Identification of genes associated with ricinoleic acid accumulation in Hiptage benghalensis via transcriptome analysis

Ricinoleic acid (12-hydroxy-9-cis-octadecenoic acid) is a hydroxy fatty acid with important industrial applications [1]. The hydroxyl group (–OH) provides unique properties to ricinoleic acid and makes this unusual fatty acid an attractive feedstock for the production of high-performance lubricants, cosmetics, polymers, surfactants, and coatings. Currently, the major commercial source of hydroxy fatty acid is castor (Ricinus communis) seed oil, which contains approximately 90% (w/w) of its fatty acids as ricinoleic acid (for a review, see [2]). However, castor is not allowed to culture for large-scale agricultural production in many countries due to the presence of the toxin ricin and allergenic 2S albumins in seeds [3]. Although the generation of genetically modified ricin-free castor lines with the RNA interference technique was recently reported [4], the seed oil content, ricinoleic acid production, and the agronomy property of the castor lines have yet to be studied. As a result, the supply of castor oil has fallen short of demand [5].

Metabolic engineering of temperate oilseed crops for hydroxy fatty acid production has been considered an attractive strategy to overcome the limitations associated with castor bean. In the past decades, numerous efforts have been put to explore the molecular mechanism of ricinoleic acid biosynthesis in castor bean and use the knowledge to produce hydroxy fatty acid in Arabidopsis thaliana (thereafter Arabidopsis) and oilseed crops [6–16]. However, it is challenging to obtain a substantial level of hydroxy fatty acids in these engineered crops. As the result, the highest content of hydroxy fatty acids achieved in transgenic plants is only about 30%, which is much lower than that in castor bean (for a review, see [17]).

Considerable efforts have also been devoted to identifying alternative plant sources of hydroxy fatty acid [18–24]. For instance, some Physaria (synonym Lesquerella) species accumulate various hydroxy fatty acids in seeds, including lesquerolic acid (14-hydroxy-11-eicosenoic acid, 20:1–OH), densipolic acid (12-hydroxy-9, 15-octadecadienoic acid, 18:2–OH) and auricolic acid (14-hydroxyeicosa-11, 17-dienoic acid, 20:2–OH), and thus have been explored for hydroxy fatty acid production [25]. Physaria fendleri is one of the most extensively studied Physaria species, which produces 24–36% of oil in seeds and has approximately 60% (w/w) of its fatty acids as lesquerolic acid [26]. This short-live perennial can grow well in semi-arid regions of North America and can tolerate freezing temperatures [25, 26]. Several advances have been achieved through breeding and agronomic research in the past decade [25, 26]. However, P. fendleri has some unfavorable agronomic characters relating to pest resistance, disease resistance, soil requirement and irrigation requirement, and thus has not been successfully domesticated as an oilseed crop for the large-scale production of hydroxy fatty acid [19, 25, 27, 28]. In addition, lesquerolic acid has different carbon length to ricinoleic acid, which may have negative effects on its potential applications in industry [29]. Therefore, it is attractive to identify new plant species containing large amount of ricinoleic acid as an alternative to castor bean.

Hiptage benghalensis is a vine-like plant native to temperate and tropical Asia region. This plant has been used for medicinal and ornamental purposes in India and Thailand for many years. The Hiptage genus (Malpighiaceae family) is composed of 20–30 species in the world and some of them contain about 50% of oil in seeds and more than 70% percent of fatty acids as ricinoleic acid, and thus have potential to be domesticated as an alternative source of castor oil [30, 31]. In the Malpighiaceae family, the seeds typically exhibit reasonably homogeneous structure with a well-developed embryo and the endosperm mostly absorbed during seed development [32]. Therefore, ricinoleic acid in H. benghalensis is likely accumulated in embryo, similar to Physaria species [21]. Unlike the extensive studies regarding ricinoleic acid biosynthesis in castor, the mechanism of ricinoleic acid accumulation in Hiptage plants is yet to be explored. Indeed, our current understanding of ricinoleic acid accumulation in plants is mainly built on the study of castor bean. Exploring the metabolic pathways of lipid metabolism from other ricinoleic acid-enriched plants, such as H. benghalensis, may bring novel insight into the molecular mechanisms underlying ricinoleic acid accumulation, and provide valuable knowledge for engineering ricinoleic acid production in other plants. Moreover, castor (Euphorbiaceae family) and Hiptage species belong to different families with long genetic distance and different evolutionary history. The study on ricinoleic acid biosynthesis and regulation in Hiptage plants may identify novel genes with unique properties, which can be further used in the breeding of proposed oilseed crops for ricinoleic acid production.

