1996) to ensure the correct mitochondrial localization of the AT5G60335 protein in yeast.

As illustrated in Figure 2, neither the empty plasmid control nor the COQ3 mitochondrial targeting element are capable of rescuing the growth deficiency of the htd2 mutant strain on glycerol, but both the native yeast HTD2 gene and the AT5G60335 CDS restored the growth of the yeast strain on glycerol medium. Therefore, this experiment establishes that the AT5G60335 gene codes for a function that overcomes the deficiency in 3-hydroxyacyl-ACP dehydratase activity that is capable of generating acyl-ACP that is used to synthesize lipoic acid in yeast.

Figure 2.

Genetic complementation of the yeast htd2 mutant by the Arabidopsis mtHD gene (AT5G60335). Expression of mtHD was controlled with the phosphoglycerate kinase promoter (pPGK) and terminator (tPGK). The dilution of the inocula for each strain is indicated. A, Yeast htd2 mutant strains were grown on glycerol as the sole carbon source. The yeast strain in each row carried the indicated constructs: the control empty plasmid (row 1); the construct to overexpress the HTD2 protein (row 2); the construct to overexpress the mitochondrial presequence of the yeast COQ3 protein (MP; row 3); and the construct to overexpress the mitochondria-targeted mtHD (MP-mtHD fusion protein; row 4). B, Same strains as in A, but they were grown on Glc as the sole carbon source.

In Vitro Characterization of the Enzymatic Activity of the AT5G60335 Coding Protein

The Escherichia coli-produced recombinant AT5G60335 coding protein was assayed for the ability to catalyze 3-hydroxyacyl-ACP dehydratase activity. As with 3-hyroxyacyl-ACP dehydratases characterized from a variety of different type II FAS systems (Kastaniotis et al., 2004; Autio et al., 2008a, 2008b), this reaction was assayed in the reverse direction, namely the hydration of enoyl-CoA substrates. In addition, this assay was used to determine the substrate specificity of the enzyme in relation to the acyl chain length of the substrate, and these data were compared with the specificity of the mtKAS enzyme.

Two enoyl-CoA thioesters (i.e. trans-Δ²-10:1-CoA and trans-Δ²-16:1-CoA) were chemically synthesized as substrates for these assays (Supplemental Fig. S1A), and eight different acyl-ACP thioesters (with 4:0, 6:0, 8:0, 10:0, 12:0, 14:0, 16:0, and cis-Δ⁹-16:1 acyl moieties) were synthesized as substrates for the mtKAS assays (Supplemental Fig. S1B). Both enzymes exhibited classical hyperbolic Michaelis-Menten activity responses to increasing concentrations of each tested substrate, and K_m and V_max values were calculated for each substrate-enzyme combination (Fig. 3). The AT5G60335 coding protein has the ability to catalyze the expected hydration of the enoyl-CoA thioesters, and the catalytic efficiency (k_cat/K_m) for the medium-chain substrate (trans-Δ²-10:1-CoA) is similar to that of the long-chain substrate (trans-Δ²-16:1-CoA; Fig. 3A). Similarly, mtKAS is able to use saturated acyl-ACP thioesters of between 4- and 16-carbon acyl chains as substrates, but its activity with the unsaturated substrate (cis-Δ⁹-16:1-ACP) was barely detectable (Fig. 3B). The catalytic efficiency of mtKAS with different saturated acyl-ACP substrates (evaluated by k_cat/K_m) is ranked in the following order: 16:0 > 14:0 > 6:0 > 8:0 > 10:0 > 12:0 > 4:0. This ranking is affected by differences in both K_m and V_max with each substrate. These characterizations indicate that plant mtFAS enzymes have the ability to synthesize saturated fatty acids of up to 18-carbon chain length.

Figure 3.

Substrate specificity of the recombinant mtHD and mtKAS enzymes. A, Substrate concentration dependence of the hydratase reaction catalyzed by mtHD with enoyl-CoA substrates of 10- and 16-carbon atom acyl chain lengths. Michaelis-Menten kinetic parameters tabulated below the graph were calculated from three replicates for substrate concentrations of 200, 100, and 70 µm, four replicates for substrate concentrations of 50 and 30 µm, six replicates for substrate concentration of 20 µm, and 10 replicates for substrate concentration of 10 µm. B, Substrate concentration dependence of the condensation reaction catalyzed by mtKAS with acyl-ACP substrates of between 4- and 16-carbon atom acyl chain lengths. Michaelis-Menten kinetic parameters tabulated below the graph were calculated from three replicates for substrate concentrations of 100 and 50 µm, four replicates for substrate concentrations of 20 and 10 µm, and 10 replicates for substrate concentrations of 5 and 2 µm.

Mitochondrial Localization of the AT5G60335 Gene Product

The subcellular localization of the AT5G60335 protein was experimentally determined in Arabidopsis with a GFP-tagged transgenic fusion protein. In this experiment, the DNA fragment encoding the N-terminal 40 residues of the AT5G60335 protein was tested for its ability to target GFP to a specific organelle. Figure 4 shows confocal micrographs of roots and leaf mesophyll cells of the resulting GFP transgenic plants. In roots, MitoTracker Orange was applied as the mitochondrial marker and it was recorded simultaneously with GFP fluorescence, while in mesophyll cells, chlorophyll autofluorescence was used as a chloroplastic marker. In plants carrying the p35S::AT5G60335-GFP transgene, GFP signals were obtained from both roots and leaf mesophyll cells in a distinct pattern that indicates localization in organelles. This GFP feature overlaps with the MitoTracker Orange in roots but is distinct from the leaf chlorophyll autofluorescence, revealing that the AT5G60335-GFP fusion protein is mitochondrially localized. In control experiments, no GFP fluorescence was detectable in roots and mesophyll cells of the wild-type plants; in transgenic plants carrying a nontargeted GFP construct (i.e. p35S::GFP), the GFP fluorescence localizes to the cytosol and nucleus, which is consistent with previous reports (Guan et al., 2015).

Figure 4.

Subcellular localization of mtHD (encoded by AT5G60335) determined with GFP-tagged transgenes. A, Fluorescence micrographs of roots of nontransgenic wild-type control plants (WT), transgenic plants carrying the p35S::mtHD_1-120-GFP transgene, and transgenic plants carrying the p35S::GFP control transgene. Confocal fluorescence micrographs imaged the emission of GFP, MitoTracker Orange, or the merged images of GFP and MitoTracker Orange. B, Fluorescence micrographs of leaf mesophyll cells of nontransgenic wild-type plants, transgenic plants carrying the p35S::mtHD_1-120-GFP transgene, and transgenic plants carrying the p35S::GFP control transgene. Confocal fluorescence micrographs imaged the emission of GFP, chlorophyll autofluorescence, or the merged images of GFP and chlorophyll autofluorescence.

In combination, therefore, these sets of experiments presented in Figures 1 to 4 lead us to the conclusions that AT5G60335 encodes the mtHD enzyme and that this organelle-targeting process is guided by the N-terminal 40-residue leader sequence. Furthermore, the enzyme kinetic data reveal that, in Arabidopsis, mtHD and mtKAS do not restrain the mtFAS system from synthesizing long-chain saturated fatty acids of up to 18 carbons in length. This contrasts with other characterized eukaryotic mtFAS systems, where the substrate specificity of either mtHD (Autio et al., 2008a) or mtKAS (Zhang et al., 2005) constrains the system to produce fatty acids of 12 carbon atoms and shorter.

Expression Patterns of mtHD and mtKAS in Different Arabidopsis Organs

The spatial and temporal expression pattern of the mtHD gene was determined by quantitative RT-PCR, and these data were compared directly with those for the other well-characterized mtFAS gene, mtKAS (AT2G04540). Quantitative RT-PCR analysis of RNA preparations extracted from different organs shows that the two mtFAS genes exhibit parallel expression patterns. The expression of both mtHD and mtKAS genes occurs in all organs tested (Fig. 5), with an approximately 3-fold difference between the highest (i.e. flowers) and lowest (i.e. siliques) levels of expression. These nearly ubiquitous expression patterns of the mtHD and mtKAS genes are consistent with the microarray data visualized by the Arabidopsis eFP Browser (Schmid et al., 2005; Winter et al., 2007) and also are comparable with the expression patterns of other already characterized mtFAS components, such as the mtPPT (Guan et al., 2015) and mtMCS (Guan and Nikolau, 2016) genes.

