Omega-3 fatty acid desaturase gene family from two ω-3 sources, Salvia hispanica and Perilla frutescens: Cloning, characterization and expression

Omega-3 fatty acid desaturase (ω-3 FAD, D15D) is a key enzyme for α-linolenic acid (ALA) biosynthesis. Both chia (Salvia hispanica) and perilla (Perilla frutescens) contain high levels of ALA in seeds. In this study, the ω-3 FAD gene family was systematically and comparatively cloned from chia and perilla. Perilla FAD3, FAD7, FAD8 and chia FAD7 are encoded by single-copy (but heterozygous) genes, while chia FAD3 is encoded by 2 distinct genes. Only 1 chia FAD8 sequence was isolated. In these genes, there are 1 to 6 transcription start sites, 1 to 8 poly(A) tailing sites, and 7 introns. The 5’UTRs of PfFAD8a/b contain 1 to 2 purine-stretches and 2 pyrimidine-stretches. An alternative splice variant of ShFAD7a/b comprises a 5’UTR intron. Their encoded proteins harbor an FA_desaturase conserved domain together with 4 trans-membrane helices and 3 histidine boxes. Phylogenetic analysis validated their identity of dicot microsomal or plastidial ω-3 FAD proteins, and revealed some important evolutionary features of plant ω-3 FAD genes such as convergent evolution across different phylums, single-copy status in algae, and duplication events in certain taxa. The qRT-PCR assay showed that the ω-3 FAD genes of two species were expressed at different levels in various organs, and they also responded to multiple stress treatments. The functionality of the ShFAD3 and PfFAD3 enzymes was confirmed by yeast expression. The systemic molecular and functional features of the ω-3 FAD gene family from chia and perilla revealed in this study will facilitate their use in future studies on genetic improvement of ALA traits in oilseed crops.


Introduction
In the family Lamiaceae, chia (Salvia hispanica, 2n = 12) and perilla (Perilla frutescens, 2n = 40), which are annual herbaceous plants, are 2 rich sources of ω-3 polyunsaturated fatty acids (PUFAs). Chia is native to Mexico and parts of South America, and perilla originated in Asia [1][2][3][4]. Chia and perilla seeds contain 25~40% oil, and ɑ-linolenic acid (ALA, 18:3 Δ9, 12,15 ) proportion in their seed oil is the highest among crop sources (about 60~71%) [5][6][7]. ALA is known as an essential fatty acid (FA) for the human daily diet because ALA cannot be synthesized in the human body due to the absence of the ω-3 FAD gene [8]. ALA has a wide variety of health benefits as it is a necessary substrate for the biosynthesis of very-long-chain ω-3 PUFAs, eicosapentaenoic acid (C20:5 Δ5, 8,11,14,17 , EPA) and docosahexaenoic acid (C22:6 Δ4, 7,10,13,16,19 , DHA). It has been reported that EPA and DHA regulate body development and growth, promote brain development, reduce blood pressure, inhibit senescence, and also have beneficial effects on neurological, cardiovascular and cerebrovascular diseases [9]. Moreover, ALA functions as a crucial component of membrane lipids and triacylglycerol seed storage lipids in higher plants [10]. Additionally, ALA is also a precursor of FA-derived signal molecules, e.g., jasmonic acid (JA), that play important roles in plant development and stress responses [11]. With respect to the ALA biosynthesis pathway, a small portion of pivotal genes has been described in perilla [12][13][14][15], but genetic and molecular studies are relatively rare for chia [16]. Hence, systemic cloning of the ω-3 FAD gene family of both chia and perilla was performed in this study, which is the first report to provide the full-length cDNA and genomic DNA (gDNA) sequences of chia ω-3 FAD genes, as well as the full-length gDNA sequences of perilla ω-3 FAD genes.
FAD3 cDNA was isolated from the developing seeds of perilla, and its mRNA accumulation manner was seed-specific [12]. Recently, the expression profiles of ω-3 FAD genes in developing seeds of perilla by transcriptome analysis have been investigated, which showed that FAD3 and FAD7/8 were determined to be pivotal genes for ALA synthesis in seeds and leaves, respectively [14]. Additionally, the subcellular location and catalytic activity of perilla FAD3 and FAD7-1/-2 were confirmed [13]. For chia, transcriptome profiles and expression analysis of Δ15/ω-3 desaturase genes in 5 different stages of developing seeds were reported [4]. Unfortunately, systemic cloning and characterization of chia ω-3 FAD genes as well as comparative study on the evolutionary relationship, exon/intron patterns, stress responses, and FAD3 catalytic activity between perilla and chia ω-3 FAD genes have not so far been conducted. The aim of this paper is to systematically clone and identify full-length sequences of the ω-3 FAD gene family from chia and perilla. This study can provide an important reference for dissecting the molecular mechanisms of their high ALA traits, and enrich our knowledge of the crucial roles of ω-3 FAD genes in response to abiotic/biotic stresses and plant hormone treatments.

