Flavonoid diversity and metabolism in 100 Rosa X hybrida cultivars

Pergamen Phyrahmlstry, VoL 3s. No. 2 pa 41349,M94 CqyrighcQ1994EbniaSdcmLtd PrintcdioGlalrbiuieAUIightarcmved 0031-942zp4 sdJ?o+0.00 FLAVONOID DIVERSITY AND METABOLISM IN 100 ROSA X HYBRIDA CULTIVARS JEAN-PHILIPPE BIOLLEY, MAURICE JAY and MARIE-ROSE VIRICEL Laboratoire de Biologic Micromolhlaire et Phytochimie I.A.S.B.S.E., UniversitC Claude Bernard, Lyon I 43, bd du 11 novembre 1918, F-69622 Villeurbanne Ccdex, France (Received in revised form 27 July 1993) IN HONOUR OF PROFESSOR JEFFREY HARBORNES SIXTY-FIFTH BIRTHDAY Key Word Index-Rosa X hybrida; Rosaceae; rose; flavonol glycosides; anthocyanins; chemotypes; biosynthesis. Abstract-The flavonoid metabolism in the petals of more than 100 cyanic cultivars of Rosa X hybrida was analysed by means of spectrophotometric and HPLC techniques. Total anthocyanin of ca 60 mg g- dry wt consisting of various mixtures of cyanidin 3,5-diglucoside and pelargonidin 3,Sdiglucoside were observed, while only pure cyanidin 3,5- diglucoside was found to accumulate in amounts above 60 mgg - 1 dry wt. Only small amount of the related 3- monoglucosides were detected and paeonidin 3,5diglucoside was rarely present. Cultivars producing almost exclusively kaempferol, as well as those dominated by quercetin exhibited chemotypes based mainly on the 3- glucosides, 4glucosides and 3-rhamnosides. The possible relative distributions, between cyanidin, pelargonidin, quercetin and kaempferol were determined revealing a range of apparently forbidden co-occurrences. With regard to our enzymatic knowledge of the flavonoid pathway established in other ornamental species, the metabolism of the rose flavonoids is discussed. INTRODUCTION Although many phytochemical reports have been de- voted to the carotenoids, ellagitannins and flavonoids of rose petals l. accurate investigations concerning the glycosidic flavonoids within a large collection of rose varieties are lacking; up to now, the relative distribution of fully identified flavonol glycosides is known for only three varieties 123. These flavonoids are particularly interesting in that they are involved in co-pigmentation 3 and are useful tools for varietal characterization. Thus, it was shown recently that flavonol glycoside HPLC patterns provide reliable fingerprints for dis- criminating red rose cultivars exhibiting similar aglycone patterns 4. In the present paper the distribution of anthocyanins will be described in a rose collection of cu 100 cuhivars using spectrophotometric and HPLC techniques. In the same way, the presence of flavonol glycosides will be used to define different chemotypes. The co-occurrence of flavonols and anthocyanidins will be discussed with respect to their B ring hydroxylation pattern. Finally, with regard to the current enzymatic knowledge estab- lished in carnation, a putative explanation for the rose flavonoid patterns will be suggested. RESULTS Anthocyanins Five well-known anthocyanins were identified as cyan- idin 3,Miglucoside, pelargonidin 3,5diglucoside, paeon- idin 3,5diglucoside, cyanidin 3-monoglucoside and pel- argonidin 3-monoglucoside. Paeonidin 3,5diglucoside was clearly demonstrated in only five varieties; it ac- counted for only 17% of the total anthocyanin pool in cultivars 535 and 566, while it exceeded 50% in three hybrids of Rosa wosa (716, 720 and 723), the species from which it was first isolated 5. Only 535 showed a rare anthocyanin pattern 63 based on the combination of paeonin (8 mgg- 1 dry wt), cyanin (19 mgg- i dry wt) and pelargonin (20 mg g- 1 dry wt), the other four cul- tivars exhibited a classical mixture of paeonidin with cyanidin exclusively. It is significant that apart from these five varieties, all others surveyed accumulated only cyanidin and pelar- gonidin glucosides, and these will be the main concern of the present paper. However, other compounds detected in trace amounts at 510 nm were visualized in the HPLC profiles, but were present in too small an amount to allow identification. If none of these minor compounds ex- 413 414 J.