Figure 5.

Expression of mtHD (white columns) and mtKAS (gray columns) genes in different organs of Arabidopsis. Quantitative RT-PCR analysis was conducted on RNA templates isolated from different organs of plants at the indicated days after imbibition (DAI). Results are means of three biological replicates ± se and are presented as normalized values relative to the expression of the ACTIN2 gene.

Morphological Alterations Associated with mthd-rnai and mtkas Mutations

The physiological significance of the mtHD gene was investigated by studying segregants identified among a family of plants generated from an Arabidopsis genetic stock that carries a T-DNA-tagged mutant allele (mthd-1) in the heterozygous state (CS856112, which is the only T-DNA insertion line available; http://signal.salk.edu). These characterizations demonstrated that the T-DNA-tagged mutant is not a null allele and that there was no obvious phenotypic difference between wild-type plants and those that were homozygous for the mthd-1 allele (Supplemental Fig. S2).

As an alternative, therefore, we generated RNA interference (RNAi) transgenic plants, which express a suppressor of the mtHD gene, under the control of the 35S promoter. Seventy-eight independent RNAi lines were recovered, and quantitative RT-PCR analysis determined the levels of the remaining mtHD transcript in the aerial organs of young seedlings at 16 DAI. Two RNAi lines that exhibit the lowest mtHD expression levels (3% and 4% of the wild-type levels) were designated as mthd-rnai-1 and mthd-rnai-2 strains, respectively, and these were used in subsequent analyses. These two mutant strains exhibit aerial organs that are significantly reduced in size compared with the wild-type plants, and their developmental appearance was classified according to the systematic system developed by Boyes et al. (2001). Thus, at 16 DAI, these mutant plants develop to stage 1.06 (i.e. exhibiting six rosette leaves that are longer than 1 mm in length), whereas the wild-type plants develop to stage 1.07 (i.e. exhibiting seven rosette leaves that are longer than 1 mm in length) within the same time period (Fig. 6A).

Figure 6.

Morphological phenotypes of the mthd-rnai and mtkas mutants. Single mutants (A) and double mutants (B) were grown in ambient air or in the 1% CO₂ atmosphere. WT, Wild type.

When these mthd-rnai mutant plants were grown in an atmosphere containing 1% CO₂, they grew nearly normally (Fig. 6A). The phenotypic reversal in the elevated CO₂ atmosphere is typical of photorespiratory deficiency (Somerville and Ogren, 1979), with the elevated CO₂ levels inhibiting the oxygenation reaction of Rubisco and, thus, reducing the levels of 2-phosphoglycolate, the starter metabolite of photorespiration. Furthermore, this phenotype also is conferred by mutations in other mtFAS components, such as mtkas (Ewald et al., 2007), mtppt-rnai (Guan et al., 2015), and aae13 (Guan and Nikolau, 2016).

To test any additive effect of mutations in the mtHD- and mtKAS-catalyzed reactions, we generated double mutant stocks by transforming the mtkas-2 mutant line with the p35S::mtHD-RNAi transgene. We recovered 41 independent transformants and selected two lines that display the lowest mtHD expression levels (8% and 11% of the wild-type level, respectively) for additional analysis. In ambient air, the double mutant lines exhibit a stronger growth defect compared with the mtkas-2 parental plants. In our experimental conditions, at 16 DAI, the growth of these double mutants is arrested at stage 1.02 (i.e. exhibiting two rosette leaves), whereas the mtkas-2 mutant plants develop to stage 1.05 (i.e. exhibiting five rosette leaves; Fig. 6B). In addition, these double mutant plants do not progress any further in development and eventually die, which contrasts with the situation with the mtkas-2 mutant, which progresses to maturity, although at a slower rate than the wild-type plants, and ultimately sets seeds. When grown in the 1% CO₂ atmosphere, the growth of these double mutants is reversed to a nearly wild-type appearance, and these plants develop to stage 1.06 by 16 DAI (Fig. 6B).

Chloroplast Alterations Associated with mthd-rnai and mtkas Mutations

More detailed insights into the growth phenotype of mthd-rnai and mtkas mutants were obtained by examining leaves of these mutants by light microscopy and transmission electron microscopy. These observations illustrate the distinct differences in leaf cell morphology and ultrastructure between the wild-type plants and the mutants (Fig. 7A). Specifically, when these plants are grown in ambient air, mesophyll cells are enlarged in the mutants, making the leaves thicker than those of wild-type plants. Despite the fact that these mutants lack a mitochondrial biochemical function, the ultrastructure of this organelle is unaffected. The most dramatic ultrastructural alterations in the mutants are associated with chloroplasts. Thylakoid membrane assembly is affected by the mutations, with the thylakoid structures being sparser and less extensive. Furthermore, the starch granule ultrastructure appears to be drastically altered in the mutant lines, with many small granules present; this compares with the typical two to five disc-shaped structures that occur in the wild-type plants.

Figure 7.

Leaf ultrastructural phenotypes of the mthd-rnai and mtkas mutants. Light micrographs of leaf cross sections (column 1) and transmission electron micrographs of chloroplasts (column 2) and mitochondria (column 3) of leaf mesophyll cells are shown for the indicated genotypes, which were grown in ambient air (A) or the 1% CO₂ atmosphere (B). WT, Wild type.

Most of these alterations in the leaf cell morphology and ultrastructure of the mutants (i.e. cell size and thylakoid membrane) are reversed when the mutant plants are grown in the 1% CO₂ atmosphere (Fig. 7B). The exception to this morphological reversal is the ultrastructure of the starch granules, which maintain the abnormal morphology observed when the plants were grown in ambient air.

Altered Protein Lipoylation Status Associated with mthd-rnai and mtkas Mutations

Mitochondrial lipoic acid biosynthesis is primed with octanoyl-ACP generated by the mtFAS system (Ewald et al., 2007). Therefore, we examined the protein lipoylation status of different enzymes in the mtFAS mutants (i.e. mthd-rnai, mtkas-2, and mtkas-2-mthd-rnai mutants). Using a western-blot procedure with anti-lipoic acid antibodies, we examined the lipoylation status of the H subunit of GDC, E2 subunits of mitochondrial PDH and KGDH, and plastidial PDH (Ewald et al., 2007; Guan et al., 2015; Fig. 8). Plants were grown in the 1% CO₂ atmosphere to eliminate any bias due to the morphological appearance associated with the mutant alleles. SDS-PAGE analysis of protein extracts does not indicate any dramatic differences in the expressed proteomes of these mutant plants. The most dramatic alteration in the lipoylation status of these proteins was associated with the H subunit of GDC, which is depleted to about 10% of wild-type levels in the mthd-rnai mutants and is undetectable in the mtkas-2 mutant and mtkas-2-mthd-rnai double mutant strains. The dramatic depletion in the lipoylation status of this protein does not affect the accumulation of the H protein itself, as indicated by measurements made using anti-H protein antibodies. Therefore, these plants harbor a large pool of inactive apo-H subunit of GDC.

Figure 8.

Protein lipoylation status in the aerial organs of mthd-rnai and mtkas mutant plants. A, Coomassie Brilliant Blue-stained SDS-PAGE analysis of extracts prepared from the indicated genotypes. B, Western-blot analysis of the H subunit of GDC detected with anti H-protein antibodies and lipoylation status of the H-protein and other indicated lipoylated proteins detected with anti-lipoic acid antibodies. WT, Wild type.

These immuno-based lipoic acid analyses also identified that the lipoylation status of other mitochondrial lipoylated proteins also is reduced in these mutant plants; these are decreased to as low as 30% of the wild-type level (i.e. E2 subunit of KGDH in the mtkas-2-mthd-rnai double mutants). In all of these mutants, the lipoylation status of the E2 subunit of plastidial PDH was unaffected.

