Protein interactomes for plant lipid biosynthesis and their biotechnological applications

Summary

Plant lipids have essential biological roles in plant development and stress responses through their functions in cell membrane formation, energy storage and signalling. Vegetable oil, which is composed mainly of the storage lipid triacylglycerol, also has important applications in food, biofuel and oleochemical industries. Lipid biosynthesis occurs in multiple subcellular compartments and involves the coordinated action of various pathways. Although biochemical and molecular biology research over the last few decades has identified many proteins associated with lipid metabolism, our current understanding of the dynamic protein interactomes involved in lipid biosynthesis, modification and channelling is limited. This review examines advances in the identification and characterization of protein interactomes involved in plant lipid biosynthesis, with a focus on protein complexes consisting of different subunits for sequential reactions such as those in fatty acid biosynthesis and modification, as well as transient or dynamic interactomes formed from enzymes in cooperative pathways such as assemblies of membrane‐bound enzymes for triacylglycerol biosynthesis. We also showcase a selection of representative protein interactome structures predicted using AlphaFold2, and discuss current and prospective strategies involving the use of interactome knowledge in plant lipid biotechnology. Finally, unresolved questions in this research area and possible approaches to address them are also discussed.

Introduction

Acyl lipids are the most abundant lipids in plants and have essential biological roles in plant architecture, development and stress responses through their functions in cell membrane formation, energy storage and signalling (Yang and Benning, 2018; Yoshihara and Kobayashi, 2022). Vegetable oils (mainly triacylglycerols, or TAGs) are among the most important agricultural commodities globally, with great value and important applications in food, fuel and oleochemical industries (Biermann et al., 2021; Lu et al., 2011; Wang, Singer, et al., 2022). Acyl lipid metabolism is highly compartmentalized and involves the coordinated action of multiple pathways in different subcellular organelles, including the de novo biosynthesis and modification of fatty acids, assembly and storage of complex acyl lipids and their molecular regulation. Despite the fact that some notable advances have broadened our understanding of lipid biosynthesis over the past few decades, considerable gaps still exist in our comprehension of these complex processes in plants (Chapman and Ohlrogge, 2012). In recent years, mounting evidence has suggested that various protein players in lipid biosynthetic pathways can form dynamic protein–protein/protein–lipid interactomes or metabolons, echoing an emerging field of interactomes/metabolons for plant metabolism (Nakamura, 2017). Moreover, dynamic lipid metabolism interactomes may exist and have important functions in lipid metabolism in other organisms. For example, similar dynamic protein interactomes have also been proposed for mammalian lipid metabolism (Coleman, 2019); however, further investigation will be required for their identification and to determine how they constitute regulatory mechanisms that facilitate lipid metabolism. The concepts of physical enzyme–enzyme interaction and metabolons (transient multi‐enzyme complexes) were first described in 1970 and 1985, respectively, as permanent or transient enzyme complexes allowing the channelling of metabolic pathway flux (Zhang and Fernie, 2020). Although it is generally assumed that interactomes/metabolons can facilitate local substrate channelling and thus are important for lipid biosynthesis, research into the nature of these interactomes and their molecular mechanisms is at an early, yet rapidly developing, stage.

Here, we review recent advances in the identification and characterization of protein interactomes involved in lipid biosynthesis in plants (Table 1) with a focus on: (i) protein complexes consisting of different subunits for sequential reactions, including acetyl‐CoA carboxylase (ACCase), fatty acid synthase (FAS) and fatty acid elongase (FAE) and (ii) transient or dynamic interactomes formed from enzymes in cooperative pathways (e.g. assemblies of membrane‐bound enzymes for TAG biosynthesis). For discussions of interactomes involving lipid droplet proteins and for specialized oil biosynthesis, the reader is encouraged to consult three recent reviews (Busta et al., 2022; Guzha et al., 2022; Pyc et al., 2017). We also predict and illustrate the structures of a selection of representative lipid biosynthetic interactomes (Figures 1 and 2) using AlphaFold2, which is an emerging machine learning‐based prediction programme for protein–protein interactions (Bryant et al., 2022; Mirdita et al., 2022) and then compared these with known biochemical results. Furthermore, we examine current progress on the use of interactome‐related knowledge to guide biotechnological efforts in the engineering of plant oil production. Finally, challenging questions that remain regarding protein interactomes in plant lipid metabolism, as well as possible approaches to investigate them further, are addressed.

Table 1.