-P. BIOLLEY et al. ANTHOCYANINS (mgb petal dry W Fig. 1. Distribution of the 93 cyanic varieties (exclusively pig- mented with cyanidin and/or pelargonidin) in a plane defined by their petal anthocyanin concentration and by their ratio between cyanidin and pelargonidin. These average values were generally calculated from five individual flower analyses per variety. Some varieties are identified by their numerical code. ceeded 1% of the total anthocyanins, their relative dis- tributions were stable enough to be useful in cultivar recognition. From a quantitative viewpoint, depending on the variety, the total anthocyanin amount ranged from 1 to 120 mg g- 1 dry wt. All varieties accumulating anthocyan- ins of more than 60 mgg - petal dry wt were character- ized by a proportion of cyanidin glycosides higher than 80% of the total anthocyanins. Other varieties, whatever their specific anthocyanin with amounts up to 60 mg g- dry wt, showed a continuous mixture from 1 to ca 100% of pelargonidin or cyanidin reciprocally (Fig. 1). While numerous cultivars exhibited a pure cyanidin content, only variety 317 was an unmixed pelargonidin type. The diglucosides predominated over the related 3- monoglucosides which generally never exceeded 5% of the total anthocyanin. Only three cultivars accumulated more than 10% of these monoglucosylated anthocyanins. Thus, cyanidin 3-monoglucoside reached 25.5% of the total anthocyanins (9 mgg- dry wt) in cultivar 816, while pelargonidin 3-monoglucoside reached 10.4% (of 8 mgg- dry wt) in 330 and 15.6% (of46 mgg- dry wt) in 565. Flavonol glycosides Flavonol glycoside patterns were revealed by HPLC analysis. A global HPLC fingerprint for the whole collec- tion of Rosa varieties showed 21 stable and distinct peaks corresponding to O-conjugation of either kaempferol or quercetin, the two aglycones detected in the petals. As suggested by Asen 2, structurally different flavonol glycosides show very similar solubility properties. Thus, chromatographic peaks could result from the co-occur- rence of two or three molecules not separated by the HPLC system that was used: a more accurate approach showed that the 21 peaks corresponded to ca 30 different flavonol glycosides. Most of the flavonol glycosides that corresponded to the highest HPLC peaks were identified (Table 1) by means of classical methods. All the molecules fully identi- fied have been already reported 2, 3, 7, 83. HPLC photodiode array detection showed that some peaks contained more than one aglycone. Thus, it was possible to distinguish whether peak I1 contained mainly kaemp- ferol glycoside (1 I a), or quercetin (I I b). In the same way, from its characteristic UV spectrum, kaempferol4-gluco- side could be identified as the major constituent of peak 13. From the 21 selected peaks, a data matrix was oh- tained, each line or individual flower being defined by the relative amounts of flavonol glycosides, the sum of which was equal to 100. The distribution of these 21 phenolic Table I. Listing of rose petal flavonoids corresponding to the main HPLC peaks detected at 340 nm HPLC Aglycone peak no. Mixture (flavonol) Main glycosides 7 YCS 8 Yes 9 Yes IO No lla Yes llb No I2 No 13 Yes 14 No 16 No 17 No 18 No 19 No 20 No 21 No Qu Qu Km Qu Km Qu Qu Km Km Km Km Km Qu Km Km Qu 3glucoside Qu 3-galloylglucoside Km 3-rhamnosylglucoside Qu 3-arabinoside Km 3glucoside + Km 3-glucuronide Qu 3-rhamnoside Qu 4-glucoside Km 4-glucoside + Km 3-galloylglucoside Km 3-xyloside Km 3-arabinoside Km 3-rhamnoside Km substituted at 3-position Qu acylated with a cinnamic acid Km substituted at 3position Km acylated with a cinnamic acid Qu: quercetin; Km: kaempferol. Flavonoid diversity and metabolism in Rosa X hybrida components in the petals of 93 Rosa varieties has led to 93 fingerprints, each computed from three to five individual profiles. On the basis of these flavonol glycoside fingerprints the cultivars can be divided into three clusters depending on the kaempferol:quercetin ratio: (i) varieties rich in kaempferol glycosides (i.e. more than 90% of the total tlavonol glycosidesk (ii) varieties which accumulate mainly quercetin (more than 80% of the total flavonol glycosides); and finally (iii) cultivars characterized by the 1: 1 mixtures of kaempferol and quercetin derivatives. Results concerning the latter group will not be shown because all these cultivars contained a complex mixture of quercetin and kaempferol derivatives and exhibited intermediate chemotypes between the patterns that will be discussed below. These intermediates could be useful for varietal discrimination, but not for the description of the main regulation processes. The chemical data of the first two groups of cultivars will be separately treated by Principal Component Analysis (performed on centred data table). Organization of petal jiauonoid patterns based on kaemp feral