Natural Product Reports 26(7): 884-915 (July 2009)
Blue flower color development by anthocyanins: from chemical structure to cell physiology
Kumi Yoshida, Mihoko Mori and Tadao Kondo

2 History and mystery of blue color development
<snip>
In the 1970s Asen proposed a new concept, self-association, to describe the stabilization of anthocyanins.³⁶ A concentrated solution of a simple anthocyanin is more stable than a dilute solution, with a non-linear increase in the molar absorption coefficient in the visible absorption spectrum;³⁶ furthermore, the peak in the spectrum broadens. The stabilization effect due to molecular interactions seemed to be similar to the co-pigmentation effect. Goto then proposed a unified mechanism for copigmentation and self-association, which he called the molecular stacking theory.^3,15,16 He suggested that both co-pigmentation and self-association occur by hydrophobic interactions between aromatic rings in co-pigments and anthocyanins, or between anthocyanins, not between anthocyanins interacting via hydrogen bonds.^{3,15,16,37,38} Therefore, the chemical driving forces for co-pigmentation and self-association are the same — the hydrophobic aromatic residues in the molecules assemble in aqueous solutions. He expanded this idea into the intramolecular stacking theory, which addressed the stability of polyacylated anthocyanins.^3,18,16 In polyacylated anthocyanins, two or more aromatic acyl residues exist and the residues engage in intramolecular stacking with the anthocyanidin chromophore through the same hydrophobic interaction. Since two chromophores with different λ_max stack closely in co-pigmentation, charge-transfer occurs between them, leading to a bathochromic shift. In the case of polyacylated anthocyanins, the same charge-transfer phenomenon should occur between the intramolecular aromatic residues and the anthocyanidin chromophore, thereby leading to a bathochromic shift. In contrast to these examples, self-associated anthocyanins do not show a bathochromic shift in λ_max, but they do show peak-broadening of the spectrum. Interaction between the same chromophores splits the excitation energy into higher and lower levels, as occurs between neighboring bases in nucleic acids,¹⁴ and they exhibit an exiton-type Cotton effect in CD spectra, depending on whether the stacking is clockwise (a positive Cotton effect) or anti-clockwise (a negative Cotton effect).^15,16

In the 1980s, the research group of Goto and Kondo reported many structures of complex anthocyanins and provided experimental evidence in support of the molecular stacking theory.^3,15,16 They also reviewed mechanisms of blue-color development; however, the mystery of blue flower coloration remained. With regard to the pH theory, there were still no direct pH measurements, except for Asen's work using glass capillaries containing pH indicator.³⁹ Concerning the metal complex theory, the entire atomic structure of commelinin was still unsolved. The intramolecular stacking conformation of polyacylated anthocyanins was supported only by long-range NOEs in NMR analysis carried out in acidic media. Briefly, scientists knew that any anthocyanin could develop a blue color when it was maintained in the anhydrobase anion form in petal cell vacuoles. The problem was how to maintain such a structure, because plant vacuoles are filled with neutral or weakly acidic aqueous media. Since that time, research into blue flower coloration has aimed at clarifying this 'magic'.

3 Metalloanthocyanins
The term metalloanthocyanin refers to a self-assembled, supramolecular metal complex pigment; this complex is composed of stoichiometric amounts of anthocyanins, flavones, and metal ions.^3,15,40 In any metalloanthocyanin, the composition is fixed at 6:6:2, respectively (Fig. 3). Three major mechanisms for blue flower coloration exist: self-association, co-pigmentation, and metal complexation. Because all the proven mechanisms for blue flower coloration (except pH) include metalloanthocyanins, they may be, in a sense, the ‘topmost pegs’ in blue flower pigments. This section discusses the structure and mechanism of blue color development of five metalloanthocyanins obtained from five different flower petals (Fig. 4). Since 1992 two X-ray crystallographic structures have been solved: commelinin (2)¹⁷ from C. communis and protocyanin (3)⁴¹ from C. cyanus. These molecular structures, as well as blue color development, are discussed in detail. In addition to these two metalloanthocyanins, three more metalloanthocyanins have been found, protodelphin (4),^42,43 cyanosalvianin (5),⁴⁴ and nemophilin (6, K. Yoshida, unpublished). Compounds 4 and 5 come from blue salvia petals, Salvia patens and Salvia uliginosa, while 6 comes from blue nemophila petals, Nemophila menziesii (Fig. 5). We discuss below the effects that structure and metal ion species have on blue coloration.