Enzyme complexes and interactomes identified for lipid biosynthesis in plants

Pathway/protein	Enzyme/protein interaction	Protein–protein interaction identification method	Plant species	References
Fatty acid biosynthesis (heteromeric ACCase)	Heteromeric ACCase complex, BC/BCCP subcomplex, α‐CT and β‐CT subcomplex	Pea chloroplast import, SEC, AP	Soybean (Glycine max)	Reverdatto et al. (1999)
	PII with BCCP, BADC	AP, MS	Arabidopsis thaliana	Feria Bourrellier et al. (2010)
	BADC with BCCP, BC	Co‐IP, MS, Y2H	Arabidopsis thaliana	Salie et al. (2016)
	BCCP/BADC/BC subcomplex	AP, SEC	Arabidopsis thaliana	Shivaiah et al. (2020)
	α‐CT and CTI	Y2H, BiFC, MST	Arabidopsis thaliana	Ye, Nikovics, et al. (2020)
Fatty acid biosynthesis (type II FAS)	KAR homotetramer with ACP	X‐ray structure and docking	Brassica napus	Fisher et al. (2000), Price et al. (2001)
Fatty acid biosynthesis (ACP and FAT)	Spinach ACP and Jatropha curcas FAT	Docking	Jatropha curcas	Srikanta Dani et al. (2011)
	ACP and FAT	Docking and cross‐linking	Chlamydomonas reinhardtii	Beld et al. (2014), Blatti et al. (2012)
Fatty acid elongation (FAE)	KCR/ECR/HCD complex	BiFC	Arabidopsis thaliana	Bach et al. (2008); Roudier et al. (2010)
	PAS1 with KCR, ECR, HCD	BiFC, AP	Arabidopsis thaliana	Roudier et al. (2010)
	CER2 with KCR, ECR, HCD, KCS	Split‐LUC	Arabidopsis thaliana	Haslam and Kunst (2021)
	PAS1/KCR/ECR/HCD complex; KCS homo‐ and heterooligomer; KCS with ECR, HCD	Y2H, BiFC	Arabidopsis thaliana	Kim et al. (2022)
Fatty acid desaturation (FAD)	SAD‐SAD	SEC	Castor (Ricinus communis)	McKeon and Stumpf (1982)
	Self‐interaction: FAD2, 3, 6, 7, 8; FAD2‐FAD3; FAD6‐FAD7; FAD6‐FAD8	Y2H, split‐LUC, cross‐linking	Arabidopsis thaliana	Lou et al. (2014)
	Castor SAD with Anabaena ferredoxin	Cross‐linking, docking	Castor (Ricinus communis) and Anabaena	Sobrado et al. (2006)
	FAD4 with PEROXIREDOXIN Q	Docking	Arabidopsis thaliana	Horn et al. (2020)
	Ferredoxin with FAD	Y2H	Chlamydomonas reinhardtii	Yang et al. (2015)
TAG assembly and acyl editing	DGAT1‐DGAT1	SEC, cross‐linking	Brasscia napus	Weselake et al. (2006)
	GPAT8‐DGAT2; GPAT8‐GPAT9; GPAT8‐GPAT8	Y2H	Tung tree (Vernicia fordii)	Gidda et al. (2011)
	GPAT9‐LPAAT; GPAT9‐LPCAT; DGAT1‐LPAAT; DGAT1‐LPCAT; GPAT9‐GPAT9	Y2H	Arabidopsis thaliana	Shockey et al. (2016)
	DGAT1‐LPCAT; DGAT1‐PDCT; DGAT2‐LPCAT; DGAT2‐PDCT; LPCAT‐PDAT; LPCAT‐PDCT; DGAT1‐DGAT; LPCAT‐LPCAT; PDCT‐PDCT	Y2H, BiFC	Flax (Linum usitatissimum)	Xu et al. (2019)
	DGAT1‐PDAT	BiFC, Co‐IP	Arabidopsis thaliana	Lee and Seo (2019)
	At: DGAT1‐PDCT; DGAT1‐LPCAT; DGAT1‐PDAT; DGAT1‐DGAT1; Rc/Gm: Rc/GmDGAT2‐Rc/GmDGAT2; Rc/GmDGAT2‐AtPDCT; Rc/GmDGAT2‐AtLPCAT; Rc/GmDGAT2‐AtPDAT	Y2H	Arabidopsis thaliana, Castor (Ricinus communis), Soybean (Glycine max)	Regmi et al. (2020)
Galactolipid biosynthesis	RBL10 with ACP4, CTI1	BN‐PAGE, SEC, Co‐IP, MS, Y2H	Arabidopsis thaliana	Lavell et al. (2021)
Lipid trafficking (TGD)	TGD2‐TGD2; TGD1,2,3 complex	BN‐PAGE, SEC, BN‐SDS 2D gel, MS	Arabidopsis thaliana	Roston et al. (2011, 2012)
	TGD5 with TGD1,2,3,4	Co‐IP	Arabidopsis thaliana	Fan et al. (2015)
Phospholipid metabolism	ACBP2 and lysophospholipase	Y2H, Co‐IP, ITC	Arabidopsis thaliana	Gao et al. (2010), Miao et al. (2019)

AP, affinity purification; BiFC, bimolecular fluorescent complementation assay; BN‐PAGE, blue native polyacrylamide gel electrophoresis; Co‐IP, co‐immunoprecipitation; ITC, isothermal titration calorimetry; MS, mass spectrometry; MST, microscale thermophoresis; SEC, size‐exclusion chromatography; split‐LUC, split‐luciferase complementation; Y2H, yeast two‐hybrid assay.

Figure 1.

Predicted structures of protein complexes for fatty acid biosynthesis using AlphaFold2. (a) Predicted structure of Arabidopsis thaliana α‐CT/β‐CT heterotetramer. The structure was superimposed by aligning with Escherichia coli α‐CT/β‐CT heterotetramer (shown in light blue and purple in the top panel, PDB: 2F9Y). The lower panel shows the predicted structure being coloured according to pLDDT values of residues. (b) Predicted structure of A. thaliana KAR homotetramer. The structure was superimposed by aligning with E. coli KAR homotetramer (shown in light colour, PDB: 1Q7B). (c) Predicted structure of A. thaliana HAD homodimer interacting with 2 ACPs. The structure was superimposed by aligning with E. coli HAD and ACP complex (shown in light colour, PDB: 4KEH). (d) The top three predicted structures of A. thaliana KCS homodimer. (e) The top three predicted structures of A. thaliana FATA and ACP interaction. The structure coloured according to pLDDT values of residues was also shown.