Metabolomic Alterations Associated with mthd-rnai and mtkas Mutations

Using multiple analytical platforms, the metabolomes of the mthd-rnai and mtkas mutants were analyzed and compared with those of wild-type plants. These analyses quantified 143 metabolites in aerial organs, and these metabolites were categorized as aqueous metabolites, fatty acids, lipids, surface cuticular lipids, and starch granule components. The metabolomes of these mutants and wild-type plants were determined when they were grown in either ambient air, when the growth phenotype associated with the photorespiratory deficiency is expressed, or in the 1% CO₂ atmosphere, when the photorespiratory growth phenotype is suppressed (Fig. 9; Supplemental Fig. S3).

Figure 9.

Metabolomic alterations in the mthd-rnai and mtkas mutants. The mthd-rnai-1 mutant (A) and the mtkas-2 mutant (B) were grown either in ambient air (black circles) or in the 1% CO₂ atmosphere (white circles). The y axis represents the individual metabolites that were identified. The x axis plots the log₂-transformed relative ratio of the abundance of each metabolite in each mutant sample, normalized to the same metabolite in the wild-type (WT) control sample. Results are presented as means ± se. MGDG, Monogalactosyldiacylglycerol; DGDG, digalactosyldiacylglycerol; SQDG, sulfoquinovosyl diacylglycerol; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PI, phosphatidylinositol; DAG, diacylglycerol.

In ambient air, the most dramatic metabolic alteration in aqueous metabolites is a 70- to 150-fold hyperaccumulation of Gly. Several additional aqueous metabolites exhibit significant, but smaller, increases in accumulation, between 25% and 4-fold of the wild-type levels. In contrast, Suc is depleted to between 5% and 12% of the level present in the wild-type plants.

All of these mutant plants exhibit significantly reduced levels of saponifiable fatty acids (e.g. 16:1, 16:2, and 16:3) that constitute membrane lipids. Consistent with the reduction in the accumulation of these fatty acids, the levels of the major leaf glycoglycerolipids (i.e. monogalactosyldiacylglycerol, digalactosyldiacylglycerol, and sulfoquinovosyldiacylglycerol) are reduced markedly to below 80% of the wild-type levels. Chlorophyll content also is reduced to below 80% of the wild-type level, as would be expected from the yellowish appearance of the mutant plants. In addition, the accumulation of most surface cuticular lipids is reduced significantly in these mutants. In contrast, however, TAG hyperaccumulates to above 5-fold of the wild-type level.

Alterations in the accumulation of phospholipids (i.e. phosphatidylcholine, phosphatidylethanolamine, phosphatidylglycerol, and phosphatidylinositol) and diacylglycerol are barely detectable. In addition, the quantities of insoluble starch and water-soluble Glc polysaccharide (WSP) were compared between the wild-type and mutant plants. Despite the difference in starch granule morphology observed by transmission electron microscopy, the total quantity of starch and WSP was essentially unchanged in the mutants.

These alterations in the metabolome appear to be a consequence of the deficiency in photorespiration, because most of these are reversed when the mutant plants are grown in a condition that inhibits photorespiration (i.e. the 1% CO₂ atmosphere). An exception is Gly, which, although it is reduced significantly when these mutant plants are grown in the 1% CO₂ atmosphere, is still 15- to 50-fold higher than that in the wild-type plants.

We also compared the aqueous metabolite profiles between the mtkas-2-mthd-rnai double mutant and its parental mtkas-2 mutant plants. Whereas malonate levels are unaffected in the mtkas-2 and mthd-rnai single mutants (Fig. 9), it is elevated by 3-fold in the double mutant (Supplemental Fig. S4), even when these plants are grown in the 1% CO₂ atmosphere, which suppresses photorespiration. This elevation in malonate levels, and the associated growth penalty, resembles the situation in the aae13-1 mutant (Guan and Nikolau, 2016), which cannot activate malonate to malonyl-CoA. Therefore, the reduced size of the mtkas-2-mthd-rnai double mutants may be associated with malonate toxicity (Guan and Nikolau, 2016), a known inhibitor of succinate dehydrogenase activity (Quastel and Wooldridge, 1928; Greene and Greenamyre, 1995).

Biochemical Mimics of the Mutant Effect on the Morphology and Metabolome

Based on the observation that growing the mthd-rnai and mtkas mutant plants in an elevated CO₂ atmosphere suppresses the changes in morphology and metabolome, we considered that some of these genetic deficiency effects are secondary to the deficiency in photorespiration. We hypothesized, for example, that the alterations in morphology and lipids are a result of the alterations in the steady-state levels of soluble metabolites that are intermediates of photorespiration. This postulate was tested by exogenously feeding these photorespiration intermediates to wild-type and mtkas mutant plants and testing whether these biochemicals complement or mimic the mtkas mutation. Specifically, we targeted the effects of treating plants with Suc, glycolate, and Gly, the metabolites associated with photorespiration, whose accumulation is either down- or up-regulated by the mutations.

By growing wild-type and mtkas-2 mutant plants on medium supplemented with 3% Suc, we tested whether the mtkas-induced reduction in Suc accumulation was the cause of the alterations in morphology and lipids. First, as indicated by directly assaying the Suc content of the tissue, we showed that these plants take up the exogenously supplied Suc (Supplemental Fig. S5C), and the Suc depletion that occurs in the mtkas mutant plants is alleviated significantly, reaching about 30% of the level obtained in the Suc-treated wild-type plants. The Suc supplementation markedly relieves the dwarf phenotype and the yellowish leaf color of the mtkas mutant plants (Supplemental Fig. S5A). In addition, externally added Suc reverses the mtkas-induced alteration in the ultrastructure of the thylakoid membrane system (Supplemental Fig. S5B). Moreover, the Suc treatment effectively complements the alterations in the profiles of fatty acids, glycoglycerolipids, and chlorophylls (Supplemental Fig. S5C). In contrast, exogenous Suc does not reverse the enlarged mesophyll cells, abnormal starch grains, and hyperaccumulation of TAG in the mtkas-2 mutant. These results demonstrate that Suc depletion in the mtkas-2 mutant contributes to the morphological phenotype (i.e. stunted growth, defective thylakoid membrane system, and loss of chlorophylls) and the alterations in glycoglycerolipids.

In corollary experiments, we evaluated the effects of other exogenous biochemicals (specifically 10 mm glycolate and Gly) on the wild-type plants to test whether the mtkas-induced hyperaccumulation of these photorespiratory intermediates is the cause of the morphological phenotype and of the altered lipidome in the mutants. These exogenously provided biochemicals appear to have been taken up by the plants, as evidenced by the 4-fold increase in glycolate tissue levels in the glycolate-treated plants (Supplemental Fig. S5C); these elevated levels are similar to those measured in the mthd-rnai and mtkas mutant plants. The application of exogenous glycolate leads to a dramatic dwarf appearance with dark-green leaf color (Supplemental Fig. S5A). In addition, this treatment increases the number of large starch granules in the chloroplasts (Supplemental Fig. S5B). In parallel, these plants also exhibit about a 6-fold increase in TAG accumulation (Supplemental Fig. S5C). These metabolic changes resemble those observed in the mthd-rnai and mtkas mutant plants, which also hyperaccumulate glycolate. In contrast, glycolate does not affect the accumulation of other portions of the tested plant lipidome in a pattern that resembles the mthd-rnai and mtkas mutants.

The contribution of Gly to the alterations in morphology and lipidome is somewhat more difficult to determine, as the exogenous application of Gly to the wild-type plants only leads to an increase of the endogenous Gly levels by about 5-fold, which is considerably lower than the over 70-fold hyperaccumulation that occurs in the mthd-rnai and mtkas mutants. Moreover, when the mtkas mutant phenotype is suppressed by growth in the 1% CO₂ atmosphere, endogenous Gly levels are still 15-fold higher than those that occur in the wild-type plants. It appears, therefore, that Gly hyperaccumulation is only a small contributor to those alterations in morphology and lipidome.