Plant materials, treatments, and nucleic acid extraction
Chia (commercial variety) and perilla (C2 cultivar) were grown in a standard experimental field of College of Agronomy and Biotechnology, Southwest University, China. For each species, roots (Ro), stems (St), leaves (Le), buds (Bu), flowers (Fl), mid-stage seeds (MS, approximately 20 days after flowering), and late-stage seeds (LS, approximately 30 days after flowering) were sampled, and early-stage seeds (ES, approximately 10 days after flowering) were also collected for chia.
For various treatments, chia and perilla seeds were grown in damp soil in plastic pots (20 seeds per pot). These pots were placed in a climatic chamber (30˚C and 56% of relative humidity) with 16/8 h of light/darkness. The 5-week-old seedling leaves of chia and perilla were subjected to various stresses. Salt and drought stresses were imitated using 300 mM NaCl and 10% PEG6000 solutions (plants were irrigated), respectively. The seedlings were sprayed with 100 μM ABA, 1 mM SA, and 100 μM MeJA solutions. Cold and heat treatments were carried out in 4˚C and 38˚C (chia) or 42˚C (perilla) chambers, respectively. Wound stress was performed as described in previous report [57]. For each stress, seedling leaves were collected at 0, 0.5, 3, 9, 24, and 48 h after treatment. All collected samples were immediately frozen in liquid nitrogen and stored at -80˚C.
Total DNA was extracted from the leaves of 2 species using the CTAB method [58]. Total RNA was extracted from differential organs or seedling leaves of 2 species subjected to various stresses using the RNAprep Pure Plant Kit (Tiangen, China). The quality and concentration of genomic DNA and total RNA were detected by agarose gel electrophoresis and spectrophotometer analysis with a Nanodrop 2000 (Thermo Fisher Scientific, USA).

Cloning of full-length sequences of ω-3 FAD gene family from 2 species
For both chia and perilla, 1 μg of an equally proportioned (w/w) mixture of total RNA from various organs was employed to generate first-strand total cDNA of 5'-and 3'-RACE, respectively, using the SMARTer™ RACE Amplification Kit (Clontech, Takara Dalian, China). Based on multi-alignment of the ω-3 FAD cDNAs from perilla, flax, olive and other plants, four gene-specific primers were designed to correspond to the conserved sites (S1 Table). Sense primers FPD153-1 and FPD153-2 were paired with the kit universal primers UPM and NUP, respectively, for primary and nested amplifications of 3'-RACE. With respect to primary and nested amplifications of 5'-RACE, the kit universal primers UPM and NUP were paired with antisense primers RPD155-1 and RPD155-2, respectively. In the primary amplification, 0.2 μL of first-strand total cDNA of 5'-or 3'-RACE was used as a template, whereas 0.1 μL of 5'-or 3'-RACE primary amplification product was employed as a template for the nested amplification. All 4 PCR programs adopted the following cycling parameters: 94˚C for 2 min; 30 cycles of 94˚C for 1 min, 58˚C for 1 min, 72˚C for 1 min; and 72˚C for 10 min. The aforementioned PCR fragments were gel-recovered and cloned into a pGEM-T easy vector (Promega, USA) and sequenced; their identities were confirmed by NCBI BLASTn.
Based on the sequence alignment of Arabidopsis and perilla ω-3 FAD mRNAs from NCBI GenBank, as well as 5'-and 3'-RACE cDNAs of the ω-3 FAD genes from 2 species obtained in this study, allele/member-specific primers of the full-length ω-3 FAD genes in chia and perilla were designed (S1 Table). Next, 0.2 μL total cDNA of mixed organs from chia or perilla was used as a template for amplifying full-length cDNA sequences, and the corresponding genomic sequences were also isolated using 0.5 μg total genomic DNA as a template from the leaves. Primer pairs FPfFAD3+RPfFAD3a and FPfFAD3+RPfFAD3b were used to isolate perilla FAD3 (PfFAD3) alleles, FPfFAD7+RPfFAD7 for perilla FAD7 (PfFAD7) and FPfFAD8+RPfFAD8 for perilla FAD8 (PfFAD8). Primer pairs FShFAD3-1+RShFAD3-1 and FShFAD3-2+RShFAD3-2 were used to isolate 2 members of chia FAD3 family (ShFAD3-1 and ShFAD3-2), FShFAD7a+RPfFAD7a and FShFAD7b+RShFAD7b were used for 2 alleles of chia FAD7 (ShFAD7a and ShFAD7b), and FShFAD8+RShFAD8 was used for chia FAD8 (ShFAD8). The thermal cycling parameters of these PCR reactions were as follows: 94˚C for 2 min; 35 cycles of 94˚C for 1 min, 60˚C for 1 min, 72˚C for 3 min; and 72˚C for 10 min. Gel recovery, TA cloning, and sequencing of amplified genes were performed following general procedures.

