Introduction

Traditional safflower oil contains about 6–8 % palmitic acid, 2–3 % stearic acid, 16–20 % oleic acid and 71–75 % linoleic acid; hence, it is regarded as one of most highly polyunsaturated vegetable oils (Velasco and Fernandez-Martinez 2001). However, there are natural mutants and breeding lines with various levels of elevated oleic acid mainly at the expense of linoleic acid (Fernández-Martinez et al. 1993). Consumption of oils with high oleic acid content (HO) is regarded as desirable because of its cholesterol-lowering effect and enhanced oxidative stability that reduces the need for hydrogenation, a process known to generate nutritionally undesirable trans fatty acids. Further, the combination of high oxidative stability and low melting point characteristics of HO safflower oil also renders it more suitable for biodiesel applications and biodegradable replacements for mineral oils, such as hydraulic oils and lubricants (Kinney and Clemente 2005). Furthermore, purified oleic acid could be used to manufacture a range of oleochemicals such as diacids that are found in the formulations of hundreds of different oleochemical products, offering an interesting suite of properties including elasticity, flexibility, impact strength and hydrolytic stability (Crandall 2002; Hill 2000).

The original HO trait in safflower, found in an introduction from India, is controlled by a partially recessive allele ol at a single locus OL (Knowles and Bill 1964), with oleic acid content of olol homozygous genotypes reaching 70–75 % (Knowles 1989). The ol allele was incorporated into safflower breeding programmes and the first HO safflower variety “UC-1” was released in 1966 in the USA, which was followed by the releases of “Oleic Leed” and the Saffola series including Saffola 317 (S-317), S-517 and S-518. The ol allele has also been used as the background genetic material for further enhancement of oleic acid content in safflower breeding programmes worldwide (Mündel and Bergman 2009; Weiske 1999).

Fatty acid biosynthesis is a highly regulated and compartmentalised process in higher plants (Ohlrogge and Jaworski 1997). Saturated fatty acids and monounsaturated fatty acids, mostly oleic acid, are synthesised in plastids. Following its exit from plastids, oleic acid can be further modified on phosphatidylcholine (PC) by the microsomal oleate Δ12 desaturase (FAD2; EC1.3.1.35), using NADH, NADH-cytochrome b5 reductase and cytochrome b5 as electron donors (Shanklin and Cahoon 1998). Fatty acids formed on PC are exchanged with the acyl-CoA pool prior to incorporation into storage lipids, mainly triacylglycerol (TAG) via the Kennedy pathway (Ohlrogge and Browse 1995). TAG can also be formed via an acyl-CoA independent pathway catalysed by phospholipid:diacylglycerol acyltransferase (PDAT) (Dahlqvist et al. 2000). Numerous biochemical, genetic and transgenic studies have clearly indicated that FAD2 is the key enzyme controlling the relative accumulation of oleic acid versus linoleic acid in safflower seed oils (Stymne and Appelqvist 1978).

To explore the biochemical nature of the olol mutation, we carried out in vitro microsomal FAD2 enzyme assay comparing HO and HL genotypes. This revealed a defective FAD2 as the ol allele. We have isolated an exceptionally large FAD2 gene family from safflower consisting of at least 11 members, including the seed-specific CtFAD2-1 (Cao et al. 2013). We have confirmed a previous finding that a single-nucleotide deletion in the coding region of CtFAD2-1 would cause a frame shift and lead to attenuated transcription (Guan et al. 2012). In this study, we carried out in vitro biochemical assays of FAD2 enzyme activity, directly associating the high oleic ol allele with reduced FAD2 activity. We also propose that the molecular basis of the olol genotype is mediated by nonsense-mediated mRNA decay (NMD) of CtFAD2-1, a process that typically degrades transcripts containing a premature termination codon (PTC). Further, we have designed a set of PCR-based perfect molecular markers for the identification of the ol allele, which may allow for rapid screening and early detection of genotypes carrying the ol allele for the management of the HO trait in safflower breeding programmes.

Materials and methods

Plant materials

A wild-type or HL safflower ‘SU’ and five HO safflower varieties, including ‘S-317’, ‘S-517’, ‘LeSaf486’, ‘CW99-OL’ and ‘Ciano-OL’, were used in this study. SU is commonly grown as a bird seed crop in Australia, and it was obtained from Heffernan seeds in NSW. A few HO varieties including LeSaf486 (PI603208, ATC 120562), CW99-OL (ATC 120561) and Ciano-OL were obtained from Australian Temperate Field Crops Collections in Horsham, Victoria, Australia. Other HO varieties, including S-317 and S-517, were supplied by Devexco International. The plants were grown from seeds and maintained under glasshouse conditions with day/night cycle of 25/22 °C and 16 h/8 h photoperiod.