Figure 2.

Predicted protein–protein interaction structures of Linum usitatissimum LPCAT and DGAT2 and Arabidopsis thaliana LPCAT and DGAT1 using AlphaFold2. (a) Predicted protein–protein interaction between L. usitatissimum LPCAT and DGAT2. (b) Predicted protein–protein interaction between A. thaliana LPCAT and DGAT1. The hydrophobic interactions are indicated by dashed yellow lines. The structures coloured according to pLDDT values of residues were also shown.

Protein complexes consisting of different subunits for sequential reactions

Acetyl‐CoA carboxylase

De novo fatty acid synthesis in higher plants occurs mainly in plastids, where the first committed step is the formation of malonyl‐CoA catalysed by ACCase (for an in‐depth discussion, see Salie and Thelen, 2016). ACCase contains four components including biotin carboxylase (BC), biotin carboxyl carrier protein (BCCP) and α‐ and β‐carboxyltransferases (CTs). Higher plants have two forms of ACCases: a heteromeric ACCase composed of four subunits to form a multi‐enzyme complex, which is the prokaryotic type and related to the well‐studied Escherichia coli ACCase, and a homomeric ACCase with all four enzymatic components concatenated into a single polypeptide. All plants have a homomeric isoform in the cytosol, which contributes to very long chain fatty acid biosynthesis and other metabolic processes. Heteromeric ACCase is the predominant form in the plastids of dicots and non‐graminaceous monocots, with β‐CT being encoded by the chloroplast genome and the remaining subunits being imported from the cytosol. Conversely, homomeric ACCase is the predominant form in the plastids of graminaceous monocots. The heteromeric ACCase is the most extensively studied isoform in plants, and this enzyme can be dissociated into two stable subcomplexes containing BC/BCCP and α‐CT/β‐CT, which are responsible for two half‐reactions to convert acetyl‐CoA to malonyl‐CoA, respectively (Reverdatto et al., 1999; Salie and Thelen, 2016). On the BC/BCCP subcomplex, a biotin cofactor that is covalently attached to BCCP via a lysine residue is carboxylated by BC using CO₂ from bicarbonate and energy from ATP hydrolysis. The α‐CT/β‐CT subcomplex then catalyses the carboxylation of acetyl‐CoA using the carboxyl group from the carboxylated biotin to yield malonyl‐CoA.

While structural information on plant heteromeric ACCase is lacking, the complex structures of E. coli orthologs have been well investigated, with the α‐CT/β‐CT subcomplex having been shown to be an α2β2 heterotetramer (Bilder et al., 2006). To further our understanding of heteromeric ACCase in plants, we predicted the subcomplex structure of the α‐CT/β‐CT heterotetramer from Arabidopsis thaliana by taking advantage of the deep learning structure prediction programme AlphaFold2. The core catalytic regions of A. thaliana α‐CT and β‐CT are well aligned with E. coli α‐CT and β‐CT in the form of a heterotetramer (Figure 1a). Nevertheless, A. thaliana α‐CT and β‐CT contain large domains (200–300 amino acids in length) with unknown functions at the C‐ and N‐termini, respectively, which are not present in the E. coli orthologs. These domains were predicted to form interactions as well, but the confidence level is very low (plDDT < 50, Figure 1a). These domains are not irrelevant to the catalytic activity of ACCase and are speculated to serve a regulatory role, potentially in association with the plastid envelope membrane (Salie and Thelen, 2016).

Furthermore, heteromeric ACCase interacts with many protein regulators to modulate its activity (Table 1) (Salie and Thelen, 2016). For example, PII, which is a homotrimeric protein involved in the regulation of nitrogen metabolism, was found to interact with heteromeric ACCase and function as a negative regulator, whereby PII binds to biotinylated BCCP and affects its ability to act as a swinging arm in the transportation of carboxylated biotin between BC and CT, thus inhibiting ACCase activity (Feria Bourrellier et al., 2010; Gerhardt et al., 2015). In addition to BCCP, a group of biotin attachment domain‐containing proteins (BADC) was also identified from the eluate of PII‐affinity chromatography (Feria Bourrellier et al., 2010). BADC shares sequence similarity with BCCP but lacks the biotinylation motif and was recently found to interact with BCCP and BC (Keereetaweep et al., 2018; Salie et al., 2016). BCCP forms a homodimer, and its interaction with BADC likely supports a regulatory role for the BCCP:BADC heterodimer in ACCase activity (Salie and Thelen, 2016). Indeed, BADC is able to bind to BC but cannot be carboxylated by BC due to the lack of the biotinylation motif, which means it might compete with BCCP for access to BC and serves as a negative regulator of heteromeric ACCase (Keereetaweep et al., 2018; Salie et al., 2016). Recently, however, biochemical characterization using purified enzymes demonstrated that BADC proteins instead have a role in stabilizing the heteromeric ACCase complex, since BADC was found to facilitate the formation of the BC/BCCP/BADC subcomplex, allowing it to more readily interact with the α‐CT/β‐CT subcomplex (Shivaiah et al., 2020). This finding suggests that BADC may activate ACCase activity rather than inhibit it (Shivaiah et al., 2020). The association of BC with BADC and BCCP was found to be modulated by light‐dependent swings in pH, whereby BC has a higher affinity for BADC over BCCP at pH 7 (dark) but shifts its preference to BCCP at pH 8 (light), suggesting their contribution to the light‐dependent regulation of ACCase activity (Ye, Fulcher, et al., 2020). Furthermore, carboxyltransferase interactor (CTI), a small plastidial membrane protein, was recently found to interact with the C‐terminal non‐catalytic domain of α‐CT, which is proposed to mediate the docking of ACCase to the plastid envelop membrane and attenuate carbon flux into fatty acid biosynthesis (Ye, Nikovics, et al., 2020).