3.1 Commelinin

3.2 Protocyanin

3.3 Other metalloanthocyanins

4 Fuzzy metal complex pigments
In blue petals, metalloanthocyanins are relatively unusual, nearly all blue flowers resulting from non-stoichiometric metalcomplex pigments stabilized by co-pigmentation. These blue pigments are less stable than those of metalloanthocyanins — they are blue only in aqueous solutions, and the color disappears during isolation or crystallization. Therefore, clarification of the mechanism of blue color development by such 'fuzzy' metal complex pigments may be more difficult than for metalloanthocyanins. For this purpose, Yoshida et al. have proposed a new methodology, "in vivo natural product chemistry" (Fig. 17).^19,20 This is combined with an intact analysis of colored vacuoles and reproduction of the blue color through reconstruction experiments. The analysis involves cell color measurement, vacuolar pH measurement, and quantitative analysis of organic and inorganic components in the cells. Color reproduction means achieving the same color by mixing components under conditions based on the analytical results. Micro-spectrophotometry is employed for color analysis, proton-sensitive microelectrodes are inserted directly into the colored vacuoles for vacuolar pH measurements, and a microHPLC system is used for quantitative analysis.^19,20 The similarity of the reproduced color to the color of the natural petals is investigated using several spectroscopic methods: Vis absorption, CD, and NMR. In the following discussion, we describe the blue coloration of such fuzzy metal complex pigments (Fig. 18).

Fig. 18. Various blue flowers (left), their transverse sections (middle), and protoplasts obtained by treatment with cellulase and pectinase (right). Scale bars for the transverse sections and protoplasts: 50 μm. (Modified from K. Yoshida et al., Plant Cell Physiol., 2003, 44, 262-268; K. Yoshida et al., Phytochernistry, 2006, 67, 992-998; and Shoji et al., Plant Cell Physiol., 2007, 48, 243-251, with permission).

4.1 Blue color in hydrangea sepals
The flowers, or more specifically the petal-like sepals, of Hydrangea macrophylla show various colors, such as red, mauve, purple, violet, and blue.⁵⁶ The most interesting phenomenon in hydrangea coloration is that all the colors are developed from one simple anthocyanin, delphinidin 3-glucoside (49);⁵⁷ in general, blue and red flowers contain different anthocyanin chromophores.⁵⁸ In blue petals, delphinidin derivatives usually exist, while in red petals, the pigment contains either pelargonidin or cyanidin chromophores. The studies on color development in hydrangea sepals have a long history, spanning more than a hundred years. The effect of acidic soils on blue coloration due to Al3+ was reported as early as the first part of the 20th century,59,60 and the existence of acylquinic acids (50-52) as co-pigments was discovered as early as the 1950s.⁶¹ From 1985 to 1990, Takeda et al. reported that a stable, blue-colored solution could be obtained by mixing 49, 5-O-acylquinic acid (51, 52), and Al³⁺.62-64 However, the mystery of how the same components could develop different colors in hydrangea sepals and why the color changed so easily remained unsolved.

In hydrangea sepals, the colored cells are located in the second layer; therefore, Yoshida et al. prepared protoplast mixtures, from which they collected and analyzed only colored cells.^19,65 Vacuolar pH (pHv) measurements of colored cells illustrated the difference between blue and red cultivars. The pH of blue cells in the blue cultivar was approximately 4.1, significantly higher than that in red cells (pH = 3.3).⁶⁵ Ito et al. analyzed the composition of anthocyanin (49) and three co-pigments (50-52) by collecting approximately 150 colored cells (Table 8, D. Ito and K. Yoshida, unpublished). The results indicated that the molar ratio of 5-O-acylquinic acids (51, 52) to 49 was much higher in the blue cells than that in the red cells. The amount of Al³⁺ was the same; in blue cells, the molar equivalent of Al³⁺ to 49 was greater than 1 eq., while the amount in red cells was lower than 0.1 eq. These results were significantly different from data obtained from whole sepal tissue.⁶⁶ This discrepancy emphasizes the importance of isolating and analyzing only colored cells in flower color studies. To measure the composition in colored cells with greater sensitivity, Yoshida et al. developed a single-cell analysis method. Monitoring the cell color by micro-spectrophotometry, a single cell was collected, and then the organic or inorganic components were quantified. These results showed an obvious correlation between cell color 'blueing' and increase in the levels of 5-O-acylquinic acid (51, 52) and Al³⁺.²⁰

