Divergent fatty acid desaturase 2 is essential for falcarindiol biosynthesis in carrot

Abstract

Carrots produce diverse falcarin-type polyacetylenes (PAs), which function as important antipathogenic phytoalexins and have potential as anticancer agents. Despite their abundance in carrot root tissues, the biosynthesis and evolutionary origins of falcarindiol, a C17-PA, remain unknown. Fatty acid desaturase 2 (FAD2) enzymes diversify PAs by introducing various double and/or triple carbon–carbon bonds into fatty acid chains. Here, we use association analysis to identify candidate FAD2 genes involved in falcarindiol biosynthesis. Using a rapid tobacco transient expression system, we found that DcFAD2 enzymes are highly functionally redundant and promiscuous. Combinatorial assays also revealed unexpected synergistic and redirective effects among FAD2 enzymes, further complicating the biosynthetic pathway. CRISPR–Cas9-mediated mutagenesis and overexpression studies identified previously overlooked DcFAD2 hub genes as essential for falcarindiol production. Evolutionary analysis suggests that the expansion of DcFAD2 genes underpins the abundance of falcarindiol in carrot, independent of the biosynthetic gene cluster previously identified in tomato. This work underscores the complexity of the falcarin biosynthetic network and identifies hub genes essential for falcarindiol biosynthesis in carrot.

This study demonstrates that the biosynthesis of falcarindiol, a bioactive polyacetylene in carrots, is driven by promiscuous and functionally redundant DcFAD2 enzymes. It further reveals complex interactions among these enzymes, including synergistic and pathway-redirecting effects, and identifies key DcFAD2 genes that serve as metabolic hubs. Unlike the biosynthetic gene cluster responsible for falcarindiol production in tomato, carrot falcarindiol biosynthesis appears to have evolved independently through a lineage-specific expansion of DcFAD2 genes.

Introduction

Polyacetylenes (PAs) are a class of specialized metabolites characterized by multiple alkynyl groups in their carbon skeletons (Minto and Blacklock, 2008). These acetylenic lipids are widely distributed across the kingdoms of life, with a notably high abundance in higher plants. More than 2,000 plant PAs have been identified, especially in families such as Asteraceae, Araliaceae, and Apiaceae (Negri, 2015). PAs are structurally diverse, with variations in chain length and patterns of dehydrogenation, oxygenation, and other modifications (Negri, 2015; Lai et al., 2023). Typically, PAs from land plants have carbon chains ranging from 10 to 18 atoms in length. The main PAs, mostly found in Apiaceae and Araliaceae, are aliphatic C17 and C18 acetylenic oxylipins, with the C17 falcarin-type PAs (falcarins) being the most common (Dawid et al., 2015; Santos et al., 2022; Lai et al., 2023). One of the best-characterized C17 falcarins, falcarindiol, is a diacetylene C17-PA typically found in the roots and rhizomes of Apiaceae and Araliaceae plants, such as carrot (Daucus carota L.) (Dawid et al., 2015), celery (Apium graveolens L.) (Anzera et al., 2005), Aegopodium podagraria L. (Wang et al., 2017), and ginseng (Panax ginseng) (Anzera et al., 2005). Falcarindiol has been shown to inhibit the growth of Botrytis cinerea in carrot (Harding and Heale, 1980) and is induced by fungal elicitors in tomato (Solanum lycopersicum) leaves (Jeon et al., 2020). It is exceptionally abundant in carrot (6–60 mg/kg fresh weight) and plays a central role in the suppression of various cancers (Kreutzmann et al., 2008; Pferschy-Wenzig et al., 2009; Thomas et al., 2023). For instance, several studies have indicated that falcarindiol effectively eradicates cancer cells by inducing apoptosis via endoplasmic reticulum stress (Jin et al., 2012; Lu et al., 2017). It has been proposed that the triple bonds in falcarins convert oxylipins into highly alkylating compounds that are reactive toward cancer resistance proteins and other biomolecules (Christensen, 2020). Falcarindiol also contains unsaturated double bonds, with the terminal double bond being particularly important for its cytotoxicity (Christensen and Brandt, 2006; Christensen, 2011). Despite its importance as a chemical defense against pathogens in carrot and its potential in cancer treatment, the biosynthesis of falcarindiol remains largely unknown.

Falcarin biosynthesis is suggested to begin with primary unsaturated fatty acids (i.e., oleic acid and linoleic acid), followed by carbon skeleton modifications primarily through fatty acid desaturase 2 (FAD2) (Busta et al., 2018) and other decorating enzymes (Jeon et al., 2020; Santos et al., 2022; Scott et al., 2022) (Figure 1A). Jeon et al. (2020) performed functional and genetic analyses in tomato leaves and identified a metabolic gene cluster activated by pathogen infection that encodes three FAD2 genes and a decarbonylase involved in falcarindiol biosynthesis. However, no synteny of this cluster was found in carrot (Jeon et al., 2020). In carrot, three Δ12-desaturases and Δ12-acetylenases have been identified in vitro; these enzymes collectively produce crepenynic acid (CA) (Figure 1A). Notably, 24 DcFAD2 genes have been identified in carrot, compared with only one FAD2 gene in the model plant Arabidopsis thaliana (Nguyen et al., 2019), indicating that much of the FAD2 repertoire remains unexplored (Busta et al., 2018). How carrots have repurposed this expanded set of FAD2 genes for falcarin biosynthesis is poorly understood.