glycosides The data matrix concerned average fingerprints of 50 cultivars accumulating almost exclusively kaempferol (more than 90% of the total accumulated flavonol glycos- ides): three white varieties (106, 112, 114), seven yellow varieties (205,207,209,211,215,216 and 229) and finally 40 cyanic cultivars. The ordination of the 50 cultivars according to the first two axes defined by P.C.A. (Fig. 2) accounted for 83.1% of the total inertia. The organization of the plant material is mainly ordered by three dominant HPLC peaks that always exhibited (within fingerprints), highest relative amounts: 1 la, 13 and 17 corresponding, respectively, to kaempferol 3-glucoside (possibly mixed with same kaempferol 3-glucuronide), kaempferol 4- glucoside and kaempferol 3-rhamnoside. Among the other 18 chemical features, only peak 09 (kaempferol 3- rhamnosylglucoside) stood out and participated weakly in the varietal make-up. Thus, it could be particularly well represented in some varieties accumulating large amounts of kaempferol 3-glucoside. Fig. 2. P.C.A. treatment (centred data table) of the petal flav- onol glycoside patterns revealed by HPLC. Ordering (Fl x F2 plane) of the 50 mean varietal patterns based on a dominant kaempferol (more than 90% of the total Aavonol glycosides). Each cluster is illustrated by some varietal chemotypes com- puted from profiles of all individuals analysed per variety. Each variety is identified by its numerical code. The corresponding ordination of the chemical variables that explains the varieties locations is also given (top right hand comer). Four clusters appeared. Cluster A, along the positive part of axis 2, contained cultivars with high relative amounts of kaempferol 4-glucoside (peak 13 exceeding 19% of the total flavonol glycoside pool). Cluster B was defined by the presence of more than 30% of kaempferol 3-glucoside (1 la), while varieties belonging to cluster C accumulated more than 31% kaempferol 3-rhamnoside (peak 17). Cluster D was made up of varieties which synthesized a number of different flavonoid glycosides. cultivars (two yellow: 223 and 239, and 20 anthocyanin pigmented varieties) dominated by quercetin glycosides (more than 80% of the total flavonol heterosides) were organized by P.C.A. The Fl x F2 ordination, accounting for 73.2% of the total variance, disclosed the main information contained in the chemical data (Fig. 3). The location of the various cultivars was determined by three peaks: 7 (quercetin 3-glucoside + an unidentified querce- tin glycoside), peak llb (quenxtin 3-rhamnoside) and peak 12 (quercetin 4-glucoside). From this analysis three major chemotypes can be described. Thus, cluster A corresponded to a chemotype based on high relative amounts of quercetin 4-glucoside (peak 12 accounting for not less than 17% of the total pool) and quercetin 3- rhamnoside (peak llb). Group B contained varieties characterized by a large per cent of peak 7 (up to 40% of the total in 717) representing two quercetin glycosides, among which only the major quercetin 3-glucoside was completely identified. Fingerprints of varieties belonging to cluster C were dominated by quercetin 3-rhamnoside (peak llb) that exceeded 39% of the total flavonol glycosides. Organization ojpetalj7avorwid patterns based on quercetin General aspects of the glycosylation process affecting glycosides pavono1s Varieties rich in quercetin (more than 90% of the total flavonol glycosides) were less numerous than those accu- mulating mainly kaempferol. Flavonoid patterns of 22 Within this structural diversity, a few molecules were seen to dominate the flavonoid fingerprints. Thus, what- ever the flavonol aglycone, the major glycosylation 416 J.-P. BIOLLEY et al. Fig. 3. P.C.A. treatment (centred data table) of the petal flav- onol glycoside patterns revealed by HPLC. Ordering (Fl x F2 plane) of the 22 mean varietal patterns based on a dominant quercetin (more than 80% of the total flavonol glycosides). Each cluster is illustrated by some varietal chemotypes computed from profiles of all individuals analysed per variety. Each variety is identified by its numerical code. The corresponding ordination of the chemical variables that explains the varieties locations is also given (bottom left hand corner). processes remained the same, i.e. 3-0-glucoside, 3-0- rhamnoside and 4-0-glucoside. In spite of these common metabolic traits, rich molecular diversity was attained, at a varietal level, by a phenolic regulation sensitive