A chalcone isomerase-like protein enhances flavonoid production and flower pigmentation

Characterization of a spontaneous efp–1 mutant

While characterizing several mutable alleles that confer flower color variegation in Ipomoea (Chopra et al., 2006), we obtained a spontaneous mutant carrying a homozygous mutable efp–1 allele that had variegated flowers with normally pigmented sectors on a pale-colored background (Figure 2a). As Tpn1-related transposons in the CACTA super-family are a common spontaneous mutagen of Ipomoea cultivars (Iida et al., 1999, 2004; Chopra et al., 2006), we used a simplified transposon display (STD) procedure (Fukada-Tanaka et al., 2001) to identify the efp–1 mutation. A 259 bp STD fragment containing a 224 bp genomic segment was amplified from all 11 efp–1 homozygous plants but was not amplified from six EFP homozygous plants (Figure 2b). To identify the origin of the 224 bp segment, we used inverse PCR (IPCR) amplification to isolate its flanking sequences from the wild-type EFP plant. The obtained 1.1 kb flanking sequence showed homology to the 5′ coding region of a cDNA sequence in the Ipomoea EST database (Morita et al., 2006). Sequence comparison of the full-length cDNA with the corresponding genomic DNA region revealed that the EFP gene comprises five exons, and the efp–1 mutant contains a 6.3 kb Tpn1-related element, Tpn13, inserted into the promoter region 129 bp upstream of the transcriptional start site (Figure 2c). From the efp–1 mutants, we independently obtained two germinal revertants, EFP–GR1 and EFP–GR2, that had normally pigmented flowers (Figure S1a). Although the Tpn13 insertion within the promoter region in the efp–1 mutant was flanked by a 3 bp target site duplication, TAA, the germinal revertants have lost Tpn13 and have footprints generated by the Tpn13 excision (Figure 2c). Southern blot analysis revealed that the EFP gene is a single-copy gene in the Ipomoea genome, and the restriction fragments of the wild-type EFP and revertant EFP–GR1 alleles are identical in size (Figure 2d). These results strongly indicate that the efp–1 mutation was caused by insertion of Tpn13 into the EFP promoter.

Identification of other efp alleles

We examined whether other Ipomoea mutants with under-tinted flowers also carried efp mutations. We screened 39 candidate mutant lines in our collections, including commercially available lines, and found two additional mutant alleles, efp–2 and efp–3, that were present in 22 and 10 lines, respectively (Figure S1a and Table S1). This result indicates that the efp mutations are a major cause of the pale-colored flower phenotype in the available Ipomoea mutants. The efp–2 mutation was caused by insertion of a 5.6 kb Tpn1-related element, Tpn14, within the coding region of exon 5, and this element was flanked by a 3 bp direct repeat of GAT (Figure 2c). Interestingly, the efp–2 allele is quite stable, and neither excision of Tpn14 from the EFP gene nor flower variegation, indicative of somatic reversion of the efp–2 allele, were detected. The efp–3 allele showed a frameshift mutation caused by a 17 bp insertion in the coding region of exon 5 (Figure 2c). The 17 bp insertion comprises an imperfect direct repeat, and 4 bp microhomology was found at the breakpoints. These results further confirm that efp mutations are responsible for the pale-colored flower phenotypes.

The EFP gene encodes a type IV CHI

The predicted EFP gene product is 206 amino acids long, and a phylogenetic analysis revealed that it is most closely related to one of the four types of CHI super-family proteins (type IV) (Ralston et al., 2005) (Figure S2a). The deduced amino acid sequence of EFP showed 64% identity to that of the Arabidopsis type IV protein (accession number At5g05270), often called AtCHIL (Ngaki et al., 2012) (Figure S2b). As in other type IV CHI proteins (Jez et al., 2000; Ralston et al., 2005; Ngaki et al., 2012), EFP lacks the conserved amino acid residues for CHI enzymatic activity (Figure S2c).

Expression analysis of EFP and structural genes

We characterized spatial and temporal expression of the EFP gene by RT–PCR and Northern blot analyses (Figure 3a,b). RT–PCR analysis revealed that the EFP gene is predominantly expressed in petal tissues, in both the limb and the tube, coincident with expression of the CHI gene encoding the type I CHI enzyme (Figure 3a). The EFP gene was also slightly expressed in other tissues, including the leaf blade and stem, in which anthocyanin accumulates. The amount of the EFP transcript in the flower limb increased during flower development and was highest 12 h before flower opening (Figure 3b). The temporal expression profile (Figure 3b) was very similar to those of structural genes for flavonoid biosynthesis as reported previously (Morita et al., 2006).