Figure 1.

Proposed falcarindiol biosynthetic pathway and falcarindiol quantification in representative carrot tissues.

(A) Proposed steps for falcarindiol biosynthesis from the primary metabolite OA (Busta et al., 2018; Santos et al., 2022). The DcFAD2 enzymes shown have been previously characterized by in vitro assays.

(B) LC–CAD–MS analysis of falcarindiol in carrot tissues at two developmental stages. CAD, charged aerosol detection. EIC, extracted ion chromatogram. m/z values for falcarindiol were measured in positive ion mode ([M–H₂O + H]⁺).

(C) Quantification of falcarindiol levels (see Supplemental Figure 1 for details). Values represent means ± standard error of the mean from biological replicates (n = 3).

With the advent of a high-quality carrot genome assembly and the development of various heterologous systems, it has become feasible to investigate the functional diversity of FAD2s and to elucidate complex biosynthetic pathways through both in vitro and in vivo assays (Iorizzo et al., 2016; Stephenson et al., 2018; Jeon et al., 2020). Busta et al. (2018) characterized three carrot FAD2 enzymes (DcFAD2-19, DcFAD2-11, and DcFAD2-12) capable of converting oleic acid (OA) to linoleic acid (LA) in yeast. However, the functional verification of FAD2 Δ12-acetylenases, which convert LA to CA, proved challenging in yeast, and instead, they were functionally tested via ectopic expression in A. thaliana (Cahoon et al., 2003; Busta et al., 2018). The Nicotiana benthamiana transient expression system is well-suited for investigating the activity of FAD2 enzymes during the early steps of the falcarindiol biosynthetic pathway, due to the abundance of the precursor LA. Cao et al. (2013) used this system to verify the functions of several FAD2 family members in safflower. Similarly, Jeon et al. (2020) used it to characterize multiple FAD2 enzymes involved in falcarindiol biosynthesis.

In the present study, we performed RNA sequencing (RNA-seq) and metabolite profiling across different carrot tissues (leaf and periderm) and growth stages (young and mature). By integrating these datasets, we conducted metabolite–transcript association analyses to identify candidate FAD2 genes that encode falcarindiol biosynthetic enzymes in carrot. Using the hypertranslational transient expression system in tobacco leaves (Reed et al., 2017; Stephenson et al., 2018), we identified functionally redundant and synergistic FAD2 enzymes associated with CA and dehydrocrepenynic acid (DCA) production in carrot. Key FAD2 genes essential for falcarindiol production were validated through both gain- and loss-of-function studies. Our results revealed that the high capacity for falcarindiol biosynthesis in carrot is partly governed by quantitative trait loci (QTLs) containing tandem arrays of FAD2 genes. These arrays likely arose through extensive tandem duplication and whole-genome duplication (WGD), with some members potentially evolving novel functions. These findings provide a foundation for breeding improved carrot cultivars and for engineering high-value falcarindiol compounds through synthetic biology.

Results

Identification of DcFAD2 genes responsible for falcarindiol biosynthesis in carrot

Root periderm has been reported as the primary tissue for falcarindiol accumulation (Busta et al., 2018). To characterize falcarindiol distribution more accurately for candidate gene mining, we selected two carrot tissues (root periderm and aerial leaf) and two developmental stages for simultaneous transcriptomic and targeted metabolite analyses. Homogenized tissues from 30-day-old (young) and 120-day-old (mature) cultivated carrot plants were analyzed using liquid chromatography–mass spectrometry (LC–MS) with charged aerosol detection (CAD). This analysis showed that young root periderm had the highest falcarindiol levels, reaching approximately 1.5 μg/mg dry weight, whereas leaf tissues at both stages had the lowest levels (∼0.03 μg/mg dry weight) (Figure 1B and 1C and Supplemental Figure 1). Transcriptomic analysis was performed on the same batch of tissues used for metabolite analysis. Given the 24 phylogenetically identified FAD2 genes in carrot (Busta et al., 2018; Feng et al., 2019) (Figure 2A; Supplemental Table 1), we examined correlations between their expression levels and falcarindiol accumulation. Twelve FAD2 genes with expression patterns correlated with falcarindiol levels are shown in Figure 2B. Of these, five have been previously reported as FAD2 Δ12-desaturases and/or Δ12-acetylenases, potentially involved in falcarindiol biosynthesis in carrot (Busta et al., 2018). The DcFAD2 genes are predominantly found in tandem gene arrays (Figure 2C) (Busta et al., 2018; Santos et al., 2022).

Figure 2.

Identification and analysis of candidate FAD2 genes involved in falcarindiol biosynthesis in carrot.

(A) Maximum-likelihood phylogenetic tree of 24 DcFAD2 proteins along with functionally characterized FAD2 enzymes, including desaturases, acetylenases (ACET), conjugases (FADX and FAC), hydroxylases (HYDX), and epoxygenases (EPOX).

(B) Expression patterns of DcFAD2 genes potentially involved in falcarindiol biosynthesis. The heatmap shows normalized transcript abundance (transcripts per million). Metabolite–gene expression correlations (Pearson correlation coefficients) are shown.