Microsomal assay of the FAD2 enzyme activity

Safflower microsomes from S-317 and SU developing seeds 15 days after anthesis (DAA) were freshly prepared as described previously (Stymne and Appelqvist 1978). The assay mixture (100 μL) contained 40 μg microsomal protein, 2 nmol [1-14C]oleoyl-CoA (10 K dpm/nmole), 10 nmole CoA and 1 % BSA in 0.1 mM potassium phosphate buffer pH 7.2 with or without 5 mM NADH. The incubations were carried out in a water bath at 30 °C with constant shaking for 10 min, followed by another 5, 10 and 20 min, respectively, after adding NADH. The reactions were stopped by adding 225 μL of methanol:CHCl3:HAc = 50:50:1 (V:V:V). The lower CHCl3 phases were recovered, dried and loaded onto a silica gel 60 thin layer chromatography (TLC) plate (MERCK, Dermstadt, Germany) and developed with a solvent mixture containing CHCl3:Methanol:HAc:H2O in the ratio of 90:15:10:3. The PC fraction was isolated and methylated with methanolic-HCl at 80 °C for 2 h as previously described (Zhou et al. 2006). The resultant fatty acid methyl esters (FAMEs) were separated on AgNO3-treated argentation TLC plate with hexane:diethyl ether:HAc (85:15:1). The plates were exposed and analysed by a Fujifilm FLA-5000 phosphorimager (Fujifilm, Fuji, Tokyo, Japan). The radioactivity of each sample was quantified with Fujifilm Multi Gauge software.

Ectopic expression of CtFAD2-1 and CtFAD2-1Δ in Saccharomyces cerevisiae

The entire coding regions of CtFAD2-1 and CtFAD2-1Δ were first PCR amplified and ligated behind the GAL1 promoter in a pYES2 vector in sense orientation for inducible gene expression. The oligo primers used to amplify the entire coding region of CtFAD2-1 and CtFAD2-1Δ were: s1: 5′-TGAAAGCAAGATGGGAGGAGG-3′ and a1: 5′-TCACAACTTTACTTATTCTTGT-3′. The resulting plasmids and the empty pYES2 vector (negative control) were introduced into bakers’ yeast (Saccharomyces cerevisiae) YPH499 cells by lithium acetate-mediated transformation. The transformed cells derived from a single colony were grown for 2 days at 28 °C in synthetic dropout medium lacking uracil, but supplemented with glucose (SD-glucose) liquid medium. The expression of CtFAD2-1 and CtFAD2-1Δ was induced by transferring the cells to a fresh SD liquid medium supplemented with galactose instead of glucose and grown with shaking for an additional 2 days. The cells were harvested by centrifugation. FAMEs were prepared by transesterification of the total fatty acids in yeast cell pellets by adding 750 μL of 1 N MeOH–HCl at 80 °C for a minimum of 2 h prior to adding 500 μL of 0.9 % NaCl. FAMEs were extracted with 300 μL of hexane and analysed by Agilent 7890A gas chromatography (GC) with a 30-m BPX70 column essentially as described (Zhou et al. 2011). Each experiment was carried out in triplicate. Total RNAs of the yeast cells expressing either CtFAD2-1 or CtFAD2-1Δ were isolated using Trizol reagent (Invitrogen, Carlsbad, USA) and real-time quantitative RT-PCR (RT-qPCR) was carried out for gene expression studies.

Transient expression of CtFAD2-1 and CtFAD2-1Δ in Nicotiana benthamiana leaves

The entire coding regions of CtFAD2-1 derived from SU and CtFAD2-1Δ derived from S-317 were each cloned in sense orientation into a modified pORE04 binary vector between the double CaMV-35S promoter and an Agrobacterium tumefaciens NOS terminator containing the polyadenylation signal sequence (Coutu et al. 2007), forming the vectors 35S:CtFAD2-1 and 35S:CtFAD2-1Δ, respectively. A vector constitutively expressing the viral suppressor protein, P19, was obtained from Dr Peter Waterhouse’s laboratory (University of Sydney). A. tumefaciens strain AGL1 harbouring the 35S:CtFAD2-1 or 35S:CtFAD2-1Δ was co-infiltrated with the 35S:P19 culture into the underside of the fully expanded N. benthamiana leaves as previously described (Voinnet et al. 2003; Wood et al. 2009). Following a period of 5 days of further growth at 24 °C, the infiltrated leaf regions were excised and immediately subjected to RNA isolation using an RNeasy Mini Kit (Qiagen, Hilden, Germany).

Ectopic expression of CtFAD2-1 and CtFAD2-1Δ in Arabidopsis

Constructs expressing either CtFAD2-1 derived from SU or CtFAD2-1Δ derived from S-317, each driven by the seed-specific promoter Fp1 derived from a truncated Brassica napus napin gene (Stalberg et al. 1993), were transformed to Arabidopsis thaliana ecotype Col-0 via the A. tumefaciens dipping method. Inoculations were performed by dipping the aerial parts of plants at flowering stage for a few seconds in 300 mL of a solution containing 5 % (w/v) sucrose, 10 mm MgCl2, resuspended A. tumefaciens cells from a 150 mL overnight culture and Silwet L-77 (Lehle Seeds, TX, USA) following Bent and Clough (1998). The A. thaliana plants were maintained in a glasshouse with constant 22 °C and 16 h photoperiod until seed maturity had been reached. Twelve lines of A. thaliana independently transformed with each of the aforementioned constructs were established following selection on 50 mg/L kanamycin on Murashige and Skoog medium. The siliques containing middle maturity developing embryos from these transgenic plants were harvested and total RNAs were isolated using the RNeasy Mini Kit (Qiagen, Hilden, Germany).