Next, we characterized genetic regulation of EFP expression by Northern blot analysis (Figure 3c,d). In the myb1 or wdr1 mutant, EFP transcripts were absent in the petals (Figure 3c). Although EFP transcripts accumulated normally in the efp–3 mutants, they were not detected in the efp–1 and efp–2 mutants (Figure 3d). In the germinal revertant (EFP–GR1) of the efp–1 mutation, EFP expression was fully restored (Figure 3d).

We further compared the accumulation of mRNA for three structural genes, CHS–D, DFR–B and 3GT (Figure 1a), between the EFP and efp–1 flowers using real-time RT-PCR. The results showed no significant difference in transcript accumulation for the genes examined (Figure 4).

Flavonoid analyses of the EFP mutants

HPLC analysis revealed that the EFP flowers contained WBA and three other minor anthocyanins. WBA accounted for 37.4 ± 0.7% (n = 3) of the total anthocyanin content (Figure S3a,b). The anthocyanin content in pale-pigmented efp–1 petals was less than approximately 22% of that in normally colored petals of the wild-type EFP plant (Figure 5a). WBA accounted for 80.1 ± 0.7% (n = 3) of the total anthocyanin content in the efp–1 mutant (Figure S3c). Decreases in the total amount of anthocyanins and increases in the relative amount of the anthocyanin end products WBA or HBA appear to be characteristic features of efp flowers. Two efp–2 mutants, Akatsuki–no–nami and Toen (Figure S1a), were reported to contain significantly lower amounts of anthocyanins than normally colored lines did. The relative amounts of the anthocyanin end products WBA in Toen or HBA in Akatsuki-no-nami were 76.9% or 81.9% of the total anthocyanin content, respectively, compared with approximately 10–40% in various lines that produce normally colored flowers (Lu et al., 1992a,b).

In addition to anthocyanins, the EFP flowers accumulated a flavonol, kaempferol glycoside (Figure S4a), but the flavonol content in the efp–1 petals was less than 27% of that in the EFP petals, a level comparable to the reduced level of anthocyanins (Figure 5b). Moreover, no chalcones were detected in the efp–1 flowers (Figure S4b). In contrast, the levels of caffeic acid derivatives, namely chlorogenic acid and 1–O-caffeoyl glucoside, markedly increased in the efp–1 petals (Figure S4b). These caffeic acid derivatives partially share a common biosynthetic pathway with p–coumaroyl CoA, which is a substrate for CHS (Figure 1a).

RNAi-mediated inactivation of the EFP homologs in petunia and torenia

To examine whether the gene corresponding to EFP enhances flavonoid production in other ornamental flowers, cDNAs for one and two EFP homologs, respectively, were isolated from petunia (PhEFP) and torenia (ThEFP–A and ThEFP–B) (Figure S2b), and used in an RNAi-mediated gene silencing strategy. For knockdown of torenia EFP genes, we used highly conserved coding regions showing more than 96% identity between ThEFP–A and ThEFP–B as RNAi triggers for construction of RNAi vectors (Figure S5). The ThEFP–A or ThEFP–B RNAi constructs were separately transformed into host plants with normally pigmented flowers. The flowers of these transgenic petunia and torenia plants were pale-colored (Figure 2e,f and Figure S1b,c). As expected, the total amount of anthocyanins and colorless flavonoids that accumulated was reduced in the knockdown plants of petunia (Figure 5c,d) and torenia (Figure 5e,f). Detailed analyses of their flavonoid components are shown in Tables S2 and S3. The relative amounts of anthocyanins in the knockdown petunia and torenia plants decreased to approximately 28% and 49% of those in the control plants, respectively (Figure 5c,e). Petunia and torenia flowers are known to contain flavonols and flavones, respectively, and the flavones in torenia clearly serve as a co-pigment in flower pigmentation (Bloor et al., 1998; Aida et al., 2000). The amounts of flavonols and flavones in the knockdown petunia and torenia decreased to approximately 12% and 31% of those in the control plants, respectively (Figure 5d,e). This concerted reduction in both anthocyanins and flavone co-pigments in torenia results in substantially reduced flower pigmentation. Thus, it is clear that the EFP homologs encode enhancer proteins of flavonoid biosynthesis in both petunia and torenia.