(C) Genomic distribution of the 24 DcFAD2 genes. Major FAD2 gene arrays are labeled. Collinearity between gene arrays, based on syntenic block analysis, is indicated by thick red lines. Genes derived from whole-genome duplication (WGD) events are marked in red (Supplemental Figure 2). QTLs previously associated with falcarindiol biosynthesis are highlighted (Dunemann et al., 2022). Black and blue circles indicate characterized Δ12-desaturases and Δ12-acetylenases, respectively (Busta et al., 2018).

To further investigate the genomic basis of FAD2 gene expansion in carrot, we analyzed the distribution of synonymous substitutions per site (Ks) within the carrot genome. The Ks plots suggested that two WGD events (Ks peaks at 0.5 and 1.5) occurred during carrot evolution (Supplemental Figure 2A). By projecting the Ks values of duplicated gene pairs onto the genome-wide Ks distribution, we found that the expanded FAD2 gene arrays were most likely generated through multiple WGD and tandem duplications (Figure 2C and Supplemental Figure 2B). Syntenic analysis also indicated that the three paired FAD2 arrays likely arose from WGD (Figure 2C). Functional evidence also supports these findings. For instance, DcFAD2-7 and DcFAD2-6, located in arrays 1.1 and 1.2 respectively, have been reported to have conserved catalytic functions (Figure 1A) (Busta et al., 2018). Notably, the number of FAD2 genes in each array varies considerably (Figure 2C), which suggests additional gene gain and/or loss events following the WGD. To date, all functionally characterized FAD2 genes belong to array 1 or 2 (Figure 2C). Despite a recent QTL mapping study implicating all three paired FAD2 arrays in falcarindiol biosynthesis (Figure 2C) (Dunemann et al., 2022), no genes from array 3 have been functionally characterized.

Identification of new Δ12-acetylenases and Δ14-desaturases required for falcarindiol production in carrot

To verify gene functions, we successfully cloned 10 of the 12 highly correlated DcFAD2 genes. Two additional FAD2 genes from array 3 (DcFAD2-9 and DcFAD2-21) that are highly correlated with DcFAD2-1 were also cloned. These 12 DcFAD2 genes were then expressed individually or in combination in the N. benthamiana transient expression system (Methods). Fatty acid methyl esters (FAMEs) were extracted from the tobacco leaves and analyzed using gas chromatography–mass spectrometry (GC–MS). Leaves expressing DcFAD2-6, DcFAD2-7, or DcFAD2-8 (encoded by array 1 genes), which were previously characterized as Δ12-acetylenases (Busta et al., 2018), showed increased production of CA (verified with an authentic standard) compared with the control (Figure 3A and Supplemental Figure 3A). Interestingly, two genes from array 3 (DcFAD2-1 and DcFAD2-9) were also found to encode Δ12-acetylenases (Figure 3A and Supplemental Figure 3A and 3B). Notably, when all 12 DcFAD2 genes were co-expressed in tobacco, CA was barely detected (Figure 3A); this suggests that one or more of the 12 DcFAD2s may participate in further modification of CA. Indeed, individual overexpression of DcFAD2-1, -6, -7, and -8 in tobacco leaves led to the production of desaturated CA, with ion spectra matching the previously reported DCA (Cahoon et al., 2003; Busta et al., 2018) (Figure 3A and Supplemental Figure 3C). Although these enzymes exhibited dual functionality, producing both CA (the major product) and DCA, this does not account for the disappearance of CA when all 12 DcFAD2 genes were co-expressed.

Figure 3.

Identification of novel Δ12-acetylenases and Δ14-desaturases in carrot.

(A) GC–MS analysis of fatty acid methyl esters (FAMEs) from N. benthamiana leaves expressing mNG (negative control) and individual or combined DcFAD2 genes. Aligned EICs or total ion chromatograms are shown along with the authentic CA standard.

(B) In the absence of an authentic dehydrocrepenynic acid (DCA) standard, previously reported DcFAD2 combinations were used for DCA identification (Supplemental Figure 3).

(C) A novel peak at m/z 292 corresponds to an unidentified polyunsaturated fatty acid (PUFA). The corresponding ion was extracted to enable comparisons across samples.

(D) Quantification of CA, DCA, and PUFA in N. benthamiana leaves transiently expressing DcFAD2 genes individually or in combination. Values represent means ± standard error (n = 6). CA was quantified using an external standard via GC–flame ionization detector (FID) (see Supplemental Figure 3F for details). Relative abundances of DCA and PUFA were assessed using total ion chromatogram peak areas.

Previously, co-expression of DcFAD2-7 and DcFAD2-19 (encoded by array 1 genes) in A. thaliana was shown to produce Δ14-DCA (Busta et al., 2018). We therefore tested whether the combination of DcFAD2-7 and DcFAD2-19 in tobacco could similarly channel metabolic flux toward DCA production. Given that DcFAD2-6, -7, and -8 belong to array 1 and are phylogenetically similar to parsley ELI12 (an enzyme that catalyzes both CA and DCA production) (Figure 2A) (Cahoon et al., 2003; Blacklock et al., 2010), we tested transient co-expression of DcFAD2-7 (or DcFAD2-6) and DcFAD2-19 in tobacco leaves. The results showed a drastic decrease in CA with a concomitant accumulation of DCA (Figure 3B and Supplemental Figure 3D). Expression of DcFAD2-19 alone caused no accumulation of either CA or DCA (Figure 3B). These results indicate that DcFAD2-7 alone primarily produces CA, but when co-expressed with DcFAD2-19, it leads to the production of DCA. Synergistic effects were also observed when DcFAD2-7 was co-expressed with DcFAD2-12 and -18 (Supplemental Figure 3D). In addition, when DcFAD2-7 was replaced by DcFAD2-1 or DcFAD2-6 (from the same phylogenetic clade) (Figure 2A and Supplemental Figure 3D), DCA was also the primary product (Figure 3B and Supplemental Figure 3D).