Depleted 3-Hydroxytetradecanoic Acid Associated with mthd-rnai and mtkas Mutations

In addition to the metabolic alterations discussed above, we also identified two novel fatty acids in Arabidopsis, 3-hydroxytetradecanoic acid and 3-hydroxyhexadecanoic acid (Supplemental Fig. S6A). These are components that have long been known to be associated with gram-negative bacterial lipid A (Raetz et al., 2007) and recently were described with Arabidopsis lipid A-like molecules (Li et al., 2011). While the accumulation of 3-hydroxyhexadecanoic acid is unaffected in the mthd-rnai and mtkas mutant plants, the accumulation of 3-hydroxytetradecanoic acid is depleted to between 20% and 30% of the wild-type level, irrespective of the atmospheric conditions that affect photorespiration (Fig. 9; Supplemental Fig. S3).

To clarify the metabolic origin of 3-hydroxytetradecanoic acid, we assayed its accumulation in two mutant strains that are deficient either in photorespiration (i.e. the shmt1-2 mutant, which is deficient in the photorespiratory Gly-to-Ser conversion; Voll et al., 2006) or in the biosynthesis of lipid A-like molecules (i.e. the atlpxa-1 mutant, which is deficient in the first reaction of the assembly of lipid A-like molecules; Li et al., 2011; Supplemental Fig. S6B). 3-Hydroxyhexadecanoic acid content is unaffected in the shmt1-2 mutant, demonstrating that its depletion is independent of the photorespiratory deficiency. In the atlpxa-1 mutant, however, the accumulation of 3-hydroxytetradecanoic acid is depleted to 15% of the wild-type level, indicating that lipid A-like molecules are the major metabolic sink of this hydroxylated fatty acid. The latter conclusion is further supported by the substrate specificity of acyl transferases (i.e. AtLpxA, AtLpxD1, and AtLpxD2) that are involved in the assembly of lipid A-like molecules; all of these enzymes recognize 3-hydroxytetradecanoyl-ACP as their optimal substrate (Joo et al., 2012; Supplemental Fig. S6C).

Parallel Alterations in Gene Expression Induced by the mthd-rnai-1 and mtkas-2 Mutations

RNA sequencing (RNA-seq) experiments were performed to assess alterations in the transcriptomes associated with two independent genetic blocks in the mtFAS system, namely mthd-rnai-1 and mtkas-2 mutant strains. In these experiments, we sequenced the transcriptomes of the wild type and the two mutant strains that were grown in either ambient air or the 1% CO₂ atmosphere. Initially, we directly measured changes in gene expression by conducting quantitative RT-PCR on nine target genes and tested the validity of using the count numbers obtained from the identical RNA sample preparations by RNA-seq analysis. Correlation coefficients of the data obtained by the two methods are above 0.9 (Supplemental Fig. S7), indicating that RNA-seq read counts can be used as a quantitative readout of global changes in gene expression in response to the two mutations.

Statistical analyses investigated whether mthd-rnai-1 and mtkas-2 elicit similar changes in the transcriptome. These analyses identified differentially expressed genes (DEGs) among the three genotypes (i.e. wild type, mthd-rnai-1, and mtkas-2) when these plants were grown in the two environmental conditions (i.e. ambient air or the 1% CO₂ atmosphere); these are comparisons 1 to 6 of Figure 10A (Supplemental Data S1). Gene Ontology (GO) functional categories associated with these DEGs were analyzed at TAIR (www.arabidopsis.org; Supplemental Data S2).

Figure 10.

Effects of the mthd-rnai and mtkas mutations on the global gene expression profile. A, The transcriptomes of the indicated genotypes, grown either in ambient air or in a 1% CO₂ atmosphere, were analyzed by RNA-seq. The six comparisons of the transcriptome data of each genotype (arrows) identified different numbers of DEGs. B, The relative abundances of individual transcripts of DEGs are compared between mutant and wild-type (WT) plants, grown either in ambient air or in the 1% CO₂ atmosphere. The log₂-transformed fold change between mthd-rnai-1 mutant and wild-type plants is plotted against the log₂-transformed fold change between mtkas-2 mutant and wild-type plants.

Comparisons 1 (7,626 DEGs; 37% of the expressed genome) and 2 (9,052 DEGs; 44% of the expressed genome) indicate that each mutation significantly alters the expression of a large proportion of the genome when plants are grown in ambient air. A χ² test for the association between the two DEG lists indicates a significant (P < 0.0001) overlap between genes affected by the mthd-rnai-1 and mtkas-2 mutations. Furthermore, Figure 10B shows the plot of fold change in expression in comparison 1 versus comparison 2 and indicates that the directions and magnitudes of the estimated effects of each mutation were strongly positively correlated (r = 0.943, P < 0.0001). However, comparison 3 demonstrates that there are a few genes (114 DEGs; less than 1% of the expressed genome) whose expression is differentially affected between the two mutants.

While the distinctions in gene expression profiles between the wild type and mutants were greatly reduced when these plants were grown in the 1% CO₂ atmosphere, similar relationships between these three genotypes are apparent. Specifically, comparisons 4 (286 DEGs; 1% of the expressed genome) and 5 (1,825 DEGs; 9% of the expressed genome) show strong correlated changes in the transcriptome relative to the differences between the two mutants, as reflected by comparison 6 (63 DEGs; less than 1% of the expressed genome). Again, a χ² test for association between the DEG lists for comparisons 4 and 5 indicates a significant overlap between the genes affected by the two mutations, and Figure 10B shows a positive correlation (r = 0.864, P < 0.0001) between the effects of each mutation. Collectively, these results demonstrate that the mthd-rnai-1 and mtkas-2 mutations induce parallel changes in the transcriptome, which supports the hypothesis that these two genes operate in the same metabolic pathway, namely mtFAS.

Transcriptomic Alterations Associated with mtFAS Deficiency But Independent of Photorespiratory Deficiency

Further statistical analyses addressed whether any photorespiration-independent but mtFAS-dependent processes are revealed by RNA-seq analysis of the mthd-rnai-1 and mtkas-2 mutants. For these analyses, we initially identified DEGs whose expression is associated with three modifiers of gene expression: (1) different genotypes (G effect; i.e. wild type, mthd-rnai-1, and mtkas-2); (2) different environmental conditions (E effect; ambient air and the 1% CO₂ atmosphere); and (3) the interaction between genotypes and environmental conditions (i.e. GxE effect). These analyses identified 13,514 DEGs (66% of the expressed genome) whose expression is associated with one or more of these three modifiers. Approximately half of these DEGs are singularly affected by the G, E, or GxE effect (Fig. 11A; Supplemental Data S3). Multiple modifiers combine to affect the expression of the remaining DEGs, these being combinations of three subsets that combine two modifiers (G and E, 15% of DEGs; G and GxE, 7% of DEGs; and E and GxE, 8% of DEGs) and one subset that combine all three modifies (17% of DEGs; Fig. 11A; Supplemental Data S3).

Figure 11.

Clustering and functional annotation analyses of genes that are differentially expressed in response to the two mtfas mutations. A, Venn diagram representing the classification of DEGs responding to genotypes (G), environmental conditions (E), and the interaction between genotypes and environmental conditions (GxE). The number of DEGs in each set and subset, and the gene expression clusters that belong to each subset (as defined in B), are indicated. B, Categorization of the expression patterns of 13,514 DEGs among 34 different expression clusters. In each graph, the y axis represents the log₂-transformed ratio between the expression value of each condition and the mean expression value across all six conditions. Each gray line indicates the expression pattern of an individual DEG, and the thick black line identifies the average expression pattern of all genes in the cluster. The x axis identifies the genotypes of the plants and the environmental conditions under which they were grown. The code for the x axis ordinances is as follows: WA, wild-type plants in ambient air; hA, mthd-rnai-1 mutant in ambient air; kA, mtkas-2 mutant in ambient air; WC, wild-type plants in the 1% CO₂ atmosphere; hC, mthd-rnai-1 mutant in the 1% CO₂ atmosphere; and kC, = mtkas-2 mutant in the 1% CO₂ atmosphere. C, Categorization of each gene expression cluster (B) associated with one of five biological functions. These functions were heuristically determined from the average differential expression pattern defined by each expression cluster.