EFP is unlikely to be a bona fide CHI, transcription factor or anthocyanin transporter

The EFP gene was found to encode a type IV CHI that contains amino acid substitutions in several catalytic residues of type I and II CHI enzymes (Figure S2c), suggesting that EFP has no CHI activity. Indeed, Ipomoea mutants in CHI produce pale yellowish flowers that accumulate chalcone 2′-O-glucoside rather than lacking anthocyanins (Iida et al., 2004; Saito et al., 2011). In addition, Arabidopsis AtCHIL is tightly co-expressed with CHI in seeds as well as other tissues (Yonekura-Sakakibara et al., 2012), and mutants in CHI yield plants lacking proanthocyanidins (Shirley et al., 1992; Dong et al., 2001). The yellow seeds of Arabidopsis CHI mutants accumulate chalcone derivatives and a reduced amount of flavonols (Bottcher et al., 2008). Thus, Ipomoea EFP and AtCHIL clearly have no CHI enzymatic activity, and are unable to complement the defects in the CHI genes in Ipomoea flowers and Arabidopsis seeds. Type IV CHI proteins probably also show no CHI activity in other species and tissues, as they lack most of the key catalytic residues of bona fide CHI proteins (Jez et al., 2000; Ralston et al., 2005; Ngaki et al., 2012), and because recombinant type IV CHI does not metabolize chalcones in vitro (Ralston et al., 2005).

Jez et al. (2000) showed that CHI super-family proteins lack any detectable homology with other proteins, and the type IV CHI proteins are most closely related to CHI types I and II among the CHI super-family (Ngaki et al., 2012). Therefore, EFP is unlikely to be a transcription factor or anthocyanin transporter. Indeed, there were no substantial differences in expression of the structural genes between the EFP and efp–1 flowers (Figure 4), indicating that EFP does not contribute to efficient anthocyanin production through transcriptional regulation. Moreover, mutants affecting the transport of anthocyanins into the vacuole in maize (Zea mays) and carnations (Dianthus caryophyllus) showed the reduced pigmentation associated with a significant reduction in anthocyanin content without alterations in the composition of anthocyanins with various chemical structures (Goodman et al., 2004; Sasaki et al., 2012). The features of these mutants with no alterations in their anthocyanin compositions clearly differ from those of the efp mutants described here (Figure S3), suggesting that EFP is unlikely to be involved in anthocyanin transport into the vacuole.

EFP regulates flower color intensity

The Ipomoea efp mutants showed an pale-colored flower phenotype that is mainly due to a reduced production of anthocyanins even in the null mutants (Figure 5a). Such anthocyanin production is not due to the presence of redundant EFP copies, as the EFP gene is a single-copy gene in the Ipomoea genome (Figure 2d). This is in sharp contrast to null mutants in the structural genes for anthocyanidin biosynthesis (Figure 1a). For example, null mutants in the CHS–D, CHI or DFR–B genes have acyanic flowers containing virtually no anthocyanins (Saito et al., 1994; Iida et al., 2004; Hoshino et al., 2009). Unlike the biosynthetic enzymes, the EFP protein is not essential for anthocyanin biosynthesis but enhances flavonoid production and flower color intensity.

Flower color intensity is an important floricultural trait, and it is mainly due to anthocyanin concentration. It depends heavily on activities for flavonoid synthesis and anthocyanin transport into the vacuole that are regulated by transcription factors. For example, the R2R3-MYB proteins of Antirrhinum species (Schwinn et al., 2006; Whibley et al., 2006) and Phlox drummondii (Hopkins and Rausher, 2011) and the bHLH proteins of petunia and two Ipomoea species, I. purpurea and I. tricolor, increase flower color intensity (Spelt et al., 2000; Park et al., 2004, 2007). In contrast, an R3-MYB protein is known to be a suppressor of flower color intensity in Mimulus species (Yuan et al., 2013). Mutations of the genes for Ipomoea bHLH proteins have been used to generate variations in flower color intensity in Ipomoea cultivars (Park et al., 2004, 2007). Mutations of efp genes in ornamental plants may also be useful in generating pale-color varieties. Indeed, 32 of the 39 Ipomoea lines that show pale-colored flowers and are characterized here are EFP mutants (Table S1).