Real-Time quantitative PCR analysis of CtFAD2 expression

Total RNAs from yeast, developing A. thaliana siliques and N. benthamiana leaves were isolated using RNeasy Mini Kit (Qiagen, Hilden, Germany). Contaminating DNA was removed by digestion with TURBO RNA-free DNaseI (Ambion, TX, USA) according to the manufacturer’s protocol. RNA concentrations were determined using a Nanodrop® spectrophotometer ND1000 (Thermo Fisher Scientific, Victoria, Australia), and concentrations were equalised before analysis. To verify RNA integrity, 1 μg of total RNAs from each sample was visualised on an ethidium bromide-stained 1.5 % agarose gel following electrophoresis.

The gene expression patterns were studied with RT-qPCR carried out in triplicate using Platinum SYBR Green qPCR SuperMix (BioRad, CA, USA) and run on ABI 7900HT Sequence Detection System. Each PCR reaction contained 20 ng of total RNA template, 800 mM each of the forward and reverse primers, 0.25 μL of reverse transcriptase and 5 μL One-Step RT-PCR Master Mix reagents, increased to 10 μL total volume with nuclease-free water. PCR was carried out with an initial cycle at 48 °C for 30 min and 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 60 s. The primers for CtFAD2-1 are sense: 5′-GTGTATGTCTGCCTCCGAGA-3′; antisense: 5′-GCAAGGTAGTAGAGGACGAAG-3′. A constitutively expressed reference gene from safflower, CtKASII, was used to normalise the relative quantities. KASII is responsible for the elongation of C16:0-ACP to C18:0-ACP in de novo fatty acid biosynthesis in plants. Safflower KASII gene (CtKASII) has been previously used as an internal reference gene because of its high expression stability in various tissues and developmental stages (Cao et al. 2013). The primers for CtKASII are sense: 5′-CTGAACTGCAATTATCTAGG-3′; and antisense: 5′-GGTATTGGTATTGGATGGGCG-3′. The calculations were made using the comparative CT method as reported (User Bulletin #2, Applied Biosystems). The data are presented as mean ± SD of three reactions performed on independent 96-well plates.

Small RNA Northern blot analysis

Approximately, 10 μg of total RNAs from 15 DAA developing embryos of SU and S-317 were separated using a 17 % denaturing polyacrylamide gel and blotted onto Hybond-N+ membranes (GE Healthcare, NJ, USA). The membranes were UV cross-linked and pre-hybridised at 42 °C for 3 h in hybridisation buffer containing 50 % formamide, 5 × SSPE (3 M NaCl, 0.2 M NaH2PO4, and 0.02 M EDTA, pH7.4), 5× Denhardt’s solution, 1 mM EDTA, 1 % BSA and 1 % SDS. DNA oligos antisense to CtFAD2-1 and CtFAD2-1Δ were end labelled by the forward reaction using 10 units of T4 polynucleotide kinase (Roche Molecular Biochemicals, Indianapolis, IN, USA) with the supplied buffer, to which 300 nM [γ-32P] ATP (3,000 Ci/mmol) was added. The reaction was incubated for 1 h at 37 °C. Unincorporated 32P-label was removed using a G-25 microcolumn (GE Healthcare, NJ, USA). Probes were added to the hybridisation buffer and hybridisation was allowed to proceed at 42 °C overnight. The membranes were then washed twice, 30 min each in 2×  SSC and 0.2 % SDS at 40 °C. Hybridisation signal was detected and analysed by a Fujifilm FLA-5000 phosphorimager (Fujifilm, Tokyo, Japan).

Deep sequencing of safflower small RNAs

Small RNAs were extracted from the mid-maturity developing seeds at 15 DAA of SU and S-317 plants using the mirVana™ miRNA Isolation Kit (Ambion, CA, USA) following the manufacturer’s instructions. Subsequently, small RNAs were subjected to 15 % (w/v) denaturing polyacrylamide gel electrophoresis (PAGE), and 18- to 25-bp portions were excised from the gel and purified. The purified small RNA molecules were then ligated to the Solexa 5′ and 3′ adaptor sequentially and converted to cDNA following the Illumina protocol. Deep sequencing was performed on the Illumina HiSeq2000 and the samples were run side by side at the Beijing Genome Institute (BGI, Shenzhen, China). After removing the low-quality reads and those that were less than 18 nt in length, the small RNA data derived from both SU and S-317 were “Blast” searched for sequences corresponding to CtFAD2-1 and CtFAD2-1Δ, respectively.