Fatty acid synthase

Malonyl‐CoA is then converted into malonyl‐acyl carrier protein (ACP) through the catalytic action of malonyl‐CoA:ACP acyltransferase. The malonyl‐ACP then enters subsequent steps of fatty acid biosynthesis catalysed by the FAS complex in plastids. FAS utilizes acetyl‐CoA as the starting unit and malonyl‐ACP as the two‐carbon unit donor to elongate the fatty acid chain up to 18 carbons, while the growing acyl chain is bound to an ACP protein during this process. Plastidial FAS is a type II FAS related to prokaryotic/bacterial FAS II and consists of four discrete enzymes catalysing each step of fatty acid elongation, which contrasts with the type I FAS in mammals and yeast, which contains all four enzymatic components in a single polypeptide in the cytosol. The plastidial FAS catalyses fatty acid elongation through the following four steps: (1) condensation of malonyl‐ACP and acetyl‐CoA/acyl‐ACP to form 3‐ketoacyl‐ACP by 3‐ketoacyl‐ACP synthase (KAS), (2) reduction of 3‐ketoacyl‐ACP to form 3‐hydroxyacyl‐ACP by 3‐ketoacyl‐ACP reductase (KAR) using NADPH, (3) dehydration of 3‐hydroxyacyl‐ACP to form enoyl‐ACP by hydroxyacyl‐ACP dehydratase (HAD) and (4) reduction of enoyl‐ACP to form fatty acyl‐ACP by enoyl‐ACP reductase (ENR) using NAD(P)H (Beld et al., 2015; Mou et al., 2000; Wu and Xue, 2010). In plants, three isoforms of KAS condensing enzymes have been identified. To produce an 18‐carbon fatty acid, the three KASs are required to cooperate together with KASIII catalysing the initial condensation of acetyl‐CoA and malonyl‐ACP to form 4:0‐ACP, KASI catalysing the subsequent condensations up to 16:0‐ACP and KASII catalysing the final condensation from 16:0‐ACP to 18:0‐ACP. 18:0‐ACP can further be desaturated to 18:1 (18:1Δ ^9cis )‐ACP by a soluble Δ9‐acyl‐ACP desaturase (SAD). The resulting 16‐carbon or 18‐carbon ACPs are then hydrolysed via the catalytic action of acyl‐ACP thioesterases (FATA or FATB) to release free fatty acids, whereby FATA prefers 18:1‐ACP as the substrate and FATB prefers saturated acyl‐ACPs such as 16:0‐ACP (For a review, see Li‐Beisson et al., 2013).

Protein–protein interactions are important for the function of FAS since the individual type II FAS subunits, including KAS, KAR, HAD and ENR, are all oligomers and interact with ACP (Chen et al., 2018; Fisher et al., 2000). While KAS (I, II and III) and HAD exist as homodimers, KAR and ENR are tetramers, ACP is known to functionally bind to all these components, as well as other enzymes (e.g. thioesterases), to enable acyl chain elongation and unloading and thus determine acyl chain length (Beld et al., 2015; Chen et al., 2018). The type II FAS from E. coli has been extensively studied and the structures of individual FAS subunits have been resolved, but how these subunits work together remains to be elucidated. When we predicted the structures of A. thaliana KAR homotetramer (Figure 1b) and HAD homodimer interacting with two ACPs (Figure 1c) using AlphaFold2, predicted structures showed good confidence and matched well with the known structure models of E. coli orthologs. We also predicted the complex structure of A. thaliana fatty acid thioesterase (FATA) and ACP (Figure 1e), since structural information on the interaction between these proteins is currently lacking. Our resulting prediction was consistent with the previous docking results of Chlamydomonas reinhardtii ACP and FAT, a unique FAT with both FATA and FATB characteristics (Blatti et al., 2012), suggesting that helix 2 of ACP may be involved in its interaction with FAT.

Fatty acid elongase

In plants, plastid‐derived fatty acids (C18) can be further elongated to very long chain fatty acids on the ER through the activity of a similar multi‐subunit enzyme complex termed FAE, using malonyl‐CoA as the two‐carbon unit donor (For a review, see Li‐Beisson et al., 2013). In a similar fashion to FAS, four steps are involved in each elongation cycle: (1) condensation of malonyl‐CoA and acyl‐CoA by 3‐ketoacyl‐CoA synthases (KCS), (2) reduction of 3‐ketoacyl‐CoA by 3‐ketoacyl‐CoA reductase (KCR), (3) dehydration of 3‐hydroxyacyl‐CoA by hydroxyacyl‐CoA dehydratase (HCD) and (4) reduction of enoyl‐CoA to form fatty acyl‐CoA by enoyl‐CoA reductase (ECR) (Zheng et al., 2005). The model plant A. thaliana possesses 21 KCS enzymes with different chain length specificities but one KCR, one HCD and one ECR. In line with this, it has been suggested that different FAE complexes are invariant except for their KCS enzyme, which allows for the production of acyl chains with different lengths (Blacklock and Jaworski, 2006; Haslam and Kunst, 2013; Joubès et al., 2008).