Intriguingly, co-expression of DcFAD2-1, -6, or -7 with either DcFAD2-17 or -13 resulted in the production of a structurally unknown polyunsaturated fatty acid (PUFA) (Figure 3C and Supplemental Figure 3D and 3E). Expression of DcFAD2-17 or -13 alone did not produce detectable levels of CA, DCA, or PUFA (Figure 3C). These results suggest that DcFAD2-1, -6, and -7 likely have dual functions, with different combinations leading to either DCA or PUFA. These findings were further supported by quantitative analysis (Figure 3D). Overall, using the rapid tobacco expression and PA detection system, we identified two new Δ12-acetylenases and six new Δ14-desaturases involved in CA and DCA production, respectively (Supplemental Figure 3B and 3C). Our results further revealed the complexity of falcarindiol precursor biosynthesis and the combinatorial effects among DcFAD2s in synergistically producing falcarindiol or redirecting metabolic flux into other falcarin branches in carrot.

Genetic evidence supporting the roles of FAD2 in carrot falcarindiol biosynthesis

In tomato, falcarindiol has been reported to originate from CA, with the encoding genes forming a biosynthetic gene cluster (BGC) (Jeon et al., 2020). However, evolutionary analysis indicates that this gene cluster is restricted to tomato, and no syntenic blocks could be identified in carrot (Jeon et al., 2020). Although our association and functional analyses identified carrot FAD2 enzymes responsible for both CA and DCA production, genetic evidence remains absent. In addition, mutagenesis that targets the CA and DCA biosynthetic steps is challenging due to the functional redundancy of the FAD2 genes in carrot (Figure 3). For this reason, we focused on the array 3 FAD2 genes, because genes in this QTL had not been functionally characterized, and because DcFAD2-1 exhibited the highest catalytic activity in CA and DCA production in N. benthamiana (Figures 2C and 3D). CRISPR–Cas9-mediated gene knockout experiments were conducted in carrot hairy roots. Two mutant lines with loss of function in DcFAD2-1 and the closely related DcFAD2-9 were obtained (Figure 2C and Supplemental Figure 4A and 4B). Notably, DcFAD2-1 and DcFAD2-9 also exhibit similar catalytic activity (Figure 3A and Supplemental Figure 3A). Compared with control 1 (hairy roots harboring a CRISPR–Cas9 empty vector), the two mutant lines showed significant reductions in falcarindiol content, with one line exhibiting an 80% decrease (Figure 4A and 4B).

Figure 4.

Identification of DcFAD2 genes essential for falcarindiol biosynthesis through knockout and overexpression analysis in carrot hairy roots.

(A) LC–MS analysis of falcarindiol content. Aligned EICs are shown for mutant and overexpression lines along with the authentic falcarindiol standard. m/z values are indicated for falcarindiol in positive ion mode (M-H₂O or M-2H₂O).

(B) Quantification of falcarindiol in mutant lines compared with the genetically unmodified control.

(C) Relative expression levels of DcFAD2-1 in overexpression lines compared with a control transformed with an empty vector.

(D) Quantification of falcarindiol in overexpression lines compared with the control.

(E) Comparison of GC–MS chromatograms between mutant and overexpression lines and their respective controls. Tobacco samples that accumulated PUFAs (via combinatorial expression of DcFAD2-1 and DcFAD2-13) and DCA (via combinatorial expression of DcFAD2-6 and DcFAD2-19) were used as positive controls.

(F and G) Relative abundance of PUFAs (F) and DCA (G) in overexpressed hairy root lines compared with the control. Control 1 represents hairy roots harboring an empty CRISPR–Cas9 vector, whereas control 2 represents hairy roots harboring an empty overexpression vector with a DsRed selection marker. Asterisks indicate significant differences from wild-type samples, as determined by a two-tailed Student’s t-test (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001). Values represent means ± SD (n = 6).

To further confirm the role of DcFAD2-1 in falcarindiol biosynthesis, we overexpressed DcFAD2-1 in carrot hairy roots, generating three independent lines (Figure 4C; Supplemental Figure 4C). The overexpression lines exhibited significantly higher falcarindiol content than the mutant lines or control 2 (hairy roots harboring an empty overexpression vector containing a DsRed selection marker) (Figure 4A and 4D). As DcFAD2-1 also showed enzymatic promiscuity toward structurally unknown PUFAs in transient tobacco assays (Figure 3C and Supplemental Figure 3D and 3E), we further investigated this effect in the CRISPR–Cas9-generated mutant lines. Although PUFAs were not detected in normally grown carrot tissues during our profiling analysis (Supplemental Figure 4D), both PUFAs and DCA accumulated to detectable levels in hairy roots generated via Agrobacterium rhizogenes transformation, as shown by GC–MS analysis (Figure 4E). These compounds were reduced to nearly undetectable levels in the two mutant lines (Figure 4E). Consistently, overexpression of DcFAD2-1 in hairy roots resulted in significantly higher levels of both DCA and PUFAs (Figure 4E–4G). These results align with those from the tobacco in vitro assays (Figure 3) and suggest that array 3 contains hub genes central to falcarindiol biosynthesis and the production of currently undiscovered unusual fatty acids in carrot roots.