Developing a perfect PCR marker and PCR for ol allele

A large intron located in the 5′UTR of CtFAD2-1 in SU or CtFAD2-1Δ in S-317 was amplified by PCR using forward oligo primer intron-s1: 5′-GAGATTTTCAGAGAGCAAGCGCTT-3′ and reverse oligo primer: intron-a1: 5′-CTTTGGTCTCGGAGGCAGACATA-3′. Based on the unique DNA sequences of CtFAD2-1Δ, a pair of oligo primers was designed to amplify a specific band of 300-bp long from the genomic DNA of the HO genotypes, including S-317, S-517, CW99-OL, LeSaf496 and Ciano-OL. The sequences of these primers are: HO-S1, 5′-ATAAGGCTGTGTTCACGGGTTT-3′; and HO-A1, 5′-GCTCAGTTGGGGATACAAGGAT-3′. A pair of oligo primers specific for the conventional HL safflower SU is designated as follows: HL-S1, 5′-AGTTATGGTTCGATGATCGACG -3′; and HL-A1, 5′-TTGCTATACATATTGAAGGCACT-3′. The primers derived from the safflower CtKASII gene, which were also used in RT-qPCR, were used as the positive control and to ensure equal loading in this experiment. PCR was conducted using HotStar mix following the manufacturer’s instructions (Qiagen, Hilden, Germany). The PCR cycle was 94 °C for 15 min, followed by 40 cycles of 94 °C for 30 s, 58 °C for 30 s and 72 °C for 1 min. The reaction products were separated by electrophoresis on a 1 % agarose gel and visualised under UV light following EtBr staining. The amplicon’s identity was confirmed by DNA sequencing.

Results

In vitro microsomal analysis of oleate desaturase activity

Microsomal oleate desaturation assays were performed on developing embryos of 15 DAA. After argentation TLC and autoradiography, examination of the autoradiographs indicated that the microsomal extracts derived from developing safflower embryos of HL safflower SU were able to rapidly desaturate [14C]oleoyl-CoA in the presence of NADH. As shown in Fig. 1, upon the addition of NADH, rapid appearance of [14C] linoleate in the PC fraction was observed following 5 min reaction, indicating efficient biosynthesis of [14C] linoleate from oleate in the HL safflower microsomes in vitro. In contrast, in microsomes prepared from developing embryos of S-317, the biosynthesis of linoleic acid via desaturation of oleate was significantly less. The ratio of [14C] oleate versus [14C] linoleate in the microsomal in vitro reaction throughout the time course was also compared and is shown in Table 1.

Fig. 1
figure 1

Argentation TLC analysis of FAMEs derived from PC isolated from in vitro microsomal reactions comparing SU and S‐317. Lanes 1 and 2 were duplicated reactions without NADH; lanes 3 and 4 were duplicated reactions with NADH for 5 min; lanes 5 and 6 were duplicated reactions with NADH for 10 min; lanes 7 and 8 were duplicated reactions with NADH for 20 min. a SU; b S‐317

Full size image
Table 1 Percentage of linoleate derived from oleate in safflower microsomes
Full size table

CtFAD2-1Δ codes for a non-functional microsomal Δ12 oleate desaturase

To verify the functionality loss of the CtFAD2-1Δ from HO genotypes, its corresponding ORF was cloned into the expression vector pYES2 behind the inducible GAL1 promoter and transformed into S. cerevisiae. CtFAD2-1 derived from the conventional HL safflower SU was used as a positive control. FAMEs were prepared from transgenic yeast cells that were induced with galactose and grown to stationary phase. Fatty acid composition of the harvested yeast cells was analysed by GC (Fig. 2). As expected, the yeast cells transformed with CtFAD2-1 showed the presence of two dienoic acids, palmitolinoleic acid (C16:2) and linoleic acid (C18:2), which were not present in the untransformed yeast or in the control cells transformed with the empty vector. In contrast, the expression of the CtFAD2-1Δ did not result in accumulation of any dienoic acid, indicating that CtFAD2-1Δ had lost its functionality as a microsomal Δ12 oleate desaturase. We also measured the transcript levels of both CtFAD2-1 and CtFAD2-1Δ in yeast cells, which were at comparable levels as shown in Fig. 3. The transcript reduction of CtFAD2-1Δ as demonstrated in HO safflower developing embryos by Guan et al. (2012) and our own experiment (data not shown) was not observed in transgenic yeast cells.

Fig. 2
figure 2

Functional identification of CtFAD2-1 and CtFAD2-1Δ by ectopic expression in yeast. a. CtFAD2‐1; b. CtFAD2‐1Δ

Full size image
Fig. 3
figure 3

RT-qPCR analysis of gene expression in yeast transformed with CtFAD2-1 and CtFAD2‐1Δ

Full size image

Northern blot and deep sequencing analysis of small RNAs

Northern blot analysis of small RNAs isolated from mid-maturity developing embryos of both HL safflower (SU) and HO genotypes S-317 and LeSaf486 revealed no discernible level of small RNA corresponding to CtFAD2-1 or CtFAD2-1Δ (Fig. 4).