Coordinated expression of EFP with the genes for flavonoid production is conserved among diverse plant species

The spatio-temporal expression patterns of EFP were very similar to those of the structural genes for flavonoid biosynthesis, and both EFP and these structural genes are under the control of at least two regulatory genes, MYB1 and WDR1 (Figure 3c) (Morita et al., 2006).

In Arabidopsis, a strong association of AtCHIL expression with expression of structural genes for flavonoid biosynthesis in transcriptional profiles has been observed in several studies. Coordinated up-regulation of AtCHIL and structural genes in the flavonoid biosynthetic pathway was observed in tissues that accumulate flavonoids when PAP1, an MYB regulatory gene, was over-expressed (Tohge et al., 2005), or when seedlings were treated with sucrose (Solfanelli et al., 2006). In response to light and UV–B radiation, the bZIP transcription factor HY5 regulates expression of the MYB regulators, MYB12 and MYB111, which results in coordinated up-regulation of expression of AtCHIL and the flavonoid biosynthetic pathway structural genes (Stracke et al., 2010). Comprehensive transcriptome co-expression analysis also revealed that AtCHIL is coordinately regulated with early biosynthetic genes in flavonoid production, namely CHS, CHI and F3H (Yonekura-Sakakibara et al., 2008, 2012).

An EFP gene in the monocotyledonous orchid Oncidium Gower Ramsey (OgEFP; Figure S2) that has been wrongly described as a CHI gene is expressed in anthocyanin-accumulating floral tissues (Chiou and Yeh, 2008). It is also coordinately regulated with flavonoid biosynthetic pathway structural genes, including CHS, DFR and ANS (Chiou and Yeh, 2008). Expression of the orchid EFP and DFR genes is activated by transient expression of a MYB gene involving anthocyanin pigmentation in flower lips. Taken together, these observations indicate that coordinated expression of EFP with the genes for flavonoid production is conserved among diverse plant species, and suggest that both AtCHIL and orchid EFP are involved in efficient production of flavonoids.

EFPs activate the early step of flavonoid biosynthesis

The Ipomoea EFP mutants and the EFP knockdown petunia and torenia plants showed significant reduction of colorless flavonoids, flavonols or flavones (Figure 5). As flavonols and flavones are produced early in the flavonoid biosynthetic pathway (Figure 1a), EFP activates at least the early steps of the flavonoid pathway. The flavone reduction in the EFP knockdown torenia plants suggests that EFP acts on CHS or CHI rather than on F3H in the early steps. However, it is unlikely that the efp mutations primarily reduce CHI activity because no chalcone 2′–O-glucoside, which is accumulated in Ipomoea CHI mutants, was detected in the efp–1 flowers (Figure S4). The levels of chlorogenic acid and 1–O-caffeoyl glucoside were markedly increased in the efp–1 petals (Figure S4). These features are substantially shared with those of null mutants in the CHS–D gene (Saito et al., 1994; Hoshino et al., 2009). These observations suggest that the EFP is likely to act as an enhancer primarily on CHS to ensure that sufficient amounts of flavonoids are produced.

Possible mechanisms for efficient flavonoid production by EFP

The detailed molecular mechanisms of how EFP enhances flavonoid production remain to be elucidated. Our working hypothesis is that EFP directly interacts with CHS, chalcone and/or related products to perform its role as an enhancer.

In the early steps of flavonoid biosynthesis in Arabidopsis, including CHS, CHI, F3H and DFR, the enzymes involved associate to form enzyme complexes that are thought to ensure efficient flavonoid production (Burbulis and Winkel-Shirley, 1999; Winkel-Shirley, 1999, 2001). In such complexes, EFP-related CHI physically interacts with CHS (Burbulis and Winkel-Shirley, 1999). It is tempting to speculate that EFP must be a component of such complexes, and that the CHI-related EFP protein may physically interact with CHS. Such an interaction would facilitate channeling of intermediates between enzymes in the complexes, and would not competitively interfere with a physical interaction between CHS/chalcones and CHI.