Additional insights on the metabolic impact of these modifiers were obtained by dissecting each of the seven subsets illustrated by the Venn diagram shown in Figure 11A into clusters that visualize the expression pattern relative to the three genotypes and two different environmental conditions (Fig. 11B). For example, the 1,390 DEGs that respond only to the G effect can be separated into two clusters: cluster 1 contains DEGs that are up-regulated in each mutant and cluster 2 contains DEGs that are down-regulated in each mutant, but all of these genes are minimally affected by the change in environmental conditions. Therefore, these are genes whose expression is affected by the mtFAS deficiency, irrespective of the atmospheric CO₂ concentration. Therefore, these DEGs probably are not associated with the photorespiratory deficiency, as the latter would be expected to show distinct gene expression patterns between plants grown in ambient air and the 1% CO₂ atmosphere.

Similarly, DEGs that respond only to the E effect can be separated into two clusters: cluster 3 contains DEGs whose expression is higher in the 1% CO₂ atmosphere, whereas cluster 4 contains DEGs whose expression is lower in the 1% CO₂ atmosphere, but all of these genes are minimally affected by differences in genotypes (Fig. 11B). Therefore, these DEGs are responding mainly to the change in the atmospheric CO₂ concentration and not to the mtFAS deficiency.

DEGs that respond to only the GxE effect also can be separated into two clusters. But these DEGs demonstrate interdependence between the G and E effects in altering gene expression (Fig. 11B). Specifically, cluster 5 contains DEGs whose expression is up-regulated in each of the mutants when grown in ambient air but down-regulated when grown in the 1% CO₂ atmosphere. DEGs in cluster 6 are oppositely affected, in that the expression of these genes is down-regulated in the mutants in ambient air but up-regulated in the mutants in the 1% CO₂ atmosphere.

The subsets that integrate two or more modifiers of gene expression are clustered in multiple expression patterns, and these appear to be additive of the patterns of each individual modifier (Fig. 11B). For example, genes whose expression is affected by both the G effect and the E effect can be separated into six clusters (clusters 7 to 12). Similarly, clusters 13 to 34 are identifiable by examining the patterns of gene expression that respond to two (i.e. G and GxE or E and GxE effects) or all three modifiers. The expression of these latter DEGs shows complex patterns that are associated with both the mtFAS deficiency and variations in the atmospheric CO₂ concentration. An example is cluster 14, which contains DEGs whose response to genotypes is restricted to only the ambient air condition. Because the altered expression of DEGs in this cluster is suppressed when plants are grown in the 1% CO₂ atmosphere, we associate these genes as responding to the mtFAS-dependent photorespiratory deficiency.

By generally applying this rationale, we were able to identify clusters as containing DEGs that are associated primarily with the mtFAS deficiency and distinguish them from those that are associated with the change in the atmospheric CO₂ concentration, with the photorespiratory deficiency, or combinations thereof (Fig. 11C).

The biological functionalities of DEGs associated with the mtFAS-dependent processes, but independent of photorespiratory deficiency, were explored by GO enrichment analyses using AmiGO (Carbon et al., 2009). These analyses focused on gene annotations associated with DEGs in expression clusters 1, 2, and 7 to 12. Supplemental Figure 8A visualizes the results in the form of heat maps that categorize the DEG annotations with GO biological processes, GO subcellular components, and GO molecular functions (Supplemental Data S4). These analyses indicate that these eight expression clusters separate into three distinct clades, namely the clade that contains cluster 1, the clade that contains cluster 10, and the clade that contains clusters 2, 7, 8, 9, 11, and 12. Despite the fact that the three clades contain DEGs whose expression patterns are distinct, the GO annotation terms that are enriched with each clade are quite similar. This similarity is visualized in Tag Clouds that identify words enriched in the GO annotation terms (Supplemental Fig. 8B). Thus, all three clades have GO annotation terms that are rich in a few key words. From the biological process category, these words are biosynthetic and metabolic processes; from the subcellular compartment category, these words are intracellular, organelle, complex, and membrane; and from the molecular function category, these words are binding, activity, enzyme classes such as hydrolase and ligase, and metabolites such as purines, nucleosides, and nucleotides. The enrichment in these terms appears to indicate that mtFAS-dependent processes that are independent of photorespiration may be associated with the metabolism of nucleotides and nucleosides, and particularly with chloroplasts or plastids. This latter term enrichment is consistent with the data presented in Figure 7, which demonstrated that the most dramatic ultrastructural alteration in these mtFAS mutants is associated with alterations in the morphology of the chloroplasts and thylakoid membrane assembly.

DISCUSSION

The occurrence of a FAS system in mitochondria was first implied by the discovery of a mitochondrially located ACP in Neurospora crassa (Zensen et al., 1992). Since then, this FAS system has been characterized in yeast and humans (Hiltunen et al., 2010), confirming its multicomponent, type II character. In plant cells, to date only the mtACP isoforms (Shintani and Ohlrogge, 1994; Meyer et al., 2007), mtKAS (Olsen et al., 2004; Yasuno et al., 2004; Ewald et al., 2007), mtPPT (Guan et al., 2015), and mtMCS (Guan and Nikolau, 2016) have been experimentally identified and characterized as components that support mitochondrial fatty acid biosynthesis. In this study, we identified and characterized the Arabidopsis genetic locus AT5G60335 as encoding the mtHD component of the Arabidopsis mtFAS system, which catalyzes the third iterative reaction of the FAS cycle, the dehydration of the 3-hydroxyacyl-ACP intermediate. Evidence that supports this conclusion includes the following: (1) the mitochondrial localization of the protein encoded by the AT5G60335 gene locus, demonstrated by the use of a transgenic GFP gene fusion; (2) the genetic complementation of the growth deficiency of the htd2 yeast mutant strain that lacks mtHD catalytic function; (3) the in vitro characterization of the recombinant Arabidopsis AT5G60335-encoded protein, which directly demonstrates that this protein catalyzes the reverse reaction, the hydration of enoyl-CoA substrates; and (4) the demonstration that mutant strains deficient in the AT5G60335 gene and those deficient in the mtKAS gene (a known catalytic component of mtFAS) exhibit nearly identical phenotypic, metabolomic, and transcriptomic responses.

mtFAS-dependent functionalities were further investigated by a combined global transcriptomic and metabolomic analysis of mutants deficient in mtHD (i.e. mthd-rnai mutants) and mtKAS (i.e. mtkas mutants), each of which reduces mtFAS activity. These analyses expanded the understanding of the central role of mtFAS in lipoic acid homeostasis, which is of major significance in supporting photorespiration via the lipoylation of the H subunit of GDC. In addition, these analyses established novel functionalities associated with mtFAS, these being the generator of the fatty acid components for the assembly of lipid A-like molecules, and via an as yet unknown mechanism affecting the morphology of chloroplastic starch granules.

mtHD Contributes to Photorespiration

Previous characterizations of the Arabidopsis mtFAS components, such as mtKAS (Olsen et al., 2004; Yasuno et al., 2004; Ewald et al., 2007), mtPPT (Guan et al., 2015), and mtMCS (Guan and Nikolau, 2016), have indicated a role for mtFAS in supplying octanoyl-ACP as the precursor to synthesize mitochondrial lipoic acid, which is an essential cofactor of the H subunit of GDC, a crucial component of photorespiration. The mthd-rnai mutants characterized herein exhibit metabolic alterations that also are associated primarily with this inability to maintain photorespiration. Specifically, photorespiratory Gly hyperaccumulates and Suc and fatty acid accumulation is decreased; these metabolic alterations are reversed when these mutant plants are grown in a condition that suppresses photorespiration (i.e. a 1% CO₂ atmosphere). Therefore, the role of mtFAS to support lipoic acid biosynthesis appears to be nearly universal among all clades of eukaryotic organisms, as evidenced by characterization of the effects in yeast, human, and plants (Wada et al., 1997; Hiltunen et al., 2010).