Kondo et al. synthesized various co-pigment analogs and carried out reconstruction studies of hydrangea colors. To obtain a stable blue color, addition of 3 eq. of 5-O-acylquinic acid (51 or 52) to 49 turned out to be essential in the presence of 1 eq. of Al³⁺ at pH 4.0. The regioisomer 3-O-caffeoylquinic acid (50) did not give a stable blue solution, but instead produced a blue precipitate, which was composed of only 49 and Al³⁺. Kondo et al. described that Al³⁺ coordinated to 49 is blue but water-insoluble; the co-pigments, i.e. the 5-O-acylquinic acid derivatives, may solubilize the blue metal complex to give a stable blue solution.⁶⁷ Reconstruction experiments using unnatural synthetic co-pigments (53-59) revealed the essential structure of the co-pigment for blue coloration in hydrangea (Table 9). The 5-O-ester, l-OH, and 1-carboxyl groups in the quinic acid parts are essential for the co-pigmentation effect and formation of the water-soluble blue metal-complex pigment.^67,68 Integrating all of these results, Kondo et al. proposed a schematic complex model in hydrangea, in which Al³⁺ coordinates to 49 through the ortho-dihydroxyl group of the B-ring and, simultaneously, the Al³⁺ may coordinate the oxygen atoms in the co-pigments (Fig. 19). In addition, the aromatic ring of the co-pigment moiety may stack on top of the chromophore of 49 through hydrophobic interactions. This supramolecule is labile and can easily fall apart depending on the concentrations of the components and the pH conditions; thus, the blue color of hydrangea is unstable.

4.2 Other blue flowers
Non-stoichiometric metal-complex pigments similar to those of hydrangea have been reported. The sky-blue petals of Meconopsis grandis contain anthocyanins (60, 61) with a cyanidin chromophore,^69,70 as well as flavonol glycosides (62, 63).⁷⁰ Yoshida et al. reported that the blue petals have λ_max values at 606 and 646 nm, and its vacuolar pH is 4.8.71 The concentrations of anthocyanins (60 and 61), flavonols (62 and 63) and Fe in the colored protoplast are 4.7 mM, 15 mM and 3.6 mM, respectively. Mixing 61, 63, and the Fe ion and Mg ion at pH 5.0 gives the same blue color, and the Vis and CD spectra of the resulting compounds are identical to those of the petals (Fig. 20). Yoshida et al. concluded that the blue petal color of M. grandis is due to a new type of metal-complex anthocyanin composed of an anthocyanin with a cyanidin chromophore (60 and 61), two or more equivalents of kaempferol derivatives (62 and 63), 1/6 equivalent of Fe ion, and excess Mg²⁺.⁷¹ They also stated that an excess of Fe ions with respect to anthocyanin yields a different blue-black-colored complex that is identical to that in the protocyanin.⁷¹

The petals of Tulipa gesneriana cv. Murasakizuisho are mostly purple, but blue near the centre of the flower. Shoji et al. reported that the blue coloration may be caused by a 'fuzzy' metal complex pigment.⁷² The only chemical difference between the purple cells and the blue cells is in the content of Fe³⁺; other factors, such as vacuolar pH, structure, and amount of anthocyanin (64) and glycosylated flavonols (65-67), are the same (Table 10).⁷² They reported that mixing anthocyanin and flavonol at pH 5.6 yields a purple color that is the same as that of the distal part of the petal, and that addition of 1 eq. of Fe³⁺ to the solution changes it to a blue color identical to that of the proximal part.⁷² In blue coloration of the tulip, the Fe³⁺ ion is essential, though the chromophore of the anthocyanin is delphinidin.⁷² Momonoi et al. recently cloned a vacuolar iron transporter gene from tulip petals and showed that the transporter plays a critical role in petal blueing by accumulating iron.⁷³

Some researchers have reported blueing mechanisms without any contributions from metal ions, but instead resulting from copigmentation by flavonols and flavones. Markham et al. reported that three geranium petals, the bluish-purple petals of Geranium "Johnson's Blue", the purplish-blue petals of G. pratense, and the bluish-magenta petals of G. sanguineum, contain the same anthocyanins: malvidin 3-O-glucoside-5-O-(6-O-acetylglucoside) (68), with a small amount of malvidin 3-O-(6-O-acetylglucoside)-5-O-glucoside(69).⁷⁴ The difference between the three colored petals is in the molar ratios of the flavonols, 3-O-glucosyl- and 3-O-sophorosyl-kaempferol and myricetin (70-73): the ratio of flavonol to anthocyanin is lower in the magenta petals than in the bluish petals. They proposed a co-pigment effect of the flavonols, but could not reconstitute the same blue color by adding flavonols (2-4 eq. to anthocyanin) at pH 5.5, the same as that of the petal juice. However, a similar blue color was obtained at pH values of 6.6-6.8; therefore, the authors concluded that the vacuolar pH may be higher than 5.5.⁷⁴

For a long time, the color of blue iris flowers has been attributed to a co-pigmentation effect.⁷⁵ When C-glycosylflavones, such as swertisin (74), O-xylosylswertisin (75), vitexin (76), isoorientin (77) and swertiajaponin (78), are added to iris anthocyanin, a blue color develops in the absence of metal ions.⁷⁵ Yabuya et al. reported that Iris ensata contains the 3-O-(4-O-p-coumaroylrhamnosyl-6-O-glucoside)-5-O-glucosides of malvidin (79), petunidin (80), and delphinidin (81), along with isovitexin (82); the delphinidin derivative is known as violanin or nasunin. Addition of 2-10 equiv. of 82 to 0.1 mM solutions of each of the anthocyanins 79-81 yields a blue coloration with a bathochromic shift of λ_max of more than 30 nm.⁷⁶

