Comprehensive Flavonol Profiling and Transcriptome Coexpression Analysis Leading to Decoding Gene–Metabolite Correlations in Arabidopsis^,

Abstract

To complete the metabolic map for an entire class of compounds, it is essential to identify gene–metabolite correlations of a metabolic pathway. We used liquid chromatography–mass spectrometry (LC-MS) to identify the flavonoids produced by Arabidopsis thaliana wild-type and flavonoid biosynthetic mutant lines. The structures of 15 newly identified and eight known flavonols were deduced by LC-MS profiling of these mutants. Candidate genes presumably involved in the flavonoid pathway were delimited by transcriptome coexpression network analysis using public databases, leading to the detailed analysis of two flavonoid pathway genes, UGT78D3 (At5g17030) and RHM1 (At1g78570). The levels of flavonol 3-O-arabinosides were reduced in ugt78d3 knockdown mutants, suggesting that UGT78D3 is a flavonol arabinosyltransferase. Recombinant UGT78D3 protein could convert quercetin to quercetin 3-O-arabinoside. The strict substrate specificity of UGT78D3 for flavonol aglycones and UDP-arabinose indicate that UGT78D3 is a flavonol arabinosyltransferase. A comparison of flavonol profile in RHM knockout mutants indicated that RHM1 plays a major role in supplying UDP-rhamnose for flavonol modification. The rate of flavonol 3-O-glycosylation is more affected than those of 7-O-glycosylation by the supply of UDP-rhamnose. The precise identification of flavonoids in conjunction with transcriptomics thus led to the identification of a gene function and a more complete understanding of a plant metabolic network.

INTRODUCTION

Plants employ many diverse metabolic pathways to produce >200,000 compounds (Dixon and Strack, 2003). Completion of the Arabidopsis thaliana and rice (Oryza sativa) genome sequencing projects has allowed (1) the annotation of genes involved in metabolic pathways based on homology and (2) genome-wide identification of whole metabolic pathways. Assigning functions to each of the genes that correlate with metabolites in a pathway is a prerequisite for developing a comprehensive understanding of complicated metabolic pathways. Because of its small genome and the development of extensive genetic, molecular, and physiological tools, Arabidopsis may be the only plant in which the complete network of secondary metabolism can currently be characterized within a single plant species.

The flavonoids comprise one of the major secondary metabolite groups, with >7000 known compounds (Harborne et al., 1999; Andersen and Markham, 2006). Formation of the basic flavonoid skeleton has been well studied in terms of molecular biology and natural products chemistry. Flavonol and anthocyanidin regulatory and biosynthetic pathways have been characterized, and the corresponding genes have been isolated from various plants (Figure 1; Andersen and Markham, 2006; Tanaka and Filippa, 2006). However, the pathways for sequential modification, such as glycosylation, acylation, and methylation, are still relatively unexplored even though modification produces a huge chemical diversity and is essential for the stable accumulation of flavonoids. A wide variety of flavonoids have been isolated and identified from specific organs, such as flowers or fruits (Andersen and Markham, 2006), but there has been no comprehensive description of flavonoid metabolism in the organs of a single plant species. Although the flavonoids are important as flower pigments, UV-B protectants, phytoalexins, signaling molecules, and regulators of auxin transport (Dooner et al., 1991; Dixon and Paiva, 1995), the precise relationship between structure and function is also largely unclear.

Figure 1.

The Arabidopsis Flavonoid Biosynthetic Pathway.

4CL, 4-coumarate:CoA ligase; CHS, chalcone synthase; CHI, chalcone isomerase; F3H, flavanone 3-hydroxylase; F3′H, flavonoid 3′-hydroxylase; FLS, flavonol synthase; OMT1, O-methyltransferase; UGT, UDP-dependent glycosyltransferase; DFR, dihydroflavonol 4-reductase; LDOX/ANS, leucoanthocyanidin dioxygenase/anthocyanidin synthase; AAT, anthocyanin acyltransferase.