To expand on the previously identified traits associated with mtFAS metabolism, we conducted a more comprehensive metabolomic analysis of the lipid species of mthd-rnai and mtkas mutants. Analyses of the lipidome and cuticle of these mutants reveal alterations in several lipid classes, including reduced accumulation of glycoglycerolipids, chlorophylls, and surface cuticular lipids and increased accumulation of leaf TAG. Somewhat unexpectedly, however, these alterations in lipid metabolism are associated with nonmitochondrial compartments (i.e. chloroplasts). These findings are evidenced by the ultrastructural phenotype of chloroplasts, namely less extensive thylakoid membranes that are composed of glycoglycerolipids and chlorophylls. Given the fact that these alterations in lipids and chloroplastic ultrastructure are completely reversed when these plants are grown in a nonphotorespiratory condition (i.e. 1% CO₂ atmosphere), they are likely to be secondary to the photorespiratory deficiency.

Some of these metabolomic changes can be interpreted with the global transcriptomic data revealed by the RNA-seq analysis of the two mtFAS mutants, such as surface cuticular lipid biosynthesis and plastidial TAG biosynthesis. In the former case, a few genes involved in the surface lipid biosynthesis (e.g. CER1, CER3, and CER5; Samuels et al., 2008) exhibit reduced expression levels in ambient air, which is reversed by the 1% CO₂ treatment. Considering the essential catalytic roles of CER1 and CER3 in producing alkanes from very-long-chain fatty acids (Bernard et al., 2012) and the role of the CER5 (i.e. an ABC transporter) in exporting surface cuticular lipids across the plasma membrane (Pighin et al., 2004), the reduced accumulation of surface cuticular lipids in the mtFAS mutant plants is consistent with these transcriptional alterations. In the latter case, the DGAT1 gene exhibited increased expression levels in ambient air, and this was reversed by the 1% CO₂ treatment. This finding reflects the importance of DGAT1 as a major regulator of the TAG biosynthetic pathway in seedling plants, as evidenced by the previous analysis of dgat1 mutants, which display a major reduction in leaf TAG accumulation (Zhang et al., 2009; Tjellström et al., 2015).

These examples illustrate the complexity of integrating transcriptomic and metabolomic data, which also is implied by principles of metabolic control analysis (Fell, 1992). Specifically, the regulation of surface cuticular lipid biosynthesis and plastidial TAG biosynthesis exemplifies a regulatory network that focuses the regulation of a metabolic pathway at a few reactions within a pathway (e.g. CER1-, CER3-, and DGAT1-catalyzed reactions).

Other complexities that can contribute to weak correlations between transcriptomic and metabolomic data are the roles of posttranscriptional (Gygi et al., 1999; Tian et al., 2004) and posttranslational regulatory mechanisms. Particularly, posttranslational regulation was exemplified in this study by the depletion of the essential lipoylation of the H protein of GDC, which inactivates the catalytic activity of this key enzyme of photorespiration, although the expression of H protein is normal in the mutant strains.

mtFAS Contributes a Fatty Acid Component for the Assembly of Lipid A-Like Molecules

It is generally believed that the substrate specificities of the mtHD and mtKAS enzymes are the major factors that determine the chain lengths of fatty acids that mtFAS systems can produce. For example, acyl substrates that are longer than 12 carbon atoms in length are poor substrates for the mtHD component in Trypanosoma brucei (Autio et al., 2008a) and the mtKAS component in humans (Zhang et al., 2005), which limits the fatty acid products of the mtFAS systems in these organisms. In contrast, our in vitro characterizations of the recombinant mtHD and mtKAS proteins demonstrate that these two enzymes display broader substrate specificity with respect to the chain lengths of substrates, up to 16 carbon atoms. Thus, we conclude that, in Arabidopsis, mtHD and mtKAS may not necessarily limit the chain lengths of the fatty acids that are generated by the plant mtFAS system. The Arabidopsis mtFAS system, therefore, may be able to produce both octanoyl-ACP for lipoic acid biosynthesis and longer chain acyl-ACP of potentially up to 18 carbons in length.

Consistent with this finding, we discovered that Arabidopsis cells appear to source 3-hydroxytetradecanoate from the mtFAS system to assemble lipid A-like molecules. This is evidenced by the fact that the accumulation of this fatty acid is depleted in the mthd-rnai and mtkas mutants, and its accumulation also is depleted in a mutant deficient in the assembly of lipid A-like molecules (i.e. atlpxa-1) but is unaffected in a photorespiratory mutant (i.e. shmt1-2), irrespective of atmospheric CO₂ conditions. In addition, the acyl transferases (i.e. AtLpxA, AtLpxD1, and AtLpxD2) that can use 3-hydroxytetradecanoyl-ACP as the substrate for the assembly of lipid A-like molecules are mitochondrially located (Séveno et al., 2010; Li et al., 2011). Considering the proteobacterial origin of mitochondria (Timmis et al., 2004), the finding that mtFAS generates the acyl moieties of lipid A-like molecules may imply that both the mtFAS system and the synthetic machinery for assembling lipid A-like molecules were acquired from the ancient proteobacterium that evolved to give rise to the modern mitochondria.

The Arabidopsis mtFAS system, therefore, generates not only the octanoyl-ACP precursor for lipoic acid biosynthesis (Ewald et al., 2007; Guan et al., 2015; Guan and Nikolau, 2016) but also the 3-hydroxytetradecanoyl-ACP that is the precursor for the assembly of lipid A-like molecules. It is interesting that the mtFAS system appears to terminate the fatty acid elongation cycle via acyl transferase mechanisms, which is distinct from the ptFAS system that terminates the elongation cycle via a hydrolytic mechanism catalyzed by acyl-ACP thioesterases (Jing et al., 2011). Specifically, mtFAS uses the acyl-transferases AtLpxA, AtLpxD1, and AtLpxD2 (Li et al., 2011) for the assembly of lipid A-like molecules and a combination of lipoyl transferase (Wada et al., 2001) and lipoic acid synthase (Yasuno and Wada, 1998) to generate lipoic acid.

The physiological functions of plant lipid A-like molecules are unclear (Armstrong et al., 2006). Because no morphological differences have been observed in plants that carry null alleles of genes that synthesize these molecules (Séveno et al., 2010; Li et al., 2011), it appears that lipid A-like molecules are not crucial for normal plant growth and development. Another unknown in this area of plant metabolism is the question of whether the four characterized lipid A-like molecules (i.e. lipid X, UDP-2,3-diacyl-GlcN, disaccharide-1-phosphate, and lipid IV_A; Li et al., 2011), which are metabolic intermediates in the assembly of lipid A in gram-negative bacteria, are intermediates of metabolism in plants or they represent the final products of the plant biosynthetic pathway. If they are intermediates, then what is the chemical nature of the final product that is produced by plants from these intermediates? Regardless, the fact that genes for the biosynthesis of lipid A-like molecules have been maintained during the evolution of plants, subsequent to the endosymbiotic events that evolutionarily gave rise to mitochondria, may be indicative of an unknown important physiological function that confers an evolutionary advantage.

Starch Granule Morphology Is Dependent on a Functional mtFAS System

An abnormal starch granule morphology occurs in the mthd-rnai and mtkas mutants, and this is maintained irrespective of an atmospheric growth condition that suppresses photorespiration. Therefore, starch granule morphology is a trait that is dependent on mtFAS function, but this trait is independent of photorespiration. The appearance of these abnormal starch granules resembles those that occur in mutant plants defective in the isoamylase-type starch-debranching enzymes (ISA), which assemble the starch polymer (Zeeman et al., 1998; Wattebled et al., 2005; Delatte et al., 2006). However, in contrast to the isa mutants that exhibit decreased accumulation of starch and increased accumulation of WSP (Facon et al., 2013; Lin et al., 2013), the mtFAS mutations do not impact starch and WSP accumulation.

The abnormal starch granule trait may be a secondary effect of the deficiency in the photorespiration-independent mtFAS functions, such as the deficiency in lipid A-like molecules, which are enriched within chloroplasts (Armstrong et al., 2006; Li et al., 2011) but are synthesized from mtFAS-derived fatty acids (i.e. 3-hydroxytetradecanoic acid). Considering that alterations in plastid membrane lipid ultrastructure result in abnormal starch granule morphology (Myers et al., 2011), the depletion of lipid A-like molecules in the mtFAS mutants could be causative in the alteration of the change in starch granule morphology.