enough for easy and pertinent varietal recognition 4. Relationships between anthocyanins and jlavonols accord- ing to their B ring hydroxylation pattern The strong positive correlation between kaempferol and pelargonidin has already been suggested in cultivars of modern roses 9, IO, as well as in other ornamental species e.g. Antirrhinum majus I l or Petunia hybrida 123. Thanks to precise measurements of the relative amounts of the four aglycones, it has become possible to specify the variability within the biosynthetic activity of t c - - - - - - - - -, 100% 100% cy g:; pg Fig. 4. Flavonoid abilities of the 93 Rosa X hybrida cultivars pigmented with cyanidin and/or pelargonidin. Cy, Pg, Qu, Km correspond to cyanidin, pelargonidin. quercetin and kaempferol, respectively. Some varieties are identified by their numerical code. the flavonoid pathway (Fig. 4). Thus, varieties defined by less than 5% of pelargonidin (i.e. containing more than 95% of cyanidin) may contain from 5 to 48% of kaemp- feral. Among anthocyanins, regardless of cultivar, when pelargonidin exceeded 20%, the associated kaempferol level was always more than 85% of the total flavonols. Obviously, a significant per cent of pelargonidin (among anthocyanins) was never associated with a large quantity of quercetin (among flavonols). From the quantitative flavonoid analysis of the varietal collection, five obvious chemotypes can be discerned (Table 2). DISClJSlON For Rosa X hybrido varieties, there is a characteristic metabolic scheme that strongly regulates the co-occur- rences of flavonols and anthocyanidins according to their respective B ring hydroxylation patterns. Flavonoid com- position may be different from one ornamental species to another, even if the correlation between kaempferol and Table 2. Distribution of flavonoids in five varieties representative of the five chemotypes of Rosa X hybrido Anthocyanins mgg-r petal Variety dry wt Cyanidin % Pelargonidin % flavonol glycosides Quercetin mgg-r dry wt % Kaempferol % Metabolic type 532 19.0*0.7* 546 89.5 f 8.0 429 20.9k3.5 317 13.5* 1.7 419 15.8 f 0.4 78.2 k 3.5 99.5 +0.2 99.4f0.5 0.3 * 0.2 56.7 + I.5 21.8_+3.5 0.5*0.2 0.6 f 0.5 99.7kO.2 43.3 f I.5 25.6k3.8 43.8 f 5.5 36.7k5.1 30.6k2.6 10.2* 1.0 9.452.1 91.7*0.9 52.1 kO.2 0.2kO.l 1.1 kO.7 90.6+2.1 8.3 kO.9 47.9kO.2 99.8k0.1 98.9 kO.7 Cy-Km CY-Qu Cy-Qu-Km Pg-Km Cy-Pg-Km *Value + SD. Cy: Cyanidin; Km: kaempferol; Pg: pelargonidin; Qu: quercetin. Flavonoid diversity ar,d m-tabolism in Rosa X hybridu 417 pelargonidin is maintained. Thus, within the genus Rho- deron, the lack of iargonidin is reflected by kaemp feral derivatives which never exceed 49% of the total flavonols 133. For Petunia X hyridu, the same lack is explained by the substrate specificity of the dihydroflav- onol4-reductase (DFR) which does not recognize dihy drokaempferol as a substrate, probably because the iso- enzyme encoded by the only gene (among three genes of DFR) transcribed in the flower is not consistent with this intermediate 14, 153. However, in terms of comparitive abilities, Rhododendron and Petunia are not the best models, especially because they have a functional B ring tri-hydroxylation route that does not exist in Rosa. More suitable for comparison are carnations and roses; indeed cyanic cultivars of Diathus cyophyll, which do not accumulate chalcone glycosides, synthesize the same four flavonoids as Rosa hybridu. A flavonoid over- view of more than 100 cyanic carnation cultivars showed that, whatever their anthocyanin pattern, all varieties were characterixed by kaempferol as the major consti- tuent (more than 75% of the total flavonols). This rule held for all but one cultivar, which was pigmented exclusively with cyanidin and exhibited quercetin and kaempferol in a 1: 1 ratio S. The co-occurrence of pelargonidin and cyanidin is exceptional in carnations, while it is frequent in roses. In the same way, carnation cultivars with only pelargonidin are numerous, whereas in Rosa X hybrida cultivars with only cyanidin are common. Furthermore, an enzymatic approach demonstrated that the accumulation of C-34disubstituted flavonoids in carnations was due to a C-3-hydroxylation process at the C-15 stage mediated by a (F3H) flavonoid 3-hydrox- ylase 17,18. More recently, Stich ef al. 19 showed that the reduction of dihydroflavonols to leucoanthocyani- dins was catalysed by the dihydroflavonol 4-reductase (DFR). Dihydrokaempferol and dihydroquercetin are both suitable subst