The four core components of the FAE complex seem to exist in oligomeric states, which is supported by the observation of homomeric and heteromeric interactions between them using bimolecular fluorescence complementation (BiFC) and/or yeast two‐hybrid assays (Kim et al., 2022), as well as the AlphaFold2 prediction of the KCS homodimer (Figure 1d). In addition, many other proteins can interact with FAE and affect very long chain fatty acid biosynthesis. For example, PASTICCINO1 (PAS1), an immunophilin protein, was observed to interact with core FAE subunits, including KCR, HCD and ECR, using BiFC and/or pulldown assays and is required for very long chain fatty acid biosynthesis, possibly by acting as a molecular scaffold (Roudier et al., 2010). Similarly, the Eceriferum2‐like protein CER2 was found to physically interact with all the four subunits of the FAE complex using a split‐luciferase assay and is necessary for the production of very long chain fatty acids up to 30‐ to 36‐carbon, likely by affecting condensing enzyme function (Haslam and Kunst, 2021).

Interactomes for triacylglycerol assembly

In addition to fatty acid biosynthesis, mounting evidence suggests that protein–protein interactions may play important roles in TAG assembly and remodelling. TAG assembly in plants involves the sequential acylation of sn‐glycerol‐3‐phosphate (G3P) by sn‐glycerol‐3‐phosphate acyltransferase (GPAT), and lysophosphatidic acid acyltransferase (LPAAT) at sn‐1 and 2 positions, respectively. The resulting phosphatidic acid (PA) is then converted to diacylglycerol (DAG), which can be further acylated by acyl‐CoA:diacylglycerol acyltransferase (DGAT) and phospholipid:diacylglycerol acyltransferase (PDAT) to produce TAG (Chen et al., 2022). In addition, enzymes such as acyl‐CoA:lysophosphatidylcholine acyltransferases (LPCAT) and phosphatidylcholine diacylglycerol cholinephosphotransferase (PDCT) are involved in acyl exchange and remodelling.

Recently, these ER membrane‐bound transferases for glycerolipid assembly and remodelling were found to form interactomes according to results from BiFC and yeast two‐hybrid assays (Table 1). A transferase interactome has been proposed to facilitate channelling of fatty acyl moieties from phosphatidylcholine (PC) to TAG in flax (Linum usitatissimum), whereby the TAG synthesizing enzymes DGAT2 and PDAT were found to directly interact with the PC‐remodelling enzymes LPCAT and/or PDCT, and interactions between LPCAT and PDCT were also observed (Xu et al., 2019). Similarly, interactomes of acyltransferases for TAG biosynthesis have also been reported in tung tree (Vernicia fordii) and A. thaliana. For example, tung tree GPAT8 was observed to interact with DGAT2 and GPAT9 (Gidda et al., 2011), while A. thaliana GPAT9 was found to interact with LPAAT and LPCAT, and DGAT1 appears to interact with LPAAT and LPCAT, but to a much lesser degree (Shockey et al., 2016). In addition, A. thaliana DGAT1, but not DGAT2, has been shown to interact weakly with PDCT and LPCAT while soybean castor DGAT2 showed strong and weak interactions with A. thaliana PDCT and LPCAT, respectively (Regmi et al., 2020). A. thaliana DGAT1 and PDAT1 have also been reported to physically associate (Lee and Seo, 2019; Regmi et al., 2020), but these same interactions have not been observed with enzymes from flax (Xu et al., 2019). The predicted protein–protein interaction structures of flax DGAT2 with LPCAT, and A. thaliana DGAT1 with LPCAT, indicated that the former had a higher confidence prediction than the latter (Figure 2), which might be due to the fact that it is easier to model membrane enzymes with simple transmembrane structure (e.g. DGAT2) than those with multiple transmembrane domains (e.g. DGAT1). Nevertheless, both predictions indicated that helices 8, 9 and 10 of LPCAT might form interaction interfaces with DGAT1 (Figure 2b) and DGAT2 (Figure 2a). Several specific interacting sites were identified in these predictions, which could be validated through biochemical means in the future.

Many enzymes involved in TAG synthesis and acyl editing, such as DGAT2, PDAT, LPCAT and PDCT, have been found to display unique substrate preferences; a phenomenon that is especially true for those enzymes derived from plants that accumulate unusual fatty acids (Demski et al., 2022; Hong et al., 2022; Lu et al., 2009; Pan et al., 2013, 2015; Xu et al., 2018). Indeed, the formation of complexes among different proteins has been suggested to facilitate the channelling of unusual fatty acids from membrane lipids into TAG in developing seeds. Multiple substrate pools of DAG and acyl‐CoA have been reported for TAG biosynthesis using isotopic tracing. For example, AtDGAT1 and AtPDAT use different PC‐derived DAG pools to form TAG (Regmi et al., 2020). These findings suggest that separate interactomes or metabolons are likely present for TAG biosynthesis, which involves highly selective enzymes and spatially distinct acyltransferase complexes that mediate metabolic channelling. Moreover, metabolons also seem to play a role in chloroplast glycerolipid biosynthesis, which is supported by the presence of possible substrate pools for phosphatidylglycerol and galactolipid biosynthesis in different plant species such as microalgae (reviewed by Xu, 2022). Correspondingly, a dynamic protein interactome for lipid channelling has also been proposed in mammals, whereby acyl‐CoA synthetase and GPAT interact and are members of larger protein interactomes (Coleman, 2019).