In Arabidopsis, at least 13 structures of flavonoids have been identified by nuclear magnetic resonance (NMR) and/or liquid chromatography–mass spectrometry (LC-MS ) with standards, and more flavonoid compounds were reported to be present (Graham, 1998; Veit and Pauli, 1999; Bloor and Abrahams, 2002; Kerhoas et al., 2006). Thirty-five flavonoid biosynthetic genes, including genes encoding 12 transcription factors, 10 structural enzymes, and 10 modification enzymes, have been identified (Tohge et al., 2005, 2007; Lepiniec et al., 2006; Fraser et al., 2007; Luo et al., 2007; Stracke et al., 2007; Yonekura-Sakakibara et al., 2007), but even in this highly studied species, the full details of the flavonoid pathway, including modification, are still unknown. The major flavonoid structures in Arabidopsis suggest that there are as yet unidentified modification enzyme genes and pathways in flavonoid metabolism. Generally, modification enzymes are encoded by multigene families. For example, there are 107 UDP-dependent glycosyltransferase (UGT) genes in Arabidopsis (D'Auria and Gershenzon, 2005). The presence of multiple genes hampers efforts to determine precise gene functions by sequence homology and to draw a complete pathway, even in Arabidopsis. The presence of unidentified minor flavonoids also suggests additional missing genes and pathways (Tohge et al., 2007).

Because of the rich research resources available for Arabidopsis, it is the best candidate for cataloging the entire flavonoid pathway. The integration of metabolome analysis with transcriptome analysis has been successfully used in Arabidopsis to decipher gene functions in biological processes (Hirai et al., 2005, 2007; Tohge et al., 2005). Transcriptome coexpression analysis is also a powerful tool for the efficient targeting of genes within a multigene family (Yonekura-Sakakibara et al., 2007; Saito et al., 2008). Use of these novel tools in combination with reverse genetics and a biochemical approach using recombinant proteins can provide robust evidence from which a whole pathway can be constructed.

In this study, we attempted to complete the identification of flavonols in each of the organs and flavonol annotation using flavonoid mutant lines. Taken together with previous reports, a total of 32 flavonol end or intermediate products were listed. Thirty of 32 flavonols were detected, and the structures of 19 compounds were annotated. Transcriptome coexpression analysis using known flavonoid genes as queries suggested 139 genes that may be involved in flavonoid metabolism. Finally, integration of comprehensive flavonol identification/annotation with transcriptome coexpression analysis identified a gene encoding flavonol 3-O-arabinosyltransferase and revealed the physiological role of a gene coding for UDP-rhamnose synthase in flavonoid biosynthesis.

RESULTS

Comprehensive Flavonol Profiling in Wild-Type and tt4 Mutants

To catalog all of the flavonol derivatives in Arabidopsis, flavonol profiling in the flowers, leaves, stems, and roots of the wild type and tt4 mutants were performed using ultraperformance liquid chromatography (UPLC)–photodiode array (PDA)–electrospray ionization (ESI)/quadrupole time-of-flight (Q-TOF)/MS (Figure 2A). The metabolites detected only in the wild type should be flavonoid derivatives because TT4 encodes chalcone synthase, the first committed enzyme in flavonoid biosynthesis, and no flavonoid derivatives were detected in the tt4 mutant (Shirley et al., 1995; von Roepenack-Lahaye et al., 2004).

Figure 2.

UPLC-PDA-MS Analyses of the Extracts from Leaves, Flowers, Roots, and Stems of Wild-Type (Col-0) and tt4 Plants.

(A) UPLC-PDA and mass chromatograms of aqueous methanol extracts of the Arabidopsis wild type and the tt4 knockout mutant. Absorbance at 320 nm was used for the detection of flavonols. Labels correspond to compounds shown in Figure 3.

(B) Principal component analyses of the data obtained by LC-MS. Proportions of the first and second components are in parentheses. Wild-type results are in red, and those from the tt4 mutant are in blue.

The flavonols were differentially distributed in Arabidopsis organs (Figure 3). Kaempferol derivatives (f1 to f3 in Figure 3) were the major flavonols in leaves, stems, and flowers. In roots, there were high levels of quercetin 3-O-glucoside-7-O-rhamnoside (f6) in addition to kaempferol glycosides (f2 and f3), but there was very little kaempferol 3-O-rhamnoside-7-O-rhamnoside (f1). As for other phenolic compounds, a sinapoyl derivative s2 was detected predominantly in the leaves and stems of both the wild type and tt4.