Quantitative RT-PCR analysis
First-strand total cDNA was generated with 1 μg of each organ or seedling leaves of each stress treatment using the PrimeScript Reagent Kit with gDNA Eraser (Takara Dalian, China). Based on full-length cDNAs of chia and perilla ω-3 FAD genes, the corresponding primer pairs in S1 Table were designed for fluorescence real-time quantitative RT-PCR (qRT-PCR) detection of expression profiles of PfFAD3, PfFAD7, PfFAD8, ShFAD3-1, ShFAD3-2, ShFAD7 and ShFAD8. First, the specificity of the qRT-PCR primers was validated using agarose gel electrophoresis of their PCR products. Then, qRT-PCR was performed with a FastStart Universal SYBR Green Master (Roche, Germany) in a total reaction volume of 10 μL, which comprised 5 μL of 2×SYBER Mix, 0.5 μL of each primer (10 μM) and 2.5 μL of cDNA. Chia and perilla 25SrRNA were used as reference genes for qRT-PCR with primer pair F25SRT+R25SRT (S1 Table), and these primers were designed according to conserved regions of 25SrRNA across the plant kingdom. The reactions were performed on CFX96 Real-time PCR system (Bio-Rad, USA) according to the manufacturer's protocol, and a melting curve analysis was conducted to test whether additional qRT-PCR products were present. The qRT-PCR experiments were performed with 3 biological replicates. All the data were analyzed by using CFX Manager 3.1 (Bio-Rad, USA) with the 2 -ΔΔCT method [63].

Yeast expression and fatty acid analysis
The ORFs of ShFAD3-1, ShFAD3-2, and PfFAD3a/b were amplified using primer pairs FShFAD3-1Y+RShFAD3-1Y, FShFAD3-2Y+RShFAD3-2Y, and FPfFAD3Y+RPfFAD3Y, respectively (S1 Table). Each of the 4 ORFs was cloned into the pGEM-T easy vector (Promega) and validated by sequencing. Then, these 4 coding regions containing BamHI and XbaI sites were individually inserted into pYES2.0 (Invitrogen, USA) via double digestion and were confirmed by sequencing. The empty vector pYES2.0, and the recombinant plasmids pYES2-ShFAD3-1Y, pYES2-ShFAD3-2Y, pYES2-PfFAD3aY and pYES2-PfFAD3bY were transformed into Saccharomyces cerevisiae strain INVScl as described by the pYES2.0 Kit User Manual. Yeast cells were grown to logarithmic phase at 30˚C in SC-Ura containing 2% (w/v) raffinose and 0.1% NP-40 using 0.5 mM LA as a feeding substrate. After adding 2% (w/v) galactose, the yeast cells were inducibly expressed and then incubated at 20˚C for 72 h. Finally, the yeast cells were collected by centrifugation, washed with sterilized water more than 3 times, and freeze-dried on a ScanVac-Coolsafe 110-4 (Denmark) for 12 h. Separation and gas chromatography (GC) analysis of FA compositions of yeast cell samples were carried out as described in a previous report [64]. Each experiment was carried out in 3 biological replicates.

Cloning of full-length ω-3 FAD gene sequences from chia and perilla
We isolated 2 heterozygous alleles containing some SNPs for PfFAD3, PfFAD7 and PfFAD8 from perilla, whereas 2 distinct ShFAD3 genes, 2 heterozygous ShFAD7 alleles, and 1 ShFAD8 gene were cloned from chia (S1 Fig; Table 1). Full-length cDNAs of the ω-3 FAD genes from the 2 species, except for PfFAD8b (5'UTR and partial CDS included), were obtained with the longest mRNAs of 1,435~1,957 bp, 5'UTRs of 34~380 bp, ORFs of 1,152~1,323 bp, and 3'UTRs of 210~325 bp (Table 1). Except for PfFAD3a and PfFAD7b, the other ω-3 FAD members/alleles from the 2 species contained 2 to 6 alternative transcription start sites. PfFAD3a/b, ShFAD3-1/-2, ShFAD7a/b, and ShFAD8 harbored 2 to 8 alternative poly(A) tailing sites, and PfFAD7a/b and PfFAD8a had 1; 1 to 2 typical and non-typical polyadenylation signals were present in these genes. Corresponding genomic DNAs were amplified using the total genomic DNA of leaves from the 2 species as a template, but amplification of genomic sequences for PfFAD3b and PfFAD7b failed. As shown in  Table). In the PfFAD8a 5'UTR, there were 2 purine-stretches (23 bp and 24 bp) and 2 pyrimidine-stretches (21 bp and 31 bp), and PfFAD8b harbored a purine-stretch (24 bp) as well as 2 pyrimidine-stretches (25 bp and 27 bp) in its 5'UTR (S1 Fig). Pairwise alignment of 5'RACE cDNAs and genomic sequences of ShFAD7a/b revealed that a small intron (148 bp) was present in the 5'UTR of alternative splicing variants ShFAD7a' and ShFAD7b', and it was 15 bp upstream of the start codon ATG, with standard GT. . .AG splicing boundaries (Fig 1; S1 Fig). However, the full-length cDNAs of ShFAD7a' and ShFAD7b' could not be isolated, which might be caused by the low abundance of these 2 splicing variants.