Other protein interactomes involved in lipid‐related pathways

Other protein–protein interactions or lipid biosynthetic protein complexes are also known to play important roles in glycerolipid biosynthesis (Table 1). One example is the TRIGALACTOSYLDIACYLGLYCEROL (TGD) complex, which is an ABC transporter that facilitates lipid transport from the ER to chloroplasts. TGD complexes consist of TGD1, 2 and 3 in the inner plastidial envelope membrane, as well as TGD4 and 5 in the outer envelope membrane (Awai et al., 2006; Fan et al., 2015; Lu and Benning, 2009; Roston et al., 2011, 2012; Xu et al., 2005, 2008). TGD1, 2 and 3 resemble ATP binding cassette transporters, which bear permease, substrate binding and ATPase domains (Roston et al., 2011, 2012). These proteins can form a stable complex in the inner envelope membrane, and this complex appears to contain 4–6 times more copies of TGD2 than TGD1/3. However, the precise structural organization of the complex remains unknown (Roston et al., 2012). TGD4, on the other hand, is a homodimer in the outer envelope membrane, while TGD5 has been suggested to bridge TGD4 and TGD1‐3 complexes via protein–protein interactions (Fan et al., 2015). This putative TGD1‐5 complex is known to facilitate the transfer of ER lipid precursors from the ER to chloroplasts but, the lipid species being transported have yet to be identified despite the fact that TGD2 has been found to bind to PA (Awai et al., 2006).

Recently, the plastid rhomboid‐like protein RBL10 has also been found to form interactomes in the inner chloroplast envelope membrane with many chloroplast proteins, including ACP4 and CTI1 (Lavell et al., 2021). RBL10 is likely involved in chloroplast lipid metabolism since its loss of function dramatically affects monogalactosyldiacylglycerol biosynthesis in A. thaliana (Lavell et al., 2019, 2021). However, the precise structural characteristics of RBL10 interactions with ACP4 or CTI1, and the relevance of these interactions in terms of lipid metabolism remain largely unknown and are currently under further investigation.

Fatty acid desaturases (FADs) have long been known to form interactions among the same group of proteins (Lou et al., 2014; Shanklin and Cahoon, 1998). For example, plastidial soluble SADs, as well as plastidial membrane‐localized ω6 (FAD6) and ω3 FAD (FAD7 and 8), can form homodimers (Lindqvist et al., 1996; Lou et al., 2014; McKeon and Stumpf, 1982). Similarly, the ER‐localized ω6 (FAD2) and ω3 FAD (FAD3) enzymes also function as homodimers (Lou et al., 2014). Above and beyond the formation of homodimers, FAD6 can also associate with FAD7 or FAD8, and FAD2 can associate with FAD3, which likely facilitates the sequential desaturation of ω9 monounsaturated fatty acids (e.g. 16:1 and 18:1) to form trienoic fatty acids (e.g. 16:3 and 18:3) without releasing intermediates (Lou et al., 2014). Furthermore, FAD enzymes also interact with electron‐donating cofactors such as plastid ferredoxin and ER cytochrome b5, which provide electrons from the reductants NADPH or NADH to the diiron centre of desaturase active sites to enable catalysis (Horn et al., 2020; Kazaz et al., 2022; Schultz et al., 2000; Shanklin and Cahoon, 1998; Sobrado et al., 2006; Yang et al., 2015). For instance, the plastidial SAD from castor has been shown to interact with ferredoxin from Anabaena (Sobrado et al., 2006), while the plastidial Δ4FAD and FAD6 from C. reinhardtii physically interact with ferredoxin (Yang et al., 2015). Recently, A. thaliana FAD4, which converts 16:0 to 16:1 ^∆3trans in chloroplasts, was also found to require a thylakoid‐associated redox protein, PEROXIREDOXIN Q, and these may interact based on docking results (Horn et al., 2020).

Furthermore, acyl‐CoA binding proteins (ACBPs), which are a group of proteins that bind acyl‐CoAs and contribute to acyl‐CoA pool maintenance and trafficking, have also been found to be involved in protein–protein interactions. For example, A. thaliana ACBP2 was found to bind to lysophosphatidylcholine, which enhances the interaction of this enzyme with lysophospholipase and facilitates its ability to degrade lysophosphatidylcholine in response to cadmium‐induced oxidative stress (Gao et al., 2010; Miao et al., 2019).