Discussion

Carrots are among the most important vegetable crops in the Apiaceae family, and their global cultivation area is rapidly expanding (Iorizzo et al., 2016; Que et al., 2019). In addition to being rich in provitamin A and various bioactive compounds (Guo and Lu, 2023), carrots contain high levels of falcarin-type PAs, which function as important antipathogenic phytoalexins and have potential as anticancer agents (Thomas et al., 2023). Falcarindiol accumulates at exceptionally high levels in carrot roots, whereas in tomato, it only accumulates in leaves upon pathogen infection. Recently, the understanding of falcarindiol biosynthesis in tomato has advanced greatly due to the identification of underlying genes, including FAD2s and a decarbonylase, which form a BGC (Jeon et al., 2020). However, the complete biosynthetic pathway remains largely unknown, including the oxygenation and decarboxylation steps. Here, we delineate the biosynthesis of falcarindiol and the genomic basis for its high levels in carrot. Through both in vitro and in vivo assays, we demonstrated that carrot falcarindiol is synthesized via the crepenynate pathway, similar to its synthesis in tomato. Although falcarindiol in tomato is mostly inducible in leaves (Lee et al., 2020), in carrot it is constitutively present in roots, even under sterile conditions. Tomato falcarindiol is largely encoded by a BGC; in carrot, FAD2 genes are extensively expanded (compared with only nine FAD2 genes in tomato) and form various gene arrays primarily resulting from a combination of WGD and tandem gene duplications (Figure 2C). No syntenic decarbonylase genes were found in the carrot gene arrays, suggesting that tomato and carrot have convergently evolved their genomic basis for falcarindiol accumulation (Figure 5).

Figure 5.

Independent evolution of falcarindiol biosynthesis in carrot and tomato.

A schematic overview of falcarindiol biosynthesis in tomato (Jeon et al., 2020) and carrot (this study) is displayed. Newly identified DcFAD2 enzymes are highlighted in bold. Unlike the BGC responsible for falcarindiol in tomato, falcarindiol biosynthesis in carrot evolved independently through lineage-specific expansion of DcFAD2 genes. OA, oleic acid; LA, linoleic acid; CA, crepenynic acid; DCA, dehydrocrepenynic acid. Tomato FAD2 homologs Solyc12g100240, Solyc12g100250, Solyc12g100260, and Solyc12g100270 are abbreviated as Sl240, Sl250, Sl260, and Sl270, respectively. The schematic was created using BioRender.

Gene and genome duplications are fundamental to plant biology and have provided the raw materials for metabolic diversification (Zhou and Liu, 2022). One underestimated effect of gene duplication is its contribution to metabolite accumulation. WGD is crucial for the generation of single-gene duplications and significantly influences plant genome function and evolution. Examples include high levels of glucosinolates in Brassicales, OA in wild olive, and triterpene acids in loquat leaves, all linked to gene and/or genome duplications (Liu et al., 2014; Unver et al., 2017; Su et al., 2021). The accumulation of falcarindiol in carrot roots likely reflects the persistent presence of pathogens in the rhizosphere. Notably, expression of most DcFAD2 genes is upregulated following pathogen challenge (Busta et al., 2018), which provides further evidence for the defensive role of falcarins in carrot. Accordingly, the expansion of carrot FAD2 genes aligns with the species’ remarkable capacity for falcarindiol production. In addition, functionally redundant FAD2 genes have been identified for the enzymatic steps that produce CA and DCA, which are essential precursors for falcarindiol biosynthesis. The expansion and functional redundancy of FAD2 genes thus likely underlie the elevated falcarindiol levels in carrot (Figure 5).

In yeast, previous studies have identified putative carrot Δ12-acetylenases and Δ14-desaturases involved in falcarindiol biosynthesis (Busta et al., 2018). In this study, we used a transient tobacco expression system, enabling rapid functional validation and combinatorial assays to identify synergistic and pathway-redirecting effects of FAD2 genes, which likely coordinate a falcarin biosynthetic network that includes the biosynthesis of falcarindiol and other PUFAs (Figure 5). Co-expression of array 3 members (e.g., DcFAD2-1) with other DcFAD2 genes resulted in the accumulation of either DCA or PUFAs (Figure 3B and 3C and Supplemental Figure 3D), suggesting that DcFAD2-1 may interact with different FAD2 partners in vivo to direct the pathway into distinct falcarindiol branches. This hypothesis is further supported by the observed localization of FAD2 proteins to the endoplasmic reticulum (ER) (Supplemental Figure 5), which provides anchoring points for several known FAD protein–protein interactions (Dastmalchi et al., 2018; Xu et al., 2023). Furthermore, CRISPR–Cas9-mediated knockout assays revealed the hub-like functions of array 3.1 FAD2 genes (Figure 5).