The tree also reveals some important evolution clues of plant ω-3 FADs. The first is the convergent evolution of FAD7/FAD8, i.e. multiple independent origination of plastidial ω-3 FADs during plant divergence process. FAD7s/FAD8s across different phylums do not cluster together to form a plant FAD7s/FAD8s large group. Rather, in most cases FAD7s/FAD8s were originated from FAD3s within respective phylums. For example, the FAD7s/FAD8s from dicot, monocot, and basal angiosperm plants form a large group, which is diverged from angiosperm FAD3s. On the other hand, this implies that dicot and monocot FAD7s/FAD8s share a common basal angiosperm FAD7 ancestor. The second is the amplification of gene numbers of ω-3 FADs in non-algae plants. Many plants contain more than 1 copy of FAD3 and/or FAD7/FAD8, especially in Physcomitrella patens, sunflower, flax, Arachis duranensis, and Musa acuminata subsp. malaccensis. The third is noticeable evolution features of ω-3 FADs in gymnosperms. Picea abies and Pinus taeda both contain only 1 plastidial ω-3 FAD and no ER-type FAD, while Ginkgo biloba contain both types. Besides, the 2 genes newly duplicated from ginkgo FAD3 contain a chloroplast transit peptide, indicating their directional evolution of subcellular localization after recent gene duplication events.

Organ-specificity of ω-3 FAD gene family from chia and perilla
To shed light on the biological function of the ω-3 FAD genes from perilla and chia, their organ-specificity patterns were investigated. In this study, we firstly validated the specificity  Table. The ω-3 FAD proteins from chia and perilla are indicated by the red squares and red triangles, respectively. of the qRT-PCR primers using agarose gel electrophoresis of their PCR products, which all showed a predicted size of specific band for each ω-3 FAD mRNA, without cross amplification using templates of other genes. Then, qRT-PCR was carried out to examine the expression patterns of chia and perilla ω-3 FAD genes in various organs. The results showed that ω-3 FAD genes from perrila and chia were expressed in various organs, but with different levels (Fig 4A-4G). PfFAD3 was mainly expressed in late-stage seeds (approximately 4000-fold compared to the roots). The expression level of ShFAD3-1 was relatively higher in stems and early-stage seeds than in other organs, while ShFAD3-2 transcripts mainly accumulated in early-stage seeds. The PfFAD7 expression levels in leaves, buds, and flowers were higher than in other organs. A smaller amount of PfFAD8 mRNA was accumulated in the roots and stems compared to other tissues. ShFAD7 was transcribed more in stems and flowers, less in middle-stage and late-stage seeds, and moderately in other organs. Except for middle-stage and late-stage seeds, ShFAD8 was expressed at relatively high levels in other tissues. These levels were relatively lower than the levels obtained in ShFAD7. From this expression analysis, we found that FAD7 genes were expressed at higher levels than FAD8 in both chia and perilla, which was also similar to the outcomes for other species, e.g., D. sophia [45] and purslane [28].

Expression profiles of ω-3 FAD gene family from chia and perilla under various stress treatments
To examine the relationship between chia and perilla ω-3 FAD genes under various stresses, real-time qRT-PCR was performed to illustrate the expression profiles of the ω-3 FAD genes from the 2 species in seedling leaves under the biotic stress of wounding and abiotic stresses, including cold, heat, drought (PEG6000), salt (NaCl), and treatments with plant hormones including MeJA, ABA, and SA. As shown in Fig 5A-5H, each ω-3 FAD gene responded to multiple treatments. Under cold treatment, PfFAD7 expression was rapidly and transiently up-regulated to 20-fold at 0.5 h, then dropped down and fluctuated, but still keeping a significant up-regulation level. PfFAD8 was weaker in response to cold stress than PfFAD7 (Fig 5A). However, both ShFAD7 and ShFAD8 showed very limited or non-distinct up-regulation. Interestingly, unlike PfFAD3 which showed only a little up-regulation by cold, ShFAD3-1/2 was steadily up-regulated by cold, though not as quick as PfFAD7 and PfFAD8 (Fig 5A). The PfFAD3 transcript was quickly reduced to the lowest level for the whole term of heat treatment (Fig 5B). ShFAD3-1 and PfFAD7/8 expression were up-regulated a short time after heat treatment, but their transcripts were inhibited over the long term. ShFAD8 expression transiently peaked at 0.5 h, but quickly dropped down to constant levels ( Fig 5B). ShFAD3-2 expression was least sensitive to cold treatment, with only a little down-regulation at 24 h and 48 h. Under drought stress (PEG treatment), ShFAD3-1/-2 expression firstly increased and then decreased, whereas PfFAD3 showed fluctuation with rounds of up-regulation and falling back. Generally, FAD7 and FAD8 genes from both perilla and chia were down-regulated by PEG treatment, though PfFAD8 was less sensitive (Fig 5C). Under NaCl treatment, all ω-3 FAD genes from perilla and chia showed down-regulation with trends similar to those in PEG treatment, though PfFAD7 and PfFAD8 were transiently up-regulated at 0.5 h and PfFAD8 was less sensitive than others (Fig 5D). Wounding stress inhibited the expression of all chia ω-3 FAD genes, but it enhanced PfFAD7 transcripts (Fig 5E). After wounding stress, PfFAD8 showed fluctuation with rounds of up-regulation and falling down, while PfFAD3 expression was first down-regulated and then up-regulated in long term (Fig 5E). After MeJA and SA treatments, all perilla and chia ω-3 FAD genes showed similar dynamics, i.e. transient up-regulation followed by declining and re-up-regulation at 48 h, but SA stimulation (peaked at 0.5 h) was quicker than MeJA (peaked at 3 h) (Fig 5F/5H). Under ABA treatment, the expression of all perilla and chia ω-3 FAD genes was inhibited, but 48 h after treatment most of them resumed to be around basal levels except ShFAD7/8 which still kept inhibition status (Fig 5G). These results suggest that chia and perilla ω-3 FAD genes all are responsive to various stresses and might play some roles in coping with adversities, but inter-specific differences between chia and perilla as well as inter-genic differences especially between FAD3 and FAD7/8 are distinct.