The aim of this study, therefore, is to explore the mechanisms of ricinoleic acid biosynthesis and regulation in H. benghalensis at the transcriptional level. First, the fatty acid composition of six Hiptage spp. was compared and H. benghalensis was identified as the one with the highest ricinoleic acid content in seeds. Second, the comprehensive transcriptome profiles of H. benghalensis seeds at different developing stages were obtained by a joint analysis with second-generation sequencing (SGS) and single-molecule real-time sequencing (SMRT) and the expression of functional transcripts encoding enzymes associated with lipid biosynthesis was analyzed. Third, the key transcription factors (TFs) and long noncoding RNAs (lncRNAs) were identified and their co-expression profiles correlated with genes in lipid biosynthesis pathways were analyzed. Finally, a network of ricinoleic acid biosynthesis and regulation in H. benghalensis was proposed and the transcriptional profiles of the identified important genes were summarized.

Ricinoleic acid is one of the most important unusual fatty acids with broad industry applications, but its production is limited by the unfavorable agronomy status of castor bean. Nevertheless, the plant kingdom is composed of more than 350,000 species and dozens of them have been reported to accumulate hydroxy fatty acids in seeds [37]. Exploring alternative plant sources with agricultural value is, therefore, pivotal for ricinoleic acid production. Moreover, the identification of ricinoleic acid accumulation mechanism in alternative plant species will also provide novel and valuable candidate genes for producing ricinoleic acid in other plants via genetic engineering. Hiptage plants contain high levels of ricinoleic acid in seed oils, and thus represent an alternative source of ricinoleic acid [31, 32]. In this study, ricinoleic acid contents in seeds of six Hiptage species were compared and H. benghalensis seeds contain as high as 81% of total fatty acid as ricinoleic acid (Additional file 1: Table S1, Fig. 1).

For a better understanding of ricinoleic acid biosynthesis and regulation in H. benghalensis at the molecular level, combined SGS short-read sequencing and SMRT full-length sequencing were performed to comprehensively analyze the transcriptome profile in developing seeds. SMRT sequencing yields kilobase-sized sequence reads which usually represent full-length or nearly full-length transcripts without the need for further assembly. However, SMRT sequencing is more expensive with relatively high error rates than SGS sequencing. On the other hand, although the low-cost SGS sequencing method can obtain transcriptome data at a greater sequencing depth with accurate sequence reads, it requires accurate assembly of the short reads based on a reference genome. Since the reference genome of H. benghalensis is not available, the combination of SMRT and SGS analyses in this study could provide effective and reliable data of the H. benghalensis transcriptome profile.

With oleic acid as the substrate, FAH12 catalyzes the production of ricinoleic acid at the sn-2 position of phosphatidylcholine (PC) [38, 39]. Two H. benghalensis FAH12 (HbFAH12-1 and HbFAH12-2) genes were identified in this study (Additional file 8: Figure S5; Additional file 9: Table S4). HbFAH12-1 showed a very low expression level at seed-developing stages S3 to S5. On the contrary, HbFAH12-2 had a very high expression level from S2 to S4 and showed a concave-rise pattern (Additional file 9: Table S4), which is consistent with the rapid accumulation of ricinoleic acid at these stages (Fig. 1). In addition, the expression of HbFAH12-2 was much higher than FATTY ACID DESATURASE 2 (HbFAD2) and HbFAD3. Therefore, HbFAH12-2, but not HbFAC12-1, most likely catalyzes ricinoleic acid synthesis in H. benghalensis. Comparing to other hydroxy fatty acid-producing plants, similar results have been observed in castor [40] but not in P. fendleri [24] (Additional file 7: Figure S4).

Analysis of co-expression networks based on the similarity in gene expression is a powerful approach to accelerate the elucidation of molecular mechanisms underlying important biological processes [34]. Specifically, co-expression gene networks can help to narrow down causal relationships among a large number of genes. In this study, cluster analysis revealed that 33 differentially expressed lipid genes, including HbKASII and HbSAD, had a similar expression pattern with HbFAH12-2 (Fig. 3a). KASII and SAD catalyze the conversion of palmitoyl-ACP to stearoyl-ACP and stearoyl-ACP to oleoyl-ACP, respectively. The high expression levels of HbKASII and HbSAD may contribute to the rapid synthesis of oleic acid (Fig. 1c), the precursor of ricinoleic acid.