An alternative hypothesis is that EFP may primarily interact with chalcone or related products, including reaction intermediates and by-products such as bis-noryangonin and coumaroyl triacetic acid lactone (Austin and Noel, 2003). X–ray crystallographic analyses of the Arabidopsis CHI-related proteins revealed that the ligand-binding pocket of AtCHIL is much more similar to that of AtCHI than to those of AtFAPs (Jez et al., 2000; Ngaki et al., 2012). The EFP protein may function in guiding the proper folding of a polyketide intermediate(s), e.g. the linear tetraketide intermediate, for correct cyclization (Austin and Noel, 2003). We may also imagine a hypothesis in which EFP traps and subsequently releases an undesirable by-product(s) that may temporarily suppress the CHS-catalyzed reaction; this rate-limiting step may be relieved by EFP, leading to efficient production of flavonoids.

Although the detailed molecular mechanisms of how EFP enhances flavonoid production remain to be elucidated, EFP appears to be a unique positive regulator protein that has an enzyme-like structure without showing enzymatic activity, and that enhances a metabolic pathway involving the regulator-related enzyme. It remains to be seen whether similar enhancers are found in metabolic pathways other than the flavonoid biosynthetic pathway.

Effect of the efp mutations on anthocyanin composition

Flowers of Ipomoea cultivars selected for favorable color tones often contain several minor anthocyanins bearing fewer glucosyl and caffeoyl moieties than WBA or HBA do (Lu et al., 1992a,b; Yamaguchi et al., 2001). A possible explanation is that spontaneous mutations may have accumulated such that addition of glucosyl and/or caffeoyl moieties became rate-limiting steps in the production of the end products WBA or HBA; such rate-limiting steps may result in accumulation of the minor anthocyanins. Inactivation of EFP changes the rate-limiting step from anthocyanin modification to a CHS-catalyzed reaction, which decreases the total amount of anthocyanins and concomitantly increases the relative amount of WBA or HBA in the total anthocyanin content (Figure 4 and Figure S3).

Evolutionary implications of EFP

Whereas CHI-related FAP genes that encode fatty acid-binding proteins are widely distributed in both land plants and green algae, EFP and CHI are detected only in land and vascular plants, respectively (Ngaki et al., 2012). EFP has been proposed to have initially arisen from the FAP gene in mosses, and may have served as the ancestor of CHI in vascular plants (Ngaki et al., 2012). Flavonoids are found ubiquitously in vascular plants and in a high proportion of bryophytes, including mosses, but are absent in algae (Markham, 1988). Flavonoids may have evolved as chemical messengers, and may have played important roles in protection against UV and UV-associated stresses during early land plant evolution (Winkel-Shirley, 2001; Pollastri and Tattini, 2011). In the model moss Physcomitrella patens, a CHS super-family gene was shown to encode an active CHS enzyme (Jiang et al., 2006); the moss also contains five putative F3H genes (Koduri et al., 2010). Although the P. patens genome was reported to contain two CHI genes (Koduri et al., 2010), they are actually EFP genes encoding type IV CHI proteins (Ngaki et al., 2012; Figure S2). CHS and one of the EFP genes, together with production of flavonols, were reportedly induced by UV–B irradiation in P. patens (Wolf et al., 2010). The flavonols of P. patens must be produced without an active CHI gene, probably through spontaneous isomerization of chalcones. Perhaps EFP acts on CHS, chalcone and/or related products to enhance flavonol production in P. patens. Even in the presence of the CHI enzyme in vascular plants, EFP retains its activity to ensure production of sufficient amounts of flavonoid pigment, thereby playing an important role in flower pigmentation of ornamental plants. Further elucidation of the EFP proteins in various plants will provide insights not only into regulation and evolution of flavonoid biosynthesis, but also into metabolic engineering strategies for flavonoid-related pigmentation and production of ‘nutraceutical’ products.

Plant materials

The wild-type Ipomoea nil Tokyo–Kokei Standard (TKS) has blue flowers, and the 78WWc–1 and NS/W1ca1 lines produce white flowers due to the myb1 and wdr1 mutations, respectively (Morita et al., 2006). The spontaneous efp–1 mutant line has pale flowers, occasionally with red spots and sectors (Figure 2a), and was isolated from a descendant of the red flower cultivar Beni Chidori (Takii & Co. Ltd, http://www.takii.co.jp). The efp–1 line, as well as the wild-type EFP line, was established from segregants of an efp–1/EFP heterozygous plant. Thirty-nine pale-colored flower lines were used for efp mutant screening (Table S1). The petunia cultivar Surfinia Purple Mini (Tsuda et al., 2004) and the torenia cultivar Summerwave Blue (Ueyama et al., 2002) were obtained from Suntory Flowers Ltd (http://www.suntory.co.jp/flower/).