In addition, protein–protein and protein–lipid interactions at membrane contacts that connect the ER with chloroplasts, the plasma membrane, mitochondria, or lipid droplets are believed to play a key role in lipid trafficking across different compartments of the cell. In A. thaliana, several proteins have recently been found to be crucial for tethering the ER to the plasma membrane, including VAMP/synaptobrevin‐associated proteins (VAPs, e.g. VAP27‐1, VAP27‐3, VAP27‐4) and synaptotagmins (SYTs) (Ishikawa et al., 2020; Siao et al., 2016; Stefano et al., 2018). VAPs are well known for their ER‐organelle tethering function and VAP27‐1 has recently been found to interact with SEIPIN at membrane contact sites between the ER and lipid droplets (Greer et al., 2020). Furthermore, lipid droplets have been found to anchor to the plasma membrane in A. thaliana via seed lipid droplet protein (SLDP) 1 and SLDP2, and lipid droplet plasma membrane adaptor (LIPA) (Krawczyk et al., 2022). Protein–protein interactions among lipid droplet proteins are also crucial for the packaging, storage and turnover of TAG (for a recent review, please see Guzha et al., (2022)). Similarly, a mitochondrial transmembrane lipoprotein complex has been identified, which plays a key role in mitochondrial lipid trafficking, most likely by regulating tethering between the outer and inner membranes of mitochondria (Michaud et al., 2016).

Engineering lipid biosynthesis by modulating interactomes

Plant species accumulate various fatty acids in their seeds, and enzymes such as FADs, FATs and FAEs are determinants of double bond formation and chain length in different fatty acids. These enzymes have been overexpressed in A. thaliana and common oilseed crops to produce different fatty acids, but only moderate accumulation of target fatty acids has been achieved in transgenic plants to date (Mietkiewska et al., 2014; Singh et al., 2001; Suh et al., 2002; Xu, Mietkiewska, et al., 2020). This is likely because heterologous systems lack the specific associated factors for lipid biosynthesis, such as interacting partners. To address this, the knowledge obtained from recent findings in lipid interactomes has, therefore, been exploited to inform a better design of lipid biosynthetic pathways for oil production. Thus far, some pioneering attempts have been devoted to engineer lipid biosynthesis by manipulating protein interactomes via the co‐expression of interacting partners, down‐regulation of negative or competing interactors, modulation of enzyme oligomer states and subunit composition of interactomes, protein fusion of interacting partners, and protein interfaces design to improve interaction, for instance.

Since many lipid metabolism‐related enzymes, such as DGAT1 and FAD, can form homo‐ or hetero‐oligomers, it is possible that introduced heterologous enzymes may form oligomers or interact with endogenous enzymes, which leads to reduced catalytic efficiency. For example, Crepis palaestina contains approximately 60% vernolic acid (12,13‐epoxy‐cis‐9‐octadecenoic acid). Fatty acid Δ11‐epoxygenase (Cpal2), which is a divergent FAD2 family enzyme, is responsible for the conversion of linoleic acid (18:2) to vernolic acid (Lee et al., 1998). The heterologous expression of Cpal2 in A. thaliana led to only low amounts of epoxy fatty acids, but substantial increases in oleic acid (18:1) and decreases in linoleic (18:2) and α‐linolenic acid (18:3). In contrast, the co‐expression of Cpal2 and FAD2 from C. palaestina restored the C18 fatty acid profile and increased the level of total epoxy fatty acids, indicating that the increase of oleic acid and decrease of polyunsaturated C18 fatty acids observed in CpCpal2 overexpressing A. thaliana was likely due to the inhibition of endogenous FAD2 by Cpal2 (Singh et al., 2001). Similar results were also reported in the production of conjugated fatty acids in transgenic A. thaliana. Pomegranate (Punica granatum) seed oil contains approximately 70% punicic acid (18:3 ^{cis9, trans11, cis13}) and a fatty acid conjugase (FADX), which is also a FAD2 family enzyme, is responsible for the conversion of 18:2 to punicic acid (Hornung et al., 2002; Iwabuchi et al., 2003). The seed‐specific expression of P. granatum FADX in an A. thaliana fad3/fae1 mutant, which exhibits an increase in 18:2 of up to 51% in its seed oil due to the inhibition of FAD3 and FAE1 activity, produced an average of 6.43% punicic acid (of total fatty acids) with a 33% decrease in 18:2 and a 34% increase in 18:1. Conversely, the co‐expression of FADX and FAD2 from P. granatum led to the production of seed oil comprising 9.2% punicic acid (of total fatty acids) and an 18:1 content comparable to that observed in untransformed A. thaliana fad3/fae1 seeds (Mietkiewska et al., 2014). These results indicated that the co‐expression of fatty acid conjugase and FAD2 from the same species led to enhanced production of unusual fatty acids (Mietkiewska et al., 2014; Xu, Mietkiewska, et al., 2020). In line with this, the co‐expression of A. thaliana or soybean cytochrome b5 along with FAD3 from the same species led to higher 18:3 fatty acid production in yeast (Kumar et al., 2012).