Figure 3.

Flavonol Glycosides in Arabidopsis.

Asterisks indicate that the compounds were identified based on a comparison of retention times and UV/mass spectra of the standards used in this study. R₁=H, kaempferol; R₁=OH, quercetin; R₁=OMe, isorhamnetin. The presence of a compound in the indicated tissue is denoted by a gray box.

The MS-detectable peaks in the wild type and tt4 were compared to extract all flavonoid-related peaks (see Supplemental Figure 1 and Supplemental Data Set 1 online). Principal component analysis was conducted with 1804 MS peaks in leaves, 1864 in flowers, 1576 in roots, and 2138 in stems, indicating the presence of clear clusters according to the wild-type and tt4 genotypes (Figure 2B). The wild-type–specific peaks (53 peaks in leaves, 322 in flowers, 152 in roots, and 75 in stems) were picked up using a high PC1 loading value, and most of the peaks are due to the ions of flavonols (see Supplemental Figure 1 online). For example, 42 of 53 peaks are three major kaempferol glycosides in leaves (f1 to f3). A total of 30 flavonol derivatives were detected in Arabidopsis (four in leaves, 25 in flowers, 14 in roots, and four in stems). Flowers, in which most flavonol derivatives were detected, were used for further analysis.

Peak Annotation by Comparison with Flavonoid Biosynthetic Gene Knockout Mutants

Twelve flavonols (f1 to f8, f17, f18, f20, and f24) in Arabidopsis have been identified by UV spectroscopy, MS, ¹H-NMR, and ¹³C-NMR spectroscopy or annotated using UPLC-PDA-ESI/Q-TOF/MS (Veit and Pauli, 1999; Jones et al., 2003; Tohge et al., 2005, 2007; Kerhoas et al., 2006). We identified two additional flavonols (f16 and f28) by comparing retention times and mass-to-charge ratio (m/z) values with the standard compounds (Table 1; see Supplemental Data Set 2 online). To annotate the remaining 18 uncharacterized flavonols, mutants lacking CHS, F3′H, flavonol 3′-O-methyltransferase (OMT1), flavonoid 3-O-glucosyltransferase (UGT78D2), flavonol 3-O-rhamnosyltransferase (UGT78D1), flavonol 7-O-rhamnosyltransferase (UGT89C1), or anthocyanin 5-O-glucosyltransferase (UGT75C1) were used for flavonol analysis. Of the 14 identified (f1 to f8, f16 to f18, f20, f24, and f28), flavonol derivatives missing from each mutant were well correlated with the loss of gene function (Figure 4, Table 1). No quercetin and isorhamnetin conjugates (f5, f6, f8, f16, f24, and f28) accumulated in tt7, which lacks F3′H. No isorhamnetin conjugates (f28) were detected in omt1 mutants. The mutants ugt78d1 and ugt89c1 had no conjugates of flavonol 3-O-rhamnoside (f1, f5, f17, f20, and f24) and flavonol 7-O-rhamnoside (f1 to f3, f5, f6, and f8), respectively. Interestingly, in the ugt78d2 mutant, accumulation of the kaempferol/quercetin 3-O-glucosides (f2, f3, f6, and f8) was drastically reduced but detectable. Other kaempferol/quercetin 3-O-glucosides (f16 and f18) also decreased. This result is in agreement with the finding that cyanidin derivatives were detected at low levels in the ugt78d2 mutant (Tohge et al., 2005). On the other hand, accumulation of the isorhamnetin 3-O-glucosides (f28) was similar or slightly increased. These data indicate that another enzyme is involved in 3-O-glucosylation of isorhamnetin. Flavonol profiles in the flowers of the wild type and ugt75c1 were essentially identical, demonstrating that UGT75C1 is not involved in flavonol glycosylation at any detectable level.

Table 1.