In summary, this study has further expanded the understanding of the mtFAS system by biochemically and genetically identifying and characterizing the fourth enzymatic component (i.e. mtHD) of the system. Moreover, by comparing the properties of the mtHD component with the properties of the mtKAS component, we have formulated a more integrated comprehensive understanding of the role of mitochondria in generating acyl chains. These combined genetic and biochemical characterizations of individual components of the Arabidopsis mtFAS system have established that mutations in the mtPPT gene block plant development during embryogenesis (Guan et al., 2015). Mutations in the mtKAS gene (Ewald et al., 2007) and the mtMCS gene (Guan and Nikolau, 2016) and knockdown lines of the mtPPT gene (Guan et al., 2015) and the mtHD gene lead to the diminished lipoylation of the H subunit of GDC (Douce et al., 2001), an enzyme complex essential in photorespiration (Bauwe et al., 2010). The resulting loss of GDC activity manifests a deficiency in photorespiration. Moreover, the characterizations reported herein has expanded the physiological role of mtFAS that, in addition to generating the acyl precursor for lipoic acid biosynthesis, mtFAS has a role in determining the complexity of starch granule morphology. However, mechanistically understanding this interorganelle coordination will require additional molecular, genetic, and biochemical studies (Mueller and Reski, 2014).

MATERIALS AND METHODS

Protein Overexpression, Substrate Syntheses, and in Vitro Kinetic Assays

Genes encoding the Arabidopsis (Arabidopsis thaliana) mtHD and mtKAS and Streptococcus pneumoniae ACP synthase (SpACPS; McAllister et al., 2006) were chemically synthesized by GenScript after codon optimization for expression in Escherichia coli (for DNA sequences, see Supplemental Table S1). The E. coli acpP, fabD, and fabG genes were cloned from the wild-type K12 strain (CGSC, Yale University). All six genes were cloned into the pET30b vector (EMD Millipore), and the resulting constructs were named mtHD-pET (primers H1 and H2), mtKAS-pET (primers H3 and H4), SpACPS-pET (primers H5 and H6), acpP-pET (primers H7 and H8), fabD-pET (primers H9 and H10), and fabG-pET (primers H11 and H12; for primer sequences, see Supplemental Table S2). These constructs express recombinant proteins with a His tag located at the C terminus. Recombinant proteins were expressed in the E. coli BL21* strain (Invitrogen) and purified using Probond Nickel-Chelating Resin (Invitrogen).

For the kinetic assays of the mtHD enzyme, the enoyl-CoA substrates (i.e. trans-Δ²-10:1 and trans-Δ²-16:1) were synthesized from a chemical reaction between CoA (EMD Millipore) and trans-Δ²-decenoic acid (TCI Chemicals) or trans-Δ²-dodecenoic acid (Sigma-Aldrich) and purified using an HPLC system (Agilent 1200 HPLC system equipped with a Thermo Scientific Hypersil ODS column [250 mm length, 4 mm i.d., and 5 μm particle size]) as described previously (Förster-Fromme et al., 2008). Structures of these CoA derivatives were determined using a liquid chromatography-mass spectrometry system (Agilent 1100 LC/MSD system) as described previously (Ding et al., 2012). The hydratase activity of mtHD was assayed (for 10 min at 22°C) as described previously (Autio et al., 2008b). Methyl esters that were derived from the reaction products (i.e. 3-hydroxydecanoyl-CoA and 3-hydroxydodecanoyl-CoA) were directly quantified as described previously (Lu et al., 2008).

For the kinetic assays of the mtKAS enzyme, the acyl-ACP substrates were prepared from an SpACPS-catalyzed enzymatic reaction between apo-ACP and acyl-CoAs (Sigma-Aldrich) as described previously (Zhang et al., 2005); the reaction results in the formation of a mixture of acyl- and holo-ACP molecules. The percentage of acyl-ACP in the mixture was determined by MALDI-QTOF analysis. Similarly, holo-ACP was prepared from a reaction of ACP mixture and CoA (EMD Millipore). The kinetic assays for the mtKAS enzyme were performed as described previously (Zhang et al., 2005). Concentrations of three ingredients were optimized (i.e. 50 μm malonyl-CoA, 50 μm holo-ACP, and 100 nm recombinant mtKAS). In addition, 300 μm NADPH and 1 μm recombinant fabG were added in the reaction mix to monitor the reaction (for 5, 10, and 15 min at 22°C) by measuring the consumption of NADPH at 340 nm.

Kinetic values for mtHD and mtKAS are calculated using Prism version 5.0 (GraphPad Software).

E. coli Strains and Lipid A Analyses

The Arabidopsis AtLpxD1 (primers H13 and H14) and AtLpxD2 (primers H15 and H16) genes (the 5′ CDS that encodes the putative mitochondrial presequences was removed) were cloned into pBE522 vector (Zhu et al., 2011), resulting in AtlpxD1-pBE and AtlpxD2-pBE. In the presence of AtlpxD1-pBE or AtlpxD2-pBE, the lpxD gene of the E. coli BL21* strain was replaced by a Gen^R cassette using a PCR-based method (Datsenko and Wanner, 2000) with primers H17 and H18. Hybrid lipid A molecules were extracted from these strains and analyzed using a liquid chromatography-tandem mass spectrometry system (Agilent 1100 LC/MSD system) as described previously (Joo et al., 2012).

Yeast Strains and Genetic Complementation

The yeast Saccharomyces cerevisiae strain deficient in the HTD2 gene (YHR067W; BY4741 background; and MATa) was obtained from Thermo Scientific. The yeast HTD2 (primers H19 and H20) and COQ3 mitochondrial presequence (Hsu et al., 1996; primers H21 and H22) of the yeast BY4741 wild-type strain were cloned into YEp351 vector (PGK promoter-driven gene expression; de Moraes et al., 1995), resulting in HTD2-YEp and YEp351M, respectively. Arabidopsis mtHD (primers H23 and H24) and mtHD_tail (primers H24 and H25) were cloned into YEp351M (the 5′ CDS of 72 bp was replaced by the yeast COQ3 mitochondrial presequence CDS), resulting in mtHD-YEp and mtHD_tail-YEp, respectively. The complementation test was performed as described previously (Autio et al., 2008b).

Plant Strains and Genetic Transformations

The Arabidopsis genetic strains carrying the mthd-1 (CS856112), mtkas-2 (SALK_022295), mtkas-3 (SALK_087186), and shmt1-2 (SALK_083735) alleles are in the Columbia-0 background and were obtained from the Arabidopsis Biological Resource Center (http://abrc.osu.edu). The atlpxa-1 mutant (Wassilewskija background) was isolated from the Arabidopsis Functional Genomics Consortium T-DNA mutant population (Krysan et al., 1999). Primers for genotyping were as follows: H26-H27-H28 for mtkas-2 and mtkas-3; H26-H29-H30 for shmt1-2; and H31-H32-H33 for atlpxa-1.

For the GFP experiments, the 5′ mtHD CDS of 120 bp was cloned into pENTR/D-TOPO vector (Invitrogen; primers H34 and H35) and subcloned into pEarleyGate103 (Earley et al., 2006) using Gateway LR Clonase II Enzyme Mix (Invitrogen), resulting in mtHD_1-120-pEG. For the RNAi experiment, the mtHD CDS of 209 bp was cloned into pENTR/D-TOPO (primers H36 and H37) and subcloned into pB7GWIWG2(II) (Karimi et al., 2002), resulting in mtHD-WIWG2. Destination vectors were used to transform Arabidopsis Columbia-0 wild-type plants (in ambient air) or mtkas-2 mutant plants (in the 1% CO₂ atmosphere) as described previously (Clough and Bent, 1998).

Seeds were sterilized and sown on Murashige and Skoog agar medium as described previously (Jin et al., 2012). The CO₂ condition was controlled as the ambient level or the 1% CO₂ level (in a growth chamber). Plants were grown at 22°C with continuous illumination (photosynthetic photon flux density of 100 μmol m⁻² s⁻¹).

Microscopy

Confocal microscopy analysis was conducted on wild-type control plants, control transgenic plants carrying the p35S::GFP transgene (Guan et al., 2015), and transgenic plants carrying the p35S::HD_1-120-GFP transgene. Seedling plants at 7 to 10 DAI were harvested for analysis as described previously (Guan et al., 2015).