Overall, we identified new FAD2 genes that are functionally redundant in catalyzing falcarindiol precursor biosynthesis and used rapid transient assays to reveal the complexity and diversity of falcarin biosynthesis in carrot. This study also provides a compelling demonstration of the diverse gene duplication mechanisms underlying metabolite accumulation in plants. The hub genes identified in this study will not only support future carrot breeding efforts but also serve as synthetic tools for the production of high-value falcarins.

Methods

Plant materials and growth conditions

Carrots (D. carota L.) were sown in pots and grown under greenhouse conditions (16 h light/8 h dark, 25°C ± 2°C, 40%–60% humidity). Plants were harvested at different stages (young and mature), and periderm and leaf tissues were sampled separately. A portion of each sample was flash-frozen, ground in liquid nitrogen, and used for RNA-Seq analysis, whereas the remaining portion was lyophilized, ground, and used for metabolite detection.

For carrot hairy root culture, 1.0 g of hairy roots was inoculated into flasks containing 100 ml of B5 liquid medium supplemented with 200 mg/L cefotaxime and incubated at 25°C with shaking at 90 rpm. After 1 month, the hairy roots were blotted, dried, and then freeze-dried using a Labconco Freezone 6PLUS. The dried roots were weighed and prepared for falcarindiol extraction. Three biological replicates were analyzed for all overexpression and mutant lines. For the transient expression experiments, 4- to 6-week-old tobacco (N. benthamiana) plants grown in a growth chamber (16 h light/8 h dark, 25°C ± 2°C, 40%–60% humidity) were used.

Chemicals

Authentic standards of falcarindiol (CAS Registry Number 225110-25-8) and CA (CAS Registry Number 2277-31-8) were purchased from Bide Pharmatech and Tansoole, respectively.

RNA-seq analysis

Total RNA was extracted from various carrot tissues using QIAzol lysis reagent (QIAGEN, Germany). RNA concentration and quality were assessed using an ND-2000 spectrophotometer (Thermo Fisher Scientific, USA). Library construction and sequencing were performed at Shanghai Majorbio Bio-Pharm Biotechnology using the Illumina platform (Illumina, CA, USA) according to the manufacturer’s instructions. Raw paired-end reads were trimmed and checked for quality using Trimmomatic 0.39 with default parameters (Bolger et al., 2014). Cleaned reads were aligned to the reference genome using HISAT2 software in directional mode (Kim et al., 2015). For each sample, mapped reads were assembled into transcripts using StringTie (Id et al., 2022) with a reference-based approach.

Phylogenetic and correlation analysis

FAD2 homologs were identified using HMMER 3.1b2 (Eddy, 2011). PF00487 (FA_desaturase), PF03405 (FA_desaturase_2), and PF10520 (Lipid_desat) were used as queries to search for FAD2 genes in Phytozome v.13 (Hajiahmadi et al., 2020). Candidate FAD2 proteins were screened based on a length range of 200–600 amino acids. Protein alignment was performed using MUSCLE v.3.8.31 (Edgar, 2004). A phylogenetic tree was constructed using FastTree (Price et al., 2009), with the reliability of the tree’s topology assessed using bootstrap values based on 1000 replications.

We used the rcorr function in R with the parameter type = “pearson” to calculate Pearson correlation coefficients. This analysis assessed the relationship between expression levels of DcFAD2 genes, quantified as transcripts per million values, and falcarindiol content, measured via LC–MS, across various tissues.

Collinearity analysis and Ks analysis

The carrot genome file was downloaded from Phytozome v.13 and filtered using the Genome Length Filter tool in TBtools (Chen et al., 2020). The GXF Gene Position & Info Extract tool was used to obtain gene location information across the genome. The genomic positions of 24 DcFAD2 genes were then extracted using the Table Row Extract tool or a filter. Synteny results were obtained through linear analysis using One Step MCScanX Super Fast, and a collinearity map was generated using Advanced Circos. Ks analysis was performed using WGDI to identify potential WGD events in carrot (Sun et al., 2022).

Extraction and detection of falcarindiol in carrot tissues and hairy roots

Freeze-dried carrot tissues or hairy roots (50 mg) were homogenized with a 3 mm tungsten bead and extracted with 1 ml of methanol under sonication for 30 min at room temperature. The extracts were centrifuged at 13 000 rpm for 15 min. The supernatant was transferred to vials, and 3 μl was used for injection and detection. Samples were analyzed by reverse-phase chromatography on a Thermo Fisher Scientific LC-ISQ-MS system using a 2.6 μm, 2.1 × 50 mm C18 HPLC column. The column temperature was set to 45°C. The mobile phase consisted of water with 0.1% formic acid (A) and acetonitrile with 0.1% formic acid (B), and was delivered at a flow rate of 0.4 ml/min with a 12-min gradient: 0–9 min, 5%–100% B; held at 100% B for 1 min; 10–10.1 min, 100%–5% B; then held at 5% B for 1.9 min. A Thermo Fisher Scientific Electrospray Ionization (ESI) mass spectrometer was used to collect MS data in positive ion mode under the following conditions: mass range 150–500 m/z, vaporizer temperature 350°C, ion transfer tube temperature 350°C, sheath gas pressure 60 pounds per square inch gauge (psig), auxiliary gas pressure 6 psig, and sweep gas pressure 0.2 psig.