The newly generated ricinoleic acid needs to be assembled into TAG by the Kennedy pathway and acyl editing [41]. The Kennedy pathway is catalyzed by glycerol-3-phosphate acyltransferase (GPAT), lysophosphatidic acid acyltransferase (LPAAT), phosphatidic acid phosphatase (PAP), and acyl-CoA: diacylglycerol acyltransferase (DGAT) [42]. Acyl editing includes PC acyl remodeling or acyl exchange between PC and diacylglycerol (DAG), PC and acyl-CoA pool, or PC and triacylglycerol (TAG), catalyzed by multiple enzymes such as PDAT, LPCAT, phosphatidylcholine:diacylglycerol cholinephosphotransferase (PDCT), cholinephosphotranferases (CPT), and phospholipase A₂ (PLA₂) (for reviews, see [43, 44]).

In H. benghalensis, several putative genes from TAG assembly pathways including LPAAT2, PAP2, PDAT2, LPCAT1, and PDCT1 showed similar expression patterns with HbFAH12-2 (Fig. 3; Additional file 8: Figure S5, Additional file 9: Table S4). PDAT plays a significant role in channeling ricinoleoyl groups from PC to TAG. PDCT catalyzes the conversion between DAG and PC and the resulting ricinoleoyl–DAG can be further utilized by DGAT to form TAG [13, 45]. Moreover, the reverse reaction of LPCAT may contribute to the enrichment of hydroxy fatty acid in castor bean, H. benghalensis, and P. fendleri [46]. The LPCAT from these three plant species displayed four–six times higher preference towards ricinoleoyl group over oleoyl group in the reverse reaction [46]. Considering the enhanced expression levels of HbPDAT, HbPDCT, and HbLPCAT during seed development (Figs. 1 and 3), these enzymes may play important roles in channeling hydroxy fatty acids from PC to TAG in H. benghalensis. On the contrary, since no different expression transcript was identified as HbCPT, this gene might only play a minor role in ricinoleic acid accumulation (Fig. 3a, Additional file 9: Table S4). The result about HbCPT is consistent with the observations in castor and P. fendleri (Additional file 7: Figure S4).

In addition to the above-mentioned lipid biosynthetic enzymes, proteins involving in lipid transport and TAG storage may play crucial roles in ricinoleic acid accumulation. ACBPs are the predominant carriers of acyl-CoA esters which have the ability to bind long-chain acyl-CoA esters to protect them from hydrolysis [47]. HpACBP6, an ortholog of the soluble Arabidopsis ACBP6, showed high expression levels and a similar expression pattern with HbFAH12-2 (Fig. 3a, Additional file 9: Table S4). Similarly, ACBP6 was the dominantly expressed isoform in castor [40] and P. fendleri [24]. These suggest that the increased expression of ACBP6 may be related to the increased levels of hydroxy acyl-CoAs which are assembled into TAG.

In oleaginous plants, most TAG is stored in oil bodies, consisting of a TAG core surrounded by a monolayer of phospholipid embedded with oil-body-membrane-associated proteins [48]. OLEs are the most abundant oil-body proteins, whereas the other two classes of proteins (steroleosins and caleosins) were also found to be associated with oil bodies [48]. H. benghalensis OLE3 was the most expressed OLE with a similar expression pattern as HbFAH12-2 (Fig. 3a, b, Additional file 6: Table S3, Additional file 8: Figure S5, Additional file 9: Table S4), though high OLE3 expression was not detected in castor bean and P. fendleri (Additional file 7: Figure S4). When HbOLE3 was compared with a peanut (Arachis hypogaea) OLE which was reported to be a bifunctional enzyme displaying both monoacylglycerol acyltransferase and PLA₂ activities [49], the major motifs of acyltransferase (HXXXXD/E) and lipase (GXSXG) are not present in HbOLE3. Therefore, HbOLE3 may be involved in ricinoleic acid accumulation by functioning in oil-body formation according to its high expression levels and co-expression with other lipid-related genes in developing seeds, instead of contributing as a monoacylglycerol acyltransferase and PLA₂ bifunctional enzyme. In addition, an Arabidopsis caleosin ortholog (PXG1) showed high expression levels in H. benghalensis and a similar expression pattern with HbFAH12-2 (Fig. 3a, b, Additional file 9: Table S4), which was not observed in castor bean [40] and P. fendleri [24]. PXG1 encodes a caleosin with peroxygenase activity, which may be involved in the formation of anti-fungal hydroxy fatty acid derivatives [50]. Therefore, PXG may play a role in catalyzing the ricinoleic acid formation in H. benghalensis seeds in addition to its function in TAG storage.