Light microscopy and transmission electron microscopy analyses were performed with plants at 16 DAI using Olympus BX-40 (Olympus America) and JEOL 2100 (Japan Electron Optic Laboratories), respectively, as mentioned previously (Myers et al., 2011).

Western Blot

Total protein was extracted from 200 mg of fresh aerial organs of plants at 16 DAI as described previously (Che et al., 2002). Western-blot analysis was performed on 50 μg of total protein with anti-lipoic acid antibodies (EMD Millipore) or with anti-H protein antibodies (a gift from Dr. David Oliver at Iowa State University) as described previously (Ewald et al., 2007; Guan et al., 2015).

Metabolomic Analyses

Metabolites (three to six replicates) were extracted from 50 mg of fresh aerial organs of plants (grown in a completely randomized design) at 16 DAI and analyzed using multiple analytical platforms: the Waters Xevo G2 Q-TOF MS equipped with the Waters ACQUITY UPLC system for glycerolipids and chlorophylls (Okazaki et al., 2015); the Agilent 1200 HPLC system equipped with a fluorescence detector for amino acids (Guan et al., 2015); the Total Starch Assay Kit (Megazyme) for starch and WSP (Smith and Zeeman, 2006); and the Agilent 7890 GC-MS system for fatty acids (Lu et al., 2008), surface cuticular lipids (Perera et al., 2010), and other soluble metabolites (Duran et al., 2003). Log₂ ratios and se were calculated as described previously (Quanbeck et al., 2012). Metabolomic data were deposited in the PMR database (Hur et al., 2013) under accession number mtFAS.

RNA-Seq

RNA-seq was performed as described previously (Li et al., 2015). Specifically, total RNA (three biological replicates) was extracted from 0.5 g of fresh aerial organs of plants (grown in a completely randomized design) at 16 DAI using TRIzol (Invitrogen) and was further purified using the RNeasy Plant Mini Kit (Qiagen) following DNase I (Invitrogen) treatment to remove any DNA contamination. The 200-bp short insert library was constructed and sequenced using the Illumina HiSeq 2000 system and V3 Reagent (www.illumina.com). Data files have been deposited in the NCBI Sequence Read Archive under accession numbers SRP052705 and SRP070190.

Raw reads were cleaned by removing adaptors and filtering reads with unknown nucleotides larger than 5% or with low quality (the percentage of bases that have quality lower than 10 is more than 20% of the read). The cleaned reads were aligned to the reference Arabidopsis genome in Phytozome version 8.0 (www.phytozome.net) using TopHat (Trapnell et al., 2009), and the mapped reads were counted using HTSeq (Anders et al., 2015).

Statistical Analyses

The count data were analyzed as a completely randomized design with six treatments, one for each combination of genotype and growth condition. Each treatment was replicated three times. Genes with an average of at least one uniquely mapped read across samples and a number of nonzero read counts at least as large as the number of treatments were tested for differential expression among genotypes within each growth condition, differential expression between growth conditions within each genotype, differential expression among genotypes averaged over growth conditions (i.e. genotype main effects), differential expression between growth conditions averaged over genotypes (i.e. growth condition main effects), and interactions between genotypes and growth conditions using the R package QuasiSeq (http://cran.r-project.org/web/packages/QuasiSeq). The negative binomial QLShrink method implemented in the QuasiSeq package as described previously (Lund et al., 2012) was used to compute a P value for each gene and each test described above. The log of each count mean was modeled as the sum of an intercept term, a genotype effect, a growth condition effect, an interaction between genotype and growth condition, and an offset normalization factor, determined for each sample by the log of the TMM normalization factor (Robinson and Oshlack, 2010). Estimates of the fold change were computed by evaluating the exponential function at estimates of effect differences. Using the P values for each comparison, an approach described previously (Nettleton et al., 2006) was used to estimate the number of genes with true null hypotheses among all genes tested, and this estimate was used to convert the P values to q values (Storey, 2002). To obtain approximate control of the false discovery rate at 5%, genes with q values no larger than 0.05 were declared to be differentially expressed.

Genes with at least one significant main effect or a significant interaction between genotypes and environmental conditions were divided into seven distinct sets, and a cluster analysis (Fang et al., 2012; Ramundo et al., 2014; Schmollinger et al., 2014) was then performed separately on each set. These sets were defined by each possible combination of the three tests, excluding genes without a significant main effect or interaction effect. For example, one subset is composed of genes with a significant genotype effect but not a significant environmental condition effect nor a significant interaction effect. Likewise, another subset consists of genes with a significant environmental condition effect but not a significant genotype or interaction effect. The remaining five sets similarly each represent a unique combination of genotype, environmental condition, and interaction effects. A model-based clustering algorithm, implemented in the R package MBCluster.Seq (Si et al., 2014), was applied to identify distinct gene expression patterns among the genes in each set. The RNA-seq read counts were assumed to be drawn from a mixture of negative binomial distributions, each of which represents a cluster. After accounting for normalization factors and each gene’s overall mean expression level, genes were assigned to clusters based on their expression profiles or the collection of estimated fold changes across treatments and genotypes. The number of clusters in each set of genes was chosen using the Akaike information criteria.

Quantitative RT-PCR

Quantitative RT-PCR was performed using the StepOnePlus Real-Time PCR System and SYBR Select Master Mix (Invitrogen) as described previously (Jin et al., 2012). RNA samples for the expression study were used as described previously (Guan et al., 2015). Gene-specific primers were used as follows: H38 and H39 for ACTIN2 (reference gene; AT3G18780), H40 and H41 for mtHD (AT5G60335), H42 and H43 for mtKAS (AT2G04540), H44 and H45 for RNP1 (AT2G32230), H46 and H47 for SQD2 (AT5G01220), H48 and H49 for PPH (AT5G13800), H50 and H51 for LPAAT (AT4G24160), H52 and H53 for DGAT1 (AT2G19450), H54 and H55 for CER1 (AT1G02205), H56 and H57 for CER2 (AT4G24510), H58 and H59 for CER3 (AT5G57800), and H60 and H61 for MCCA (AT1G03090).

Accession Numbers

Sequence data from this article can be found in the GenBank/EMBL data libraries under the following accession numbers: mtHD, AT5G60335, gene ID 10723110; and mtKAS, AT2G04540, gene ID 814996.

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. Mass spectrometric characterization of the acyl-CoA and acyl-ACP substrates synthesized for the in vitro enzyme assays of mtHD- and mtKAS-catalyzed reactions.

• Supplemental Figure S2. Characterizations of the mthd-1 mutant allele.

• Supplemental Figure S3. Metabolomic alterations in the mthd-rnai-2 and mtkas-3 mutants.

• Supplemental Figure S4. Metabolic alterations in the mtkas-2-mthd-rnai double mutant.

• Supplemental Figure S5. Morphological, ultrastructural, and metabolomic alterations in response to the feeding of 3% Suc or 10 mm glycolate.

• Supplemental Figure S6. Mass spectrometric characterization and identification of 3-hydroxytetradecanoic acid and lipid A.

• Supplemental Figure S7. Validation of RNA-seq gene expression data by quantitative RT-PCR analysis.

• Supplemental Figure S8. GO enrichment analysis of DEGs associated with the mtFAS biological function.

• Supplemental Table S1. Codon-optimized synthetic gene sequences.

• Supplemental Table S2. Nucleic acid sequences of DNA primers used in this study.

• Supplemental Data S1. Statistical test for DEGs among treatment groups.

• Supplemental Data S2. Gene counts associated with each GO functional category in each comparison group.

• Supplemental Data S3. Statistical test for genes whose expression is associated with main effects.

• Supplemental Data S4. GO terms enriched in each gene cluster relative to biological processes, subcellular components, and molecular functions.

Glossary

TAG

triacylglycerol

CDS

coding sequence

RT

reverse transcription

DAI

days after imbibition

WSP

water-soluble Glc polysaccharide

RNA-seq

RNA sequencing

DEG

differentially expressed gene

GO

Gene Ontology