Nucleic acid procedures

General nucleic procedures, including genomic DNA and total RNA isolation, PCR, RT–PCR, real-time RT–PCR, Northern blotting and DNA sequencing analyses were performed as described by Morita et al. (2006). PCR primers used in this study are listed in Table S4. We used the primer sets InCHI–F1 and InCHI–R2, InCHI–B–F2 and InCHI–B–R1, CHS–D–F2 and CHS–D–R1, DFR–B–F1 and DFR–B–R2, 3GT–F1 and 3GT–R1, and ATPase–F3 and ATPase–R1 to amplify the CHI, EFP, CHS–D, DFR–B, 3GT and ATPase cDNA fragments, respectively, in RT–PCR analyses.

EFP cloning

We used the simplified transposon display (STD) method (Fukada-Tanaka et al., 2001), with minor modifications, to clone the EFP gene. Genomic DNA isolated from the efp–1 mutants and wild-type EFP plants was cleaved using TaqI and ligated to a TaqI adapter (5′-GAGGATGAGTCCTGAG-3′ and 5′-CGCTCAGGACTCAT-3′). We used the primer sets TaqI and TIR for pre-amplification and TaqI–C and TIR–T for subsequent selective amplification by PCR (Table S4), and observed co-segregation of a 259 bp STD fragment and efp–1. For amplification of the efp–1 allele-specific 224 bp fragment, we used the primers TaqI–F1 and TIR–R. We subsequently isolated the flanking DNA of the STD fragment from the EFP plant by inverse PCR (IPCR) using TaqI and PCR primers IPCR–A1 and IPCR–B1 for the first round of amplification and IPCR–A2 and IPCR–B2 for subsequent nested PCR. Based on the IPCR fragment sequence, we also obtained EFP cDNA clones from an EST library (Morita et al., 2006), and screened 27 648 BAC clones from wild-type TKS to isolate the genomic EFP gene. Genomic EFP fragments of the EFP–GR1, EFP–GR2, efp–1, efp–2 and efp–3 alleles were amplified using appropriate primers.

Characterization of EFP in petunia and torenia

Conserved EFP cDNA sequences were amplified from petunia (PhEFP) and torenia (ThEFP–A and ThEFP–B) using degenerate primers CHI–B–F2 and CHI–B–R2 (Table S4). A full-length PhEFP cDNA was obtained using a GeneRacer kit (Invitrogen, http://www.lifetechnologies.com), and ThEFP–A and ThEFP–B cDNA clones were isolated from a cDNA library. RNAi vectors for suppressing EFP expression were constructed (Figure S5) and introduced into petunia (Tsuda et al., 2004) and torenia (Ueyama et al., 2002). We obtained and analyzed 4 petunia mutants (PT258-5, PT258-18, PT258-19 and PT258-23) as well as 3 (TT48-3, TT48-11 and TT48-27) and 4 (TT49-1, TT49-4, TT49-17 and TT49-19) torenia mutants harboring the ThEFP-A and ThEFP-B RNAi construct, respectively.

Flavonoid analyses

Flavonoid pigments of Ipomoea were extracted from individual petals and characterized by HPLC as described by Toki et al. (2001) and Saito et al. (2006). Anthocyanin and flavonol contents were measured based on absorbance at 530 and 360 nm, respectively. In petunia and torenia, flavonoids were extracted from 0.5 g of petals and were characterized using HPLC as described by Katsumoto et al. (2007).

Accession numbers

The DDBJ accession numbers for the sequences/mutants described in this paper are given in parentheses: EFP mRNA (AB545799); EFP gene from wild-type (AB545800); efp–1 mutant (AB545801); EFP-GR1 plant (AB545802); efp–2 mutant (AB545803); efp–3 mutant (AB545804); PhEFP mRNA (AB543054); ThEFP–A mRNA (AB543055); ThEFP–B mRNA (AB543056).