Catalytic activity identification of ω-3 FADs in chia and perilla using yeast expression
Yeast has been shown to be an ideal model system for identifying the function of ER-located desaturases, including FAD2 and FAD3 [17,71], but it was not suitable for heterologous expression of plastidial desaturases (e.g. FAD6/7/8) due to their requirements for electron transport chains from the chloroplast [16]. To determine the function of PfFAD3a, PfFAD3b, ShFAD3-1 and ShFAD3-2, corresponding ORFs were individually cloned into the expression vector pYES2.0 under an inducible GAL1 promoter and transformed into S. cerevisiae. The GC analysis of FA compositions in transformed yeast strains showed a high content of LA that is absent in wild-type yeast, which confirmed the correct uptake of supplemented substrate. As shown in Fig 6, ALA was not present in yeast cells transformed with empty vector pYES2.0, but ALA production was detected in yeast cells transformed with pYES2-ShFAD3-1Y, pYES2-ShFAD3-2Y, pYES2-PfFAD3aY, and pYES2-PfFAD3bY. The percentage of ALA in transgenic yeast cells was 8.84~16.91% of the total FA, and the conversion of LA to ALA was 8.84~16.91% (Table 3); the desaturation ratio of PfFAD3a/b was not as high as that of ShFAD3-1/-2. This result showed that yeast cells overexpressing PfFAD3a, PfFAD3b, ShFAD3-1, and ShFAD3-2 performed the desaturation of LA to ALA, implying that these four FAD3 genes all encode a functional linoleate Δ-15 desaturase. Unfortunately, ALA production was not detected in yeast cells transformed with recombinant vectors pYES2-PfFAD7--T2A-BnFD2, pYES2-PfFAD8-T2A-BnFD2, pYES2-ShFAD7-T2A-BnFD2, and pYES2-Sh-FAD8-T2A-BnFD2, although various conditions had already been optimized according to current theories and this experiment was repeated for many times.

Evolutionary features of ω-3 FAD genes from chia, perilla, and plant kingdom
In this study, we have systematically isolated and characterized the ω-3 FAD gene family from chia and perilla. Chia ShFAD3 contains 2 member genes, including ShFAD3-1 and ShFAD3-2, and a single ShFAD8 gene was cloned from chia, while ShFAD7 and perilla PfFAD3/7/8 are also single-gene loci possessing 2 heterozygous allele sequences. For all chia ω-3 FAD genes, i.e. ShFAD3-1/2, ShFAD7a/b, and ShFAD8, we have obtained their both full-length cDNAs and corresponding gDNAs. For all perilla ω-3 FAD genes, i.e. PfFAD3a/b, PfFAD7a/b, and PfFAD8a/b, we isolated their full-length cDNAs (except PfFAD8b), but we failed to obtain the genomic sequences of PfFAD3b/7b/8b though we used allele-specific primers to screen numerous TA-colonies in gDNA cloning process. Nevertheless, unlike the distinct divergence between ShFAD3-1 and ShFAD3-2 on both nucleotide and protein levels, the nucleotide sequences of ShFAD7a, PfFAD3a, PfFAD7a, and PfFAD8a are extremely similar to ShFAD7b, PfFAD3b, PfFAD7b, and PfFAD8b, respectively, with encoded proteins completely identical to each other or differed by only 1 to 2 similar aa substitution. Until now, no completed genome data from chia and perilla could be used as an important reference. Though traditional literatures report that chia is a diploid/amphiploid while perilla is an amphidiploid [1,3], the intersister identities within each of PfFAD3a/b, PfFAD7a/b, and PfFAD8a/b pairs are as high as that of ShFAD7a/b pair. This result is likely that these 4 pairs are 4 heterozygous allelic pairs other than 8 independent genes, and perilla might be originated from amphidiploidization between 2 subspecies or 2 closely related species. As 1 parent of perilla is P. citriodora [3], the other parent should be sought among wild or cultivated subspecies or species with very close relationships to P. citriodora. A simply feasible way to identify this unknown parent is to clone and compare the whole set of PfFAD3a/b, PfFAD7a/b, and PfFAD8a/b pairs from the candidate subspecies or species.