However, such a phenomenon is not always the case. For example, the seed endosperm of Coriandrum sativum (coriander) synthesizes the unusual monoenoic acid 18:1^Δ6 through the activity of Δ4 16:0‐ACP desaturase, and the glandular trichomes of Pelargonium hortorum (garden geranium) produce 16:1^Δ11, 18:1^Δ13 and derived products through the activity of Δ9 14:0‐ACP desaturase (Cahoon et al., 1992; Schultz et al., 1996). In line with this, the expression of coriander Δ4 16:0‐ACP desaturase and geranium Δ9 14:0‐ACP desaturase in A. thaliana resulted in the accumulation of a low amount of the corresponding unusual monoenoic acids, respectively (Suh et al., 2002). However, neither the co‐expression of coriander Δ4 16:0‐ACP desaturase with its cofactor coriander ACP‐I, or the co‐expression of geranium Δ9 14:0‐ACP desaturase with its cofactor Anabaena sp. 7120 ferredoxin, further enhanced the production of the unusual fatty acids in A. thaliana (Suh et al., 2002), regardless of the fact that acyl‐ACP desaturase activity could be enhanced in vitro with these specific ACP and ferredoxin isoforms (Schultz et al., 2000; Suh et al., 1999). Similarly, Umbellularia californica (California Bay) and Cuphea hookeriana produce medium chain fatty acids (10:0 and 12:0 in U. californica and 8:0 and 10:0 in C. hookeriana, respectively) through the activity of medium chain‐specific FATBs. The introduction of the medium chain‐specific FATBs from both plants, respectively, did not increase medium chain fatty acid production in C. reinhardtii, which was suggested to be due to the low affinity of heterologous FATBs with the endogenous ACP (Blatti et al., 2012). However, the subsequent co‐expression of a medium chain‐specific FATB and ACP from 10:0‐producing Cuphea lanceolata did not significantly increase medium chain fatty acid yield, which indicated that other factors were required (Inaba et al., 2017). In an alternative approach, structure‐guided protein engineering has also been used in an attempt to match protein interfaces of ACP and thioesterase to improve medium chain fatty acid production, whereby FATB activity was enhanced by alternating the protein surface of FATB to better recognize endogenous E. coli ACP (Sarria et al., 2018).

Interacting protein partners that negatively affect lipid enzyme activity have been engineered to improve lipid production. For example, the down‐regulation/mutation of ACCase negative regulators CTI and BADC was shown to enhance ACCase activity and seed oil content by up to 32% and 22% (relative) in seeds and leaves of A. thaliana, respectively (Keereetaweep et al., 2018; Salie et al., 2016; Ye, Nikovics, et al., 2020).

Modulating the molecular ratio of subunits in an enzyme complex has been found to be effective in the engineering of oil biosynthesis. For example, the α‐CT and β‐CT subunits of ACCase form a heterotetramer, but the abundance of α‐CT is 3–10‐fold less than that of β‐CT in developing A. thaliana seeds based on quantitative proteomics (Ke et al., 2000; Wang, Garneau, et al., 2022). Increasing the concentration ratio of α‐CT by overexpression of the corresponding gene successfully increased ACCase activity and resulted in increased oil production up to 8–15% (relative) in A. thaliana and camelina seeds (Wang, Garneau, et al., 2022). Similarly, ratiometric adjustments in the concentrations of different subunits of E. coli FAS can effectively improve fatty acid yields and modify acyl chain length profiles (Ruppe et al., 2020). By using a kinetic model of E. coli FAS along with a reconstituted in vitro system, the concentration ratio of a promiscuous FAT (TesA) and a KAS (FabB or FabH) were found to be crucial in determining the fatty acid chain length profile in E. coli, and shifts in the titres of different fatty acids were achieved by modulating their ratio (Mains et al., 2022; Ruppe et al., 2020).

The fusion of interacting partners has also been shown to be effective in enhancing enzyme activity and lipid biosynthesis. Certain plant desaturases, such as the fatty acid Δ6 desaturase from borage (Borago officinalis), are naturally fused to their electron donor cytochrome b5 (Napier et al., 1997; Sperling and Heinz, 2001). In line with this, the artificial fusion of cytochrome P450 monooxygenase and its interacting partner cytochrome P450 reductase have been used to improve the production of many valuable compounds previously (Sadeghi and Gilardi, 2013). One example is the fusion of Starmerella bombicola cytochrome P450 monooxygenase with A. thaliana cytochrome P450 reductase, which led to enhanced hydroxy fatty acid production in yeast compared to the simple co‐expression of the encoding genes (Liu et al., 2019). Similarly, the fusion of DGAT with acyl‐CoA binding proteins has been shown to improve enzyme kinetics and oil production (Xu, Caldo, et al., 2020). Taken together, it is clear that this strategy shows promise in the design of novel genetic engineering approaches based on our knowledge of protein interactomes, and as such, the expansion of this knowledge is likely going to be critical for the further advancement of plant lipid biotechnology.

Closing comments

Plant lipid biosynthesis likely involves various protein–protein and protein–lipid interactomes or metabolons, but our current understanding of the nature of these interactomes/metabolons is very limited. As such, it would be of considerable interest to systematically explore the formation of interactomes in the context of lipid biosynthetic enzymes as a means of revealing their overall role in plant lipid biosynthesis. Approaches such as yeast two‐hybrid assays, qualitative and quantitative mass spectrometry (pulldown assays or organelle separation), proximity labelling (Mair et al., 2019), and a combination of deep learning‐based prediction programmes (e.g. AlphaPulldown (Yu et al., 2023)) and biochemical verification will likely be valuable in terms of addressing these challenges. It will also be necessary to explore the nature of interactomes to address unknowns concerning specific interacting regions, the strength of the interactions and the conditions required for transient interactions. Moreover, although current evidence demonstrates the potential for physical interactions among proteins, it remains unclear whether and how substrates are channelled in the interactomes. Molecular simulation/docking, as well as structural identification and characterization, would provide further structural insight into protein interactomes. These efforts will not only enable a better understanding of the mechanisms driving the formation of interactomes/metabolons and their roles in lipid biosynthesis, but will also facilitate future synthetic biology efforts to engineer lipid biosynthesis.

Conflict of interest

There are no conflicts of interest with the content of this article.