Petals of Anagallis monelli contain 3-glucosides of pelargonidin (83), delphinidin (49) and malvidin (84).^77,78 Blue- and red-coloured forms have upper and lower surfaces of the same color, but an intermediate hybrid with violet-and-lilac petals shows different colors; the petals are blue in the adaxial epidermis and red in the abaxial epidermis. The pigment in the blue surface was identified as 84 and that in the red surface as 49; this is in contrast to the standard assumption that methylation of the -OH residue at the B-ring causes a hyperchromic shift.⁷⁸ The violet-and-lilac petal surface has a mixture of blue and red cells, with the red cells containing a higher level of 49. Furthermore, the upper surface tissue, composed of bluer cells, is more acidic than the lower surface.⁷⁸ This is an unusual phenomenon considering that the color change of anthocyanins depends on pH, although the authors did not provide any conclusive mechanism. Resolving this question may require single-cell analysis.

5 Anthocyanins with intramolecular stacking
Polyacylated anthocyanins, which are anthocyanins with two or more aromatic acyl residues, such as p-coumaroyl, caffeoyl, feruloyl, and p-hydroxybenzoyl moieties, are found mainly in bluish flower petals (Fig. 21).^{1,3,7,8,15,16} Therefore, polyacylated anthocyanins have been thought to possess some kind of special mechanism for blue color development. In 1971 platyconin (85) from the bluish-purple Chinese bellflower Platycodon grandiflorum, was described,⁷⁹ and the first complete structure of a polyacylated anthocyanin, that of gentiodelphin (86) from blue Gentiana makinoi, was published in 1982.⁸⁰ Until 1992, Goto and Kondo published many structures of polyacylated anthocyanins, 81–87 and they proposed intramolecular sandwich-type stacking of aromatic acyl residues onto the anthocyanidin chromophore in order to account for the stability and blue color.^3,15,16 Yoshida et al. reported evidence for intramolecular stacking in gentiodelphin; they observed that a caffeoyl residue in the B-ring stacks on top of the anthocyanidin nucleus, even under strongly acidic conditions.88 So far, hundreds of polyacylated anthocyanins have been identified,⁵ but no X-ray structures have been solved because of the difficulties in crystallizing these molecules. Therefore, very little direct chemical structural evidence exists to explain the unique characteristics of polyacylated anthocyanins. In the following discussion, the cause of petal color change from red to blue during flower opening of the morning glory is described. We then mention recently reported characteristics of polyacylated anthocyanins in blue flower color development and newly identified anthocyanins covalently linked to flavonoids.

5.1 Blue morning glories

5.2 Other polyacylated anthocyanins

5.3 Anthocyanins covalently linked to another flavonoid

---

7 Perspectives
Over the past 15 years, mechanisms of blue flower coloration have been clarified to a considerable extent. This progress is due to development of instrumental analysis and new techniques of single-cell treatment and analysis. In the case of metalloanthocyanins, we expect that more varieties of such pigments will continue to be found. The mechanism of blue coloration may be common among blue flowers, and so by screening such pigments by direct Vis and CD measurements of intact petals, researchers should be able to find new metalloanthocyanins. The effects of structural differences on the formation of metalloanthocyanins – a combination of matching pair and nonmatching pair – should extend our understanding of the molecular recognition chemistry involved.

However, ‘fuzzy’ unstable blue coloration may remain a problem. Structural determination of such weakly associated pigments in aqueous solutions and a ‘breakthrough idea’ is sorely needed. Petals of blue clover [Trigonella caerulea] and blue Muscari species also contain simple anthocyanins; therefore, these cases of blue flower coloration may be due to a similar ‘fuzzy’ metal complex. Polyacylated anthocyanins seem to be classified into two types. One is a group consisting of gentiodelphin, HBA, and ternatins; aromatic acyl residues may stack onto both sides of the anthocyanidin chromophore. The other is a group including tecophilin, phacelianin and cyanodelphin, in which intramolecular stacking of aromatic acyl residues onto the chromophore and intermolecular stacking of chromophores occur simultaneously. Except for morning glory petals, the true blueing mechanisms have not yet been clarified; therefore, physicochemical studies on the charge transfer effect on the bathochromic shift need to be undertaken. The atomic structure of intra- and intermolecular stacking also needs to be resolved. In addition to chemical studies, mechanisms of metal and pigment transport into vacuoles remain a major unresolved question, and we expect that molecular biological studies related to the chemical mechanisms for blue coloration will progress in the next decade. For a true blue rose to be developed, a multilateral strategy is necessary.

Flower Pigments