Sequence similarity of mRNAs and proteins showed that ShFAD7a/b were more identical to PfFAD7a/b than ShFAD8, whereas ShFAD8 was more similar to PfFAD8a/b than ShFAD7a/ b. The evolutionary relationship showed that ShFAD7a/b was clustered with PfFAD7a/b, and ShFAD8 was clustered with PfFAD8a/b. These results suggested that, in Lamiaceae family, the gene duplication event leading to FAD7 and FAD8 was prior to the divergence between genus Salvia and genus Perilla, i.e. possibly the whole Lamiaceae family has evolved 2 plastidial ω-3 FAD genes, FAD7 and FAD8. Finally, though FAD7 and FAD8 from Lamiaceae are not respective real orthologues of FAD7 and FAD8 from Brassicaceae, since they are the results of respective duplications after divergence between order Lamiales and order Brassicales, we still prefer to name them as FAD7 and FAD8 other than as two FAD7 genes as reported previously [13]. There are 3 reasons. Firstly, on both nucleotide and protein levels, they differ from each other significantly. Secondly, in BLAST analyses, FAD7 and FAD8 from perilla and chia show a little deviation toward FAD7 and FAD8 from A. thaliana respectively. Thirdly, many previous reports already adopted this method to name the duplicated plastidial ω-3 FAD genes from non-Brassicales plants.
Though the evolution of front-end desaturases as a whole has been reported previously, to date there is no systemic study on the evolution of ω-3 FAD genes. In studying safflower ω-3 FAD genes, Guan et al. constructed a phylogenetic tree of ω-3 desaturases from different plants, but the tree was still not systemic enough and they did not deeply analyse the evolution rules [27]. On the other hand, sequencing, annotation, and releasing of more and more plant The full names of the FAs are described in Fig 6. The data are the mean ± SD from3 biological replicates; "-", no detection. https://doi.org/10.1371/journal.pone.0191432.t003 genomes make it possible in this research to thoroughly identify the evolutionary features of ω-3 desaturases from various plant taxa. All the 3 algae species own a single copy ω-3 desaturase gene, either ER-type or CP-type, while all non-algae plants contain more than 1 except for the conifer species P. abies and P. taeda. In microalga Chlamydomonas reinhardtii, the single FAD7 gene can impact both plastidic and extraplastidic membrane lipids [72], whether the single gene status of other algae and conifer plants have also evolved similar mechanisms deserve future study. On the contrary to these species, most plant species contain both ER-type and CP-type D15D genes, which means that divergently evolution and keeping of both types are necessary for most higher plants. However, our phylogenetic study distinctly indicates that plant FAD7/FAD8 genes are convergently evolved in respective phylums, i.e. origin of FAD7/ FAD8 genes through gene duplication events after the formation of individual phylums. Furthermore, recent duplication events of FAD3 or FAD7/FAD8 genes in certain lower taxa are very common, i.e. in P. patens, ginkgo, wild banana, flax, sunflower, etc.

Noticeable structural features of ω-3 FAD genes from chia and perilla
We found that exon/intron numbers, intron phases, and splicing boundaries between ω-3 FAD genes from chia, perilla, and Arabidopsis were highly conserved (Fig 1) [30,31,73], which indicates that they were derived from a common ancestral gene. Except for the partial PfFAD8b, the remaining ω-3 FADs from the two species all contain 3 histidine boxes that are essential for maintaining their catalytic activity [33], a conserved domain FA_desaturase (pfam PF00487), and 4 strong transmembrane helices, which are typical characteristics of plant membrane-bound desaturases [74,75]. Alternative splicing is an important regulatory mechanism for controlling gene expression at a post-transcriptional level, and intron retention is more prevalent in plants [76]. A pairwise alignment showed that the alternative splicing variants ShFAD7a'/b' contained 1 5'-UTR intron, which was consistent with soybean GmFAD8 [17], and both variants had intron retention, but the corresponding full-length cDNA sequences could not be isolated from chia leaves. This might result from the relative low abundance of ShFAD7a'/b' transcripts due to certain environmental stress factors or during different development stages of chia. Under cold treatment, ShFAD7a/b expression is more similar to GmFAD8-1 [17] compared the remaining ω-3 FAD genes in these 2 species, which is likely due to the existence of a 5'UTR intron. Additionally, as reported in previous studies [77,78], the 5'UTR intron also possessed promoter activity, which enhanced transcriptional expression of the target gene under various environmental factors. Therefore, it is necessary to carry out the isolation and characterization of these two ShFAD7a/b isoforms in a future study. Additionally, there were 1 to 2 purine-stretches and 2 pyrimidine-stretches in the 5'UTRs of PfFAD8a/b (S1 Fig), which suggests their possible roles in modulating the transcription and translation of PfFAD8 genes. Except for PfFAD8b, the remaining ω-3 FAD genes for 2 species have 1 to 2 canonical or non-canonical poly(A) signals in the 3'UTRs, which may play a crucial role in determining alternative poly(A) sites.