Heterologous expression in N. benthamiana

The method for hypertranslational high-level transient protein expression in tobacco leaves was adapted with minor modifications from previous studies (Stephenson et al., 2018; Jeon et al., 2020). DcFAD2 genes were cloned into the pEAQ-xhoI vector with mNeonGreen (mNG, encoding a monomeric yellow-green fluorescent protein) included as a negative control. The constructs were transformed into A. tumefaciens (strain LBA4404), followed by selection on LB plates containing 50 μg/ml kanamycin, 50 μg/ml streptomycin, and 20 μg/ml rifampicin at 28°C for 2–3 days. Primer sequences are listed in Supplemental Table 2. Single colonies were cultured overnight in lysogeny broth (LB) liquid medium containing the appropriate antibiotic at 28°C with shaking at 220 rpm. The A. tumefaciens cells were centrifuged at 4000 rpm for 15 min, and the supernatant was discarded. The bacterial pellet was then resuspended in 5 ml of induction medium (10 mM 2-morpholinoethanesulphonic acid (MES) [pH 5.6], 10 mM MgCl₂, and 150 μM acetosyringone) and incubated at room temperature for 2 h. A. tumefaciens suspensions (optical density 600 [OD₆₀₀] = 0.2 for each strain) were infiltrated into the underside of N. benthamiana leaves using a needleless 5 ml syringe. Leaves from different plants (n = 6) were harvested 3 days post-infiltration, lyophilized, ground uniformly, and subjected to lipid analysis.

Lipid and methyl ester (FAME) profiling

Lyophilized N. benthamiana leaves were ground to a fine powder using a mortar and pestle. Tissue powder (50 mg) was extracted using 1 mL of a methanol:chloroform mixture (2:1 [v/v]) in an ultrasonic bath (40 kHz) at room temperature for 30 min. The suspension was centrifuged at 13 000 rpm for 10 min, and the supernatant was evaporated to dryness using a centrifugal evaporator. Next, 500 μl of methanolic HCl (Supelco, PA, USA) was added for methyl esterification of fatty acids, and the mixture was heated at 60°C for 2 h. After cooling, 600 μl of 0.9% NaCl aqueous solution and 600 μl of hexane were added. The mixture was shaken vigorously and left to stand at room temperature for 10 min. The upper hexane layer was carefully transferred to a new 2 ml centrifuge tube. The organic phase was dried using a centrifugal evaporator, and 200 μl of hexane was added to the residue. The resulting samples were then prepared for GC–MS analysis. FAMEs were analyzed using a Thermo Fisher Scientific TRACE 1310 gas chromatograph coupled to an ISQ 9000 triple quadrupole mass spectrometer equipped with a CP-Sil 88 column (60 m × 0.25 mm, 0.2 μm). The column temperature was programmed as follows: initial temperature 120°C (held for 1 min), ramped to 165°C at 15°C/min, to 185°C at 5°C/min, to 190°C at 1°C/min, to 195°C at 0.5°C/min (held for 2 min), then to 200°C at 0.5°C/min, and finally to 220°C at 10°C/min (held for 3 min). A split ratio of 20:1 was used, with the inlet temperature set to 200°C and a nitrogen flow rate of 1.5 mL/min. The mass spectrometer operated in full scan mode over an m/z range of 50–500. Peak integration was performed using Thermo Fisher Scientific Chromeleon Console software.

CRISPR–Cas9-mediated mutagenesis and gene overexpression analysis in carrot hairy roots

The procedures for designing single-guide RNAs (sgRNAs) and constructing CRISPR–Cas9 plasmids are described in detail in Zhang et al. (2016) and Feng et al. (2018). In brief, sgRNAs were designed using CRISPR-GE (http://skl.scau.edu.cn/) and CRISPOR Tefor (http://crispor.gi.ucsc.edu/). The sequences, selected to achieve effective simultaneous knockout of DcFAD2-1 and DcFAD2-9, are provided in Supplemental Table 2. sgRNA production followed a defined protocol: the forward and reverse primers were denatured at 37°C for 30 min, incubated at 95°C for 5 min, and then cooled at a rate of 6°C per minute until the temperature reached 25°C. The double-stranded sgRNA1 and sgRNA2 were ligated into the AtU6 18-TM and At7SL 18-TM vectors, respectively, both of which had been predigested with BbsI, to generate AtU6-DcFAD2-1&2-9-sgRNA1 and At7SL-DcFAD2-1&2-9-sgRNA2, respectively. The AtU6-DcFAD2-1&2-9-sgRNA1 fragment was released by HindIII and XhoI digestion, and the At7SL-DcFAD2-1&2-9-sgRNA2 fragment by XhoI and SalI digestion. These fragments were subcloned into the pCDC45-Cas9-1300 vector to generate the final construct pCDC45-Cas9-1300-DcFAD2-1&2-9-sgRNA1+2. All restriction enzymes were sourced from New England Biolabs. The final construct and the empty pCDC45-Cas9-1300 vector (used as a negative control) were transformed into A. rhizogenes strain Ar1193 using the freeze–thaw method. The resulting strains were used to induce hairy roots from carrot leaf and petiole explants.

For gene overexpression, the DcFAD2-1 cDNA segment was cloned into the entry vector pENTR-SDD-TOPO. The insert was then transferred into the overexpression vector pK7WG2R (Zhao et al., 2019) using the Gateway LR Clonase Enzyme Mix (Thermo Fisher Scientific). The resulting recombinant construct was introduced into A. rhizogenes strain Ar1193 using the freeze–thaw method. Strain Ar1193 harboring the empty pK7WG2R vector was used as a negative control. Primer sequences are listed in Supplemental Table 2.