The Flavonol Profiles in Acidic Methanol-Water Flower Extracts of Wild-Type Plants and Flavonoid Mutant Lines

Peak No.	Rt (min)	ESI-MS (m/z)	Fragment (m/z)a	Col-0	tt4	tt7	omt1	ugt78d2 (Fd3GlcT)	ugt78d1 (F3RhaT)	ugt89c1 (F7RhaT)	ugt75c1 (A5GlcT)
f1b	5.38	579	433, 287	5.14	ND	6.05	5.71	5.61	ND	ND	4.63
f2b	4.88	595	433, 287	3.33	ND	4.87	3.96	0.19	4.25	ND	2.94
f3b	4.14	741	595, 449, 287	6.47	ND	7.75	7.49	0.29	7.33	ND	5.93
f5b	4.94	595	449, 303	2.09	ND	ND	4.14	2.04	ND	ND	1.86
f6b	4.47	611	449, 303	2.16	ND	ND	3.51	0.12	2.54	ND	1.97
f8b	3.88	757	611, 449, 303	1.11	ND	ND	2.37	0.04	1.58	ND	1.09
f14	4.99	625	463, 317	2.71	ND	ND	ND	2.08	3.47	ND	2.51
f15	4.49	625	479, 317	0.05	ND	ND	ND	0.41	ND	1.12	0.05
f16b	5.33	465	303	0.07	ND	ND	0.12	ND	0.10	2.36	0.08
f17b	6.52	433	287	0.03	ND	0.13	0.04	0.02	ND	3.54	0.02
f18b	5.84	449	287	0.08	ND	0.19	0.06	ND	0.10	3.18	0.05
f19	4.95	565	433, 287	0.07	ND	0.28	0.11	0.66	0.13	ND	0.08
f20	4.34	595	449, 287	0.10	ND	0.32	0.18	0.14	ND	1.96	0.09
f21	5.02	611	449, 287	2.37	ND	3.27	3.10	ND	2.07	3.00	1.97
f22	4.83	711	565, 433, 287	0.02	ND	0.03	0.03	0.18	0.02	ND	0.02
f23	5.72	435	303	0.01	ND	ND	0.02	0.02	0.01	0.50	0.01
f24b	5.92	449	303	0.05	ND	ND	0.14	0.07	ND	2.09	0.06
f25	4.66	581	449, 303	0.36	ND	ND	0.72	1.16	0.40	ND	0.33
f26	4.60	627	465, 303	1.66	ND	ND	1.86	0.06	1.74	1.77	1.35
f27	6.64	463	317	0.05	ND	ND	ND	0.07	ND	1.89	0.04
f28b	5.97	479	317	0.14	ND	ND	ND	0.15	0.16	2.99	0.15
f29	5.17	595	463, 317	0.25	ND	ND	ND	2.46	0.34	ND	0.24
f30	5.50	609	463, 317	2.09	ND	ND	ND	4.38	ND	ND	1.83
f31	5.07	641	479, 317	Trace	ND	ND	Trace	0.11	0.01	0.01	0.01
f32	4.27	771	625, 463, 317	0.03	ND	ND	ND	0.03	0.05	ND	0.05
s1b	4.00	387	207	0.06	0.12	0.07	0.02	0.07	0.05	0.07	0.05
s2b	5.46	341	207	0.08	0.21	0.08	0.03	0.10	0.10	0.12	0.07

Flavonols were quantified by measuring mass peak area using a response value of the peak area of an internal standard (naringenin 7-O-glucoside). ND, not detected; trace, <0.01; s1 and s2, sinapoyl derivatives.

^aDetected in mass and/or tandem mass data.

^bIdentified based on a comparison of retention times and UV/mass spectra of the standards.

Figure 4.

UPLC-PDA-MS Analyses of the Extracts from Flowers of Wild-Type (Col-0) and Flavonoid-Defective Mutants.

Absorbance at 320 nm was used for detection. A5GlcT, anthocyanin 5-O-glucosyltransferase; F3RhaT, flavonol 3-O-rhamnosyltransferase; F7RhaT, flavonol 7-O-rhamnosyltransferase; Fd3GlcT, flavonoid 3-O-glucosyltransferase. Labels correspond to compounds shown in Figure 3. The labels in parentheses indicate compounds at <5% in the peak.

The basic flavonol structures and glycosylation patterns were revealed by superimposing the flavonol profiles of mutants over the wild type (Table 1). The structures of additional nine flavonols, including pentosides (f14, f15, f19, f23, f25, f27, f29, f30, and f32), were identified in Arabidopsis. A total of 