Functional identification of ω-3 FAD genes from chia and perilla
In this study, heterologous yeast expression confirms the catalytic activity of FAD3s of chia and perilla, i.e. they both encode a functional linoleate Δ-15 desaturase. The conversion ratio of ALA (8.84~10.61%) in yeast overexpressing PfFAD3a/b was higher than that of PfrFAD3 (3.89~6.00%) described in previous reports [13,79], which might be due to a modified Kozak sequence containing 6 adenine nucleotides (AAAAAAATG, S1 Table) [80] in this PfFAD3a/b yeast expression system. In addition, the higher conversion ratio of LA to ALA obtained for ShFAD3-1/-2 compared to PfFAD3a/b in transformed yeast suggest that ShFAD3-1/-2 contributes to the ALA content to a greater extent than PfFAD3a/b does. Additionally, previous reports indicated that both the removal of N-terminal chloroplast transmit peptide and the ferredoxin co-expression were necessary to increase heterologous expression activity of plant plastidial FAD7/8 desaturases in yeast [23,24]. Here, although chloroplast transmit peptides of both PfFAD7/8 and ShFAD7/8 were deleted and rapeseed ferredoxin BnFD2 was meanwhile co-expressed, catalytic activity of the conversion of LA to ALA was not detected in transgenic yeast cells harbouring PfFAD7/8 or ShFAD7/8. However, functionality of perilla PfrFAD7-1 and PfrFAD7-2 genes were confirmed using their coding regions despite no deletion of chloroplast transmit peptide in N-terminus [13]. Hence, there is a need to identify the function of FAD7/8 of chia and perilla in yeast, using the original open reading frame without deletion of chloroplast transmit peptide, in the future study. More importantly, future study on yeast expression of plants FAD7/FAD8 genes should further clarify the issues regarding to chloroplast transmit peptide and ferredoxin.

Transcriptional expression characteristics of ω-3 FAD genes from chia and perilla
In higher plants, trienoic acids (TAs), including ALA, are structural components of membrane lipids and seed storage lipids and function as precursors of signaling molecules, e.g., JA [11,81]. FA unsaturation in the cellular membrane plays crucial role in temperature stress and adaption [82]. The JA-signaling pathway also functions in plant growth and development, as well as defense responses [83]. In Arabidopsis, the formation of TAs was catalyzed by two types of ω-3 FADs: ER-located FAD3 and chloroplast-located FAD7/8 [30,31,72]. To date, it has been reported that ω-3 FAD genes in a wide variety of plants function in ALA biosynthesis in various organs/tissues and in response to various environmental stimuli. Organ-specificity expression showed that all members of chia and perilla ω-3 FAD genes were expressed in various organs, but there was divergence and complementation in their expression patterns. In perilla, PfFAD3 mRNA accumulation in late-stage seeds was the most abundant (approximately 4000-fold of root, Fig 1A), which was not consistent with the seedspecific expression of PfFAD3 [12] but imitated the expression pattern of PfrFAD3 in various organs [13], revealing its key roles of ALA biosynthesis in seeds. Similar to PfrFAD7-1/-2 [14,15], PfFAD7/8 showed higher expression levels in leaves than seeds, which indicated that they play preferential roles in ALA accumulation of vegetative organs. ShFAD3-1/-2 was mainly transcribed in early-stage seeds, whereas the difference between ShFAD3-1 and ShFAD3-2 could be attributed to higher transcripts of ShFAD3-1 in stems than that of ShFAD3-2, which indicated that there was a small functional partition between the two copies of ShFAD3. A previous study [4] indicated that ShΔ15 and Shω-3 were mainly expressed in the early stages of seed development, which are almost in agreement with transcriptional pattern of ShFAD3-2 and ShFAD8, respectively, and suggests that they both play an important role in ALA biosynthesis of early seed stages.