Other differentially expressed lipid genes, such as DGAT2 and LONG-CHAIN ACYL-COA SYNTHETASE 8 (LACS8), also had high expression levels at stages S3–S5 and thus appear to contribute to ricinoleic acid accumulation (Fig. 3a, Additional file 6: Table S3, Additional file 8: Figure S5; Additional file 9: Table S4). As the last enzymes catalyzing TAG formation, both DGAT and PDAT, have determinant roles in channeling the hydroxy acyl flux into TAG [51]. The relative expression of DGAT and PDAT in H. benghalensis was compared with those of castor bean and P. fendleri (Additional file 7: Figure S4). The result indicated that the three hydroxy fatty acid-producing plant species might have different routes to produce TAG (Additional file 7: Figure S4). In H. benghalensis, DGAT2 and PDAT2 were dominantly expressed during ricinoleic acid accumulation and thus may contribute primarily to the formation of ricinoleic acid-enriched TAG. On the other hand, DGAT2, rather than PDAT, was the major player in the enrichment of TAG with ricinoleic acid in castor, whereas in P. fendleri, DGAT1, DGAT2 and PDAT2 all contributed to the hydroxy fatty acid enrichment. It should be noted that LACS8 was the predominantly expressed LACS isoform in H. benghalensis, whereas LACS9 was dominant in castor bean and P. fendleri (Additional file 7: Figure S4). Indeed, LACS8 has also been proposed to be involved in channeling modified fatty acids from PC to the acyl-CoA pool together with PLA₂, which can further be used by DGAT or other enzymes in the Kennedy pathway to form TAG [52].

In this study, co-expression analyses were also performed to identify potential TFs and lncRNAs involving in the regulation of lipid-related genes for ricinoleic acid accumulation. TFs are key regulators in metabolic networks, in which one TF can simultaneously regulate the expression of multiple genes and one gene can be simultaneously regulated by multiple TFs [53]. Previous studies indicated that WRI1, FUS3, LEC1 (NF-YB6), and ABI3 were key TFs regulating oil biosynthesis [54, 55]. The analysis of H. benghalensis seed transcriptome showed that these four TFs had high expression levels and were all co-expressed with some lipid-related genes (Fig. 4a, Additional file 10: Table S5). Moreover, bZIP67 showed high expression levels and high degrees of interaction with important lipid-related genes in H. benghalensis developing seeds (Fig. 4a, Additional file 10: Table S5). bZIPs have been previously reported to be a group of crucial regulators on seed lipid production in Arabidopsis [56]. The high correlations between bZIP67 and lipid-related genes including SAD, KASII, LPCAT1, PDAT2, OLE3 and ACBP6 (Fig. 4a) may suggest a possible role of bZIP67 in the regulation of ricinoleic acid accumulation in H. benghalensis seeds. In addition, TFs such as TCP, EN61, IDD4, REM16, TZF3, TZF4, ARF10, and NFYA1 also showed high expression levels and were co-expressed with lipid-related genes (Fig. 4A; Additional file 10: Table S5). Therefore, they may also be correlated with ricinoleic acid accumulation. From another perspective, both HbFAH12-2 and HbOLE3 have co-expression linkages with multiple TFs, indicating the importance of these two genes in ricinoleic acid accumulation and the interactions between lipid-related genes and TFs. Further analysis of these genes with forward and backward genetic methods would explain their interaction in H. benghalensis in detail. In addition, it is also interesting to study how FAH12-2 and OLE3 interact with TFs in Arabidopsis and oilseed crops.

Recently studies showed that plant lncRNAs are important players in various biological pathways, though only some of them have been thoroughly studied [36, 57]. In this study, 746 lncRNAs, including 664 novel ones, were identified. Among them, 35 DElncRNAs were co-expressed with the major lipid-related genes (Fig. 4b, Additional file 11: Table S6, Additional file 12: Table S7). Further analysis revealed that lncRNAs PB24473, PB56338, PB3144, and PB6000 had high expression levels at mid-developing stages in seeds (Additional file 12: Table S7) and high correlations with several important lipid-related genes (Fig. 4b). Therefore, these four lncRNAs may play substantial roles in ricinoleic acid accumulation in H. benghalensis.

Nevertheless, it should be noted that lncRNAs are not only positively co-expressed with FAH12-2. A few lncRNAs were indeed negatively co-expressed with FAH12-2. Similar results were also identified in the co-expression networks of TFs (Fig. 3a). As an unusual fatty acid, ricinoleic acid is predominantly present in the form of TAG in H. benghalensis seed (Additional file 2: Figure S1), which may also play a possible role in defense against pests in plants [58]. Also similar to other unusual fatty acids, ricinoleic acid is likely deleterious to cell membranes [59]. The positive and negative co-expression of lncRNAs may contribute to the dynamic balance of ricinoleic acid in the embryo and endosperm cell membranes in H. benghalensis seeds. Moreover, the high co-expression of HbFAH12-2 and HbOLE3 with lncRNAs may indicate the important roles of these two genes in the regulation process.