Hairy root induction was performed using a previously reported method (Liu et al., 2024). Single Ar1193 colonies were screened on LB agar plates containing the appropriate antibiotics and incubated overnight in LB liquid medium at 28°C. After incubation, bacterial cells were collected by centrifugation at 4000 rpm for 10 min at 25°C. Carrot leaf and petiole explants were collected, sterilized in 10% bleach for 10 min, and rinsed 5 times with sterile water. The explants were then cut into fragments using scissors and immersed in a suspension of A. rhizogenes strain Ar1193 for 10 min. They were then co-cultivated in B5 liquid medium supplemented with 100 μM acetosyringone in the dark at 25°C for 3 days. After co-cultivation, explants were transferred to B5 solid medium supplemented with 400 mg/L cefotaxime, and monitored for hairy root initiation at the wound sites over a period of 3–4 weeks. Once the hairy roots reached a length of 3–5 cm, they were transferred to B5 solid medium containing 5 mg/L hygromycin for selection. Creamy white hairy roots were selected for genomic DNA (gDNA) extraction and genotyping. Lines with the correct genotype were chosen for large-scale subculture in B5 liquid medium at 25°C with shaking at 90 rpm. After 1 month of growth, the hairy roots were harvested for gDNA extraction and metabolite analysis.

Quantitative real-time PCR analysis

Total RNA was extracted using the FastPure Universal Plant Total RNA Isolation Kit (Vazyme, China). First-strand cDNA was synthesized using the HiScript III RT SuperMix for qPCR (+gDNA wiper) (Vazyme). Relative transcript abundance was quantified by RT–qPCR using ChamQ Universal SYBR qPCR Master Mix (Vazyme) on an ABI 7500 system (Applied Biosystems, USA). The primers used for RT–qPCR analysis are listed in Supplemental Table 2. All experiments were performed in triplicate.

Subcellular localization analysis

The three DcFAD2 genes were cloned into the pEAQ-SalI-nNG vector and transformed into A. tumefaciens strain GV3101. Primer sequences are listed in Supplemental Table 2. A positive colony was cultured in 5 ml of LB liquid medium at 28°C with shaking at 200 rpm until the OD₆₀₀ reached 0.8–1.0. The culture was then centrifuged at 4000 rpm for 10 min, and the pellet was resuspended in 5 ml of MMA buffer (10 mM MES, 10 mM MgCl₂, and 200 μM acetosyringone in Milli-Q water). The suspension was incubated at 28°C in the dark for 1 h. The A. tumefaciens suspension (final OD₆₀₀ = 0.2 for each strain) was infiltrated into the abaxial side of 4-week-old N. benthamiana leaves using a needleless 1 ml syringe. Infiltrated plants were kept in the dark for 2 days, and then protein localization was visualized using a laser confocal microscope (A1 HD25, Nikon, Japan). Fluorescence detection parameters were set as follows: mNG was excited at 488 nm and detected at 525 nm, mCherry was excited at 561 nm and detected at 595 nm, and chloroplasts were excited at 488 nm and detected at 700 nm. The endoplasmic reticulum marker was ER-rk CD3-959, which contains a synthetic oligonucleotide encoding an tetrapeptide HDEL sequence at the C terminus and the AtWAK2 signal peptide at the N terminus of the mCherry fluorescent protein.

Statistical analysis

Statistical analysis was performed on data from at least three independent experiments using GraphPad Prism software (v.9.5.0). Figures containing bar graphs depict the mean ± SD. Statistical significance was determined using one-way ANOVA. Significance levels are indicated as follows: ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Data availability

The raw sequencing data supporting this study have been deposited in the China National Center for Bioinformation (CNCB; https://www.cncb.ac.cn) under accession number PRJCA035379 and are publicly available without restrictions.

Funding

We thank Prof. Qian Shen (Shanghai Jiao Tong University) for providing the entry vector pENTR-SDD-TOPO, Prof. Qing Zhao (Chenshan Science Research Center, CAS Center for Excellence in Molecular Plant Sciences) for providing the overexpression vector pK7WG2R, and Prof. Kexuan Tang (Shanghai Jiao Tong University) for providing the ER marker (ER-rk CD3-959). In addition, we acknowledge the metabolic service platforms at the School of Agriculture and Biology and Shanghai Jiao Tong University for assistance with metabolic analysis. We also extend our gratitude to all lab mates in the Z.L. research group for their critical reading and suggestions. Research in the Z.L. group is supported by the National Key R&D Program of China (2022YFC2602000) and the Key Project at the Central Government level: the ability establishment of sustainable use for valuable Chinese medicine resources (2060302).

Acknowledgments

No conflict of interest is declared.

Author contributions

Z.L. conceived and designed the project. J.Z. performed most of the experiments. N.A., J.W., Z.C., and C.S. assisted with material collection and preparation for metabolomics and transcriptomics. N.A. contributed to figure design and drawing. X.Z. conducted transcriptome and coexpression analyses. W.H. performed overexpression analysis. J.W. supervised metabolite analysis. J.Z., N.A., and Z.L. wrote the manuscript with contributions from all authors.

Published: March 27, 2025