The current results indicate that in the chia seedling leaves, ER ω-3 FAD expression is tightly regulated under cold treatment, while no obvious change was detected for plastidial ω-3 FAD. This outcome is consistent with previous observations for soybean GmFAD3A, GmFAD7-1/-2 and GmFAD8-1/-2 [17]. In perilla, plastidial ω-3 FADs play more important roles in response to cold than ER ω-3 FADs. Heat treatment inhibits the PfFAD3 transcripts but first up-regulates and then down-regulates PfFAD7/8, which suggests that low levels of TAs are critical for the heat response and tolerance over the long-term. At 48 h after heat treatment, chia ShFAD3-1/-2 and ShFAD7 transcripts were reduced slightly, whereas ShFAD8 expression increased to a small extent, which indicates that ShFAD8 might respond to heat treatment at a higher temperature or over a longer period. This hypothesis needs to be tested in further study. The different ω-3 FAD gene expression patterns between chia and perilla under cold and heat suggest that there is diversity between the two species for response time and speed in these opposite stresses. The lima bean PlFAD3 transcript was induced by drought stress [42]; ω-3 FAD genes in the present study were also enhanced at different levels by drought. Salt stress suppressed chia ω-3 FAD and PfFAD3 gene expression, which was similar to lima bean PlFAD3 [42], but an up-regulated expression of PfFAD7/8 at 0.5 h was obtained, which implied that processes, such as enhanced membrane FA unsaturation, are essential for perilla response to salt stress.
Wounding normally exemplifies biotic stresses, e.g., insect feeding and herbivory. Multiple structurally distinct molecules function in wound signaling, including plant hormones (e.g., JA, ABA and ethylene), oligosaccharides, and oligopeptides [84]. It has been reported that ALA for JA biosynthesis is derived from ER (FAD3) and plastid (FAD7/8) membranes [85]. In this study, PfFAD3/7/8 expression was up-regulated by wound stress, which is consistent with the report on ω-3 FAD genes from D. sophia [45]. Wounds can activate the JA biosynthesis pathway and lead to an increase in JA accumulation by converting ALA to JA, which plays a critical role in the transcriptional regulation of wound-inducible genes [86,87]. However, wound treatment down-regulated the transcripts of chia ω-3 FAD genes. Previous reports showed that there is a complex wound signaling network in plants, which notably has speciesspecific variations [84]. Accordingly, the participation of any deduced signal in the activation of wound response depends on plant species, which indicates a different wound-induction mechanism for the ω-3 FAD genes in chia and perilla.
Plants have a set of defense mechanisms against microbial pathogen attacks in which plant hormones, e.g., JA, SA and ABA, play indispensable signaling roles [88]. In general, cooperative or antagonistic crosstalk between these hormones plays a pivotal role in maintaining the disease resistance [88]. Therefore, the induction of ω-3 FAD gene expression occurs under MeJA, SA and ABA treatment, which occurs due to ω-3 FAD products serving as precursors of JA biosynthesis. Under SA and MeJA treatment, ω-3 FAD genes from perilla and chia showed either up-regulated or down-regulated expression patterns, which indicates that they play important roles in the SA and JA signaling pathways. Moreover, ABA treatment suppressed the transcription of chia and perilla ω-3 FAD genes, which is almost consistent with CsFAD7 and CsFAD8 in the tea plant [49].
The different expression patterns observed between FAD3 and FAD7/8 under various stresses imply that there is an obvious divergence of ER-type and plastid-type ω-3 FAD genes in stress response and adaptation. This result also suggests that plants need to maintain membrane fluidity for each stress/treatment adaptation by modulating PUFAs (including ALA) compositions. In addition, the response variation between the two species may be associated with long-stage artificial domestication and the selection of cultivated species of chia and perilla in different places of origin, i.e., Mexico and Asia, respectively.
In conclusion, this is the first report to provide a systemic and comparative study of ω-3 FAD gene family from chia and perilla, which are two plant sources containing the most abundant ω-3 PUFAs (namely, ALA). In this study, we systemically isolated the ER-type FAD3 and chloroplast-type FAD7/8 genes from these two species and comparatively analyzed sequence characters, genomic organization, phylogenetic relationships, organ-specificity, stressinducibility and enzymatic activities. This work provides a basis for revealing the molecular mechanism of high ALA traits via the FA desaturation pathway and facilitates our understanding of the chia and perilla ω-3 FAD genes in response to multiple stresses. Besides, this study also reveals some important evolution features of plant ω-3 FAD genes. In further study, it is very important to carry out transgenic manipulation of chia and perilla ω-3 FAD genes for exploring ALA traits in oilseed crops because these are dedicated steps for ALA biosynthesis.
Supporting information S1 Table. Degenerate and non-degenerate primers used in this study. a N: A or G or C or T; V: A or G or C; D: G or A or T. b Both degenerated bases and the restriction sites that were introduced are underlined, and Kozak sequences [80] are italicized in bold face. c T2A sequence [65] is wave-lined, and overlapped regions between two primers are in bold face. All primers were synthesized by Genscript (Nanjing, China) and Sangon Biotech (Shanghai, China). (DOCX) S2 Table. Basic parameters of ω-3 FAD genes from chia and perilla. (DOCX) S3 Table. Identities of mRNAs (italic) and proteins among ω-3 FADs from chia, perilla and Arabidopsis Ã . Ã Pairwise-alignment and identity analysis of mRNAs and proteins of ω-3 FADs from perilla, chia and Arabidopsis were performed using the ClustalW method in Vector NTI advance 11.5.1 (Invitrogen, USA).