Fig. 4
figure 4

Northern blot analysis of small RNAs corresponding to CtFAD2‐1 and CtFAD2‐1Δ. The top panel was hybridised with U6 RNA as a positive control. The lower panel was hybridised with CtFAD2‐1. Lane 1 SU, lane 2 S‐317, lane 3 Lesaff496, lane 4, RNA ladder

Full size image

The lack of significant levels of small RNAs corresponding to CtFAD2-1 or CtFAD2-1Δ was confirmed by small RNA deep sequencing. Solexa sequencing of small RNAs derived from SU and S-317 generated 3,741,194 and 3,585,498 sequences ranging from 18 to 30 nucleotides (nt), respectively. These two small RNA databases were BLAST searched with CtFAD2-1 and CtFAD2-1Δ, respectively. A comparable low number of small RNAs corresponding to the CtFAD2-1 in conventional HL safflower SU and CtFAD2-1Δ in HO genotype S-317 were detected. The small RNAs corresponding to the CtFAD2-1Δ sequence in S-317 were not significantly higher than those in SU (Fig. 5). The small RNAs corresponding to CtFAD2-1Δ were not expected to be sufficient to generate any substantial effect of posttranscriptional gene silencing in S-317.

Fig. 5
figure 5

The number of small RNAs corresponding to the cDNA sequence in conventional HL safflower SU CtFAD2‐1 (filled circle) and HO genotype S‐317 CtFAD2‐1Δ (open circle)

Full size image

Transient expression of CtFAD2-1Δ in N. benthamiana leaves

To test whether the ectopic expression of CtFAD2-1Δ harbouring a PTC would result in NMD-mediated mRNA degradation in a plant system, we first investigated its expression level measured by RT-qPCR in a heterologous plant host N. benthamiana transient expression system. N. benthamiana leaves infiltrated with A. tumefaciens culture containing either the 35S:CtFAD2-1 or the 35S:CtFAD2-1Δ expression vectors were harvested 5 days after infiltration and total RNAs were isolated for RT-qPCR analysis. As shown in Fig. 6, the expression level of CtFAD2-1Δ derived from S-317 was significantly lower than that of CtFAD2-1, with the former expressed at levels that were observed to be at least twofold lower than the latter.

Fig. 6
figure 6

Real-time RT‐qPCR analysis of CtFAD2‐1 and its mutant CtFAD2‐1Δ that were transiently expressed in N. benthamiana leaves, controlled by a 35S promoter. CtFAD2‐1 (filled circle) and CtFAD2‐1Δ (open circle)

Full size image

Seed-specific expression of CtFAD2-1Δ in Arabidopsis

A. thaliana transgenic lines were produced to investigate whether the NMD phenomenon observed in the expression of CtFAD2-1Δ could also be observed in the developing seeds of a heterologous host A. thaliana. A truncated B. napus napin promoter Fp1 was used to drive the seed-specific expression of the transgenes. The Fp1:CtFAD2-1 and Fp1:CtFAD2-1Δ plasmids were introduced into A. tumefaciens that were used to transform wild-type A. thaliana. Based on kanamycin selection, 12 primary transgenic lines were generated from the transformations involving each plasmid. Kanamycin-resistant T1 plants were grown to flowering and developing siliques containing T2 immature embryos were harvested for further analysis. The gene expression patterns of CtFAD2-1 and CtFAD2-1Δ were comparatively analysed by RT-qPCR. As shown in Fig. 7, in the majority of the 12 independently transformed A. thaliana lines, the average expression level of CtFAD2-1 was significantly higher than CtFAD2-1Δ in developing A. thaliana siliques.

Fig. 7
figure 7

Real-time RT-qPCR analysis of CtFAD2‐1 and its mutant form CtFAD2‐1Δ expressed in A. thaliana embryos, controlled by a seed-specific Fp1 promoter. Twelve transgenic lines were renumbered according to their expression levels of CtFAD2-1. CtFAD2‐1 (filled circle) and CtFAD2‐1Δ (open circle)

Full size image

Perfect PCR markers for ol allele controlling the HO trait

The sequence polymorphism between CtFAD2-1 and CtFAD2-1Δ was exploited to develop a highly efficient molecular marker for tracking the HO mutant ol allele. The coding regions of CtFAD2-1 and CtFAD2-1Δ share 98 % nucleotide sequence identity. For CtFAD2-1Δ, in agreement with previous finding by Guan et al. (2012), the deletion of a single base pair “C” was identified at the position 606 bp downstream of the ATG initiation codon. This deletion causes a shift in the open reading frame that creates a stop codon and generates a putatively truncated protein of 245 amino acids in length. In addition, there are 22 nucleotide substitutions in S-317 CtFAD2-1Δ compared to SU CtFAD2-1. This is different from the earlier observation by Guan et al. (2012) that the single base pair “C” deletion was the sole sequence variation between CtFAD2-1 and CtFAD2-1Δ. Interestingly, the nucleotide substitutions seem not to be randomly distributed within the 1,142-bp long putative coding region of CtFAD2-1Δ in that 13 nt of the 22 (63.6 %) substitutions occur within 123 bp downstream of the deletion site. Nevertheless, it is apparent that the single-nucleotide polymorphisms (SNPs) revealed in the coding region of CtFAD2-1 and CtFAD2-1Δ may not be sufficient to design PCR-based molecular markers. However, the DNA sequence of a large intron situated in the 5′ UTR of CtFAD2-1 and CtFAD2-1Δ, which is 1,083-bp long in CtFAD2-1, and 1,144-bp long in CtFAD2-1Δ, was found to be highly divergent (Fig. 8). The two intron sequences share about 76.8 % sequence identity. Several short stretches of highly variable sequences between HL and HO genotypes enabled the designing of PCR primers specific for each of the two genes (Fig. 8). The gene-specific PCR products are shown in Fig. 9. The amplicon of the CtFAD2-1Δ intron was a 315-bp fragment that was present in all the five HO varieties including S-317, S-517, CW99-OL, LeSaf496 and Ciano-OL, while it was absent in the conventional HL safflower SU. The amplicon of the CtFAD2-1 intron, 603 bp in length, was present in the conventional HL safflower SU, while absent in all five HO varieties tested. A 198-bp band derived from the CtKASII gene was present in all the safflower varieties tested, acting as a positive control to distinguish between PCR failures and absence of a target gene-specific band.

Fig. 8
figure 8

Alignment of DNA sequences of the 5′ UTR intron of CtFAD2-1 derived from HL safflower SU (GenBank accession number KC886425) and CtFAD2-1Δ derived from HO genotype (GenBank accession number KC886424). Oligo primers designated for the amplification of genotype-specific fragment is boxed and orientation of the primer is indicated by arrow

Full size image
Fig. 9
figure 9

Development of a perfect DNA marker for ol allele safflower varieties. A ~300-bp band is distinctive in all five high oleic genotypes including S-317, S-517, CW99-OL, Ciano-OL and LeSaf496, while it is absent in the wild-type SU. A 600-bp fragment is specific for wild-type SU. A 198-bp band derived from CtKASII gene is present in all the lines tested and is used as a positive control

Full size image

Discussion

The original HO trait with oleic acid content of up to 75 % of total fatty acids in safflower seed oil is controlled by a partially recessive allele ol at a single locus OL (Knowles and Bill 1964). The ol allele has been commercially exploited and new safflower oil with elevated oleic acid content is widely available for food and industrial uses. In the current study, we have explored the biochemical and molecular features of the ol allele and established perfect molecular markers for tracking the ol allele in both homozygous and heterozygous states.

The biochemical analysis of oleate desaturase activity as determined by TLC and autoradiography analyses clearly indicated that the microsomal oleate desaturase enzyme CtFAD2 was defective in S-317. Among the numerous FAD2 enzymes, CtFAD2-1 was expressed strongly and exclusively in the developing embryos of the conventional HL safflower and, therefore, it likely plays an essential role in the biosynthesis of linoleic acid in developing seeds (Guan et al. 2012; Cao et al. 2013). This is direct evidence that CtFAD2-1 is the site of the ol allele as proposed by Hamdan et al. (2012). When mutated, the CtFAD2-1Δ disrupts the biosynthesis of linoleic acid, leading to a buildup of oleic acid in the seeds, as has been shown in various other oilseed crops such as sunflower (Helianthus annuus L.), peanut (Arachis hypogaea L.) and soybean (Glycine max L.) (Falentin et al. 2007; Lacombe et al. 2009; Perez-Vich et al. 2002; Pham et al. 2010; Schuppert et al. 2006).

It is noticeable that some residual linoleic acid still exists in the seed oil of the olol genotype, despite CtFAD2-1, which makes a major contribution to the production of linoleic acid, being defective. It is likely that genetic redundancy amongst the CtFAD2 genes, such as CtFAD2-2 and/or some other CtFAD2 genes, contributes to the accumulation of the residual linoleic acid in S-317. The chloroplast-localised oleate desaturase FAD6 could also be a contributor to the residual linoleic acid. However, its role is likely minor, if any, as safflower embryo is white and the number of mature and fully functional chloroplasts in developing embryos are relatively low (Martínez-Rivas et al. 2001).

Additional germplasm with even higher level of oleic acid than the olol genotype has also been reported (Fernández-Martinez et al. 1993; Mündel and Bergman 2009). Oleic acid content up to 89 % in safflower seed oil was reported by Fernandez-Martinez et al. (1993) in the germplasm accession PI401472 originally sourced from Bangladesh. The Montola series developed by safflower breeders in the USA (Mündel and Bergman 2009) contains more than 80 % oleic acid, clearly beyond the upper level of oleic acid of the olol genotype. An EMS mutant S901 containing up to 90 % oleic acid in its seed oil has also been obtained (Weiske 1999). Crossing and segregation analysis of the HO and very high oleic lines suggested that the very high oleic acid content is generated by the combination of ol allele and modifying gene(s) with a small positive effect on oleic acid (Hamdan et al. 2009, 2012).

The coding regions of CtFAD2-1Δ differed from that of the conventional HL safflower CtFAD2-1 in one nucleotide deletion and 22 nucleotide substitutions in S-317 CtFAD2-1Δ. The single-nucleotide deletion in the middle of the coding region in CtFAD2-1Δ caused a shift in the ORF that created a PTC and a non-functional microsomal oleate desaturase. A single base pair insertion or deletion (Indel) in the coding region of FAD2 leads to the formation of PTC in several oilseed crops. For example, in G. max, a single-nucleotide deletion 232 bases downstream of the initiation codon ATG of GmFAD2-1a was found in an HO mutant resulting from X-irradiation (Anai et al. 2008). This deletion resulted in a frame-shift mutation and led to increased oleic acid content from 27 % in wild type to 47 % in the HO mutant. A PTC was created by EMS mutagenesis in the seed-expressed FAD2 gene in Camelina sativa, which caused a moderate increase of oleic acid from 17 % in the wild type to 27 % in the mutant (Kang et al. 2011). In allotetraploid (AABB) A. hypogaea, the homoeologous FAD2-A and FAD2-B sequences share a high degree (99 %) of identity with a few polymorphic SNPs among normal and HO lines. In the HO variety with 80 % oleic acid, a missense mutation (G448A) was identified in FAD2-A and a single base pair insertion (442insA) in FAD2-B producing a frame shift and downstream PTC (Lopez et al. 2000). Both of these mutations (G448A and 442insA) are necessary to produce an HO peanut phenotype (Jung et al. 2000).

Interestingly, the expression of CtFAD2-1Δ in HO genotypes was drastically reduced in developing seeds, compared to the conventional HL safflower indicating specific down-regulation of CtFAD2-1Δ operating in the HO genotypes (Guan et al. 2012). In eukaryotes, gene silencing involves 21 and 24-nt siRNA produced from double-strand RNA resulting from transcription of antisense or hairpin RNA (Brodersen and Voinnet 2006). We carried out Northern blot analysis of small RNAs that indicated the absence of CtFAD2-1Δ siRNA in HO developing embryos. This was also verified by deep sequencing of small RNAs isolated from developing embryos. Only a negligible amount of small RNA corresponding to CtFAD2-1Δ could be detected. This was also the case for CtFAD2-1 in the conventional HL safflower. Further, previous research indicated that the ol allele is semi-recessive, which is distinct from the small interference RNA (siRNA)-mediated gene silencing system in which the HO trait would act as the dominant phenotype as is the case in the HO H. annuus genotype Pervenets (Lacombe et al. 2009).

We therefore propose that the specific attenuation of CtFAD2-1 transcripts in the olol genotypes is governed by nonsense-mediated mRNA degradation (NMD), an mRNA surveillance pathway involved in the degradation of aberrant mRNAs that contain a PTC resulting from mutation, transcriptional errors or alternative splicing. NMD protects eukaryotic cells from potentially harmful effects of truncated proteins that might have dominant negative effects on functional homologues. While only sporadically reported in plants, NMD is universally present in eukaryotes and has been extensively studied in yeast and mammals (Culbertson 1999; Conti and Izaurralde 2005; Trcek et al. 2013). One of the most defining features of NMD substrates is a PTC that gives rise to a long 3′ UTR that is sensed in a translation-dependent manner by the evolutionarily conserved RNA helicase UPF1 (Hogg and Goff 2010). In yeast, the presence of a downstream instability element relative to the PTC was found to be required for NMD (Culbertson 1999). In mammals, splicing of at least one intron is required for NMD and occurs only if the PTC is more than 50 nucleotides upstream of the last intron (Nagy and Maquat 1998). In plants, several studies have shown that PTC-harbouring genes both with and without an intron can trigger NMD. The NMD phenomenon was initially reported in the G. max Kunitz trypsin inhibitor gene (Kti3), phytohemagglutinin gene (PHA) from common bean (Phaseolus vulgaris L.) (Jofuku et al. 1989; Voelker et al. 1990) and pea (Pisum sativa L.) ferredoxin gene (FED1) (Dickey et al. 1994), none of which contain an intron. In contrast, the splicing of an intron upstream of the PTC in the rice (Oryza sativa L.) waxy gene affected the efficiency of NMD (Isshiki et al. 2001). CtFAD2-1Δ contains a reasonably large intron of approximately 1.1 kb in length in the 5′ UTR that is more than 600 bp upstream of the PTC. We were able to reproduce the NMD phenomenon by expressing the coding region of the CtFAD2-1Δ without its 5′ UTR intron in both N. benthamiana leaves and transgenic A. thaliana seeds. This suggests that an intron located at 5′ UTR is not required in the activation and maintenance of NMD in HO safflower. However, the expression of the intron-less CtFAD2-1Δ in yeast failed to generate NMD, indicating that a mere PTC present in the CtFAD2-1Δ coding region was not sufficient to induce NMD in transgenic yeast. The fact that the CtFAD2-1Δ transcript containing a PTC was able to be detected in the HO varieties of safflower indicates that not all these defective transcripts have been subjected to NMD degradation. This is conceivable since NMD is not 100 % efficient, but generally reduces the abundance of nonsense-containing mRNAs to ~5–25 % of the normal level (Isken and Maquat 2007). Whether an RNA molecule will be degraded or not is determined by competition between cytoplasmic poly(A)-binding protein 1 and UPF1 for binding to translation release factors at the terminating ribosome (Silva et al. 2008). This interaction may be modulated by numerous structural features within the mRNA. In mammals, the presence of splicing boundaries downstream of a stop codon acts as a strong enhancer of NMD (Nicholson et al. 2010). In plants, NMD is activated by either a long 3′ UTR or by a premature termination codon in close proximity to an exon junction (Kertesz et al. 2006). By introducing PTCs at various positions in the trypsin proteinase inhibitor (TPI) gene of Nicotiana attenuate, it was observed that NMD efficiency was correlated with PTC location (Wu et al. 2007).

We have proposed that NMD is the molecular basis of the high oleic genotype of olol mutant based on the following observations. Firstly, as the result of single-nucleotide deletion in the middle of its coding region, CtFAD2-1Δ contains an extra long 3′ UTR that is the prerequisite for the NMD phenomenon. Secondly, we have ruled out the possibility of siRNA or microRNA-mediated RNA degradation because of the lack of high-level presence of siRNA corresponding to CtFAD2-1Δ. Thirdly, we have successfully re-produced the NMD phenomenon in N. benthamiana leaves using an A. tumefaciens infiltration transient assay of CtFAD2-1Δ. Transgenic A. thaliana plants expressing the PTC-containing CtFAD2-1Δ without the 5′ UTR intron in a seed-specific manner also showed more highly attenuated mRNA levels than did the plants expressing the wild-type CtFAD2-1 gene.

Conventional breeding selects HO genotypes based on the analysis of fatty acid composition on safflower half seeds. This phenotype-based method is time consuming and environmentally sensitive as temperature influences the FAD2 oleate desaturase activity. Further, since the ol allele controlling the HO trait is partially recessive, the heterozygote is not clearly distinguishable from the wild-type homozygote. This fact makes it difficult to select plants carrying the ol allele in backcrossing programmes focusing on the HO trait. Therefore, the use of molecular marker-assisted selection can overcome such a limitation and greatly support the introgression of the ol allele into elite safflower lines. Recently, an SSR-based molecular marker for the OL locus has been described (Hamdan et al. 2012). In the current study, we have developed a dominant Indel marker that is not only diagnostic for the OL locus, but also able to resolve the Ol and ol alleles.

Using a pair of oligo primers that are based on the specific regions of the CtFAD2-1Δ intron sequence that lacks homology with the conventional HL safflower CtFAD2-1 intron, the PCR amplification products are produced only for genotypes carrying the ol allele, in either homozygous or heterozygous states. Such a perfect molecular marker is completely linked to the HO mutation and is able to detect the ol allele regardless of genetic background. Because such a molecular marker is dominant, it does not allow the distinction between homozygous and heterozygous genotypes at the OL locus. We have subsequently used the HL-specific PCR marker in both wild-type and heterozygous HO lines. It is anticipated that both HO-specific and HL-specific markers would be present in the heterozygous (Olol) plants simultaneously. This combination of both Ol and ol molecular markers can be useful in breeding programmes, as they will allow for rapid screening and early detection of not only the genotypes carrying the ol allele, but also the progenies with olol homozygotes.

In conclusion, this research reveals the molecular mechanism of the HO trait generated by the olol mutation originally discovered by Knowles and Bill (1964). The combination of in vitro microsomal FAD2 enzyme assays and gene expression studies indicated the seed-specific CtFAD2-1 as the OL locus. A single-nucleotide deletion in the coding region of CtFAD2-1Δ as previously described (Guan et al. 2012) possibly triggers an NMD phenomenon leading to highly attenuated transcript levels of CtFAD2-1Δ in the HO varieties carrying the olol mutation. We have explored the NMD mechanism by ectopic expression in yeast, N. benthamiana leaves and A. thaliana seeds. We have developed molecular marker assays that allow for the selection of the desired ol allele, differentiating its presence in either homozygous or heterozygous states. It is anticipated that the development of such a perfect molecular marker will make it possible for safflower breeders to quickly incorporate the HO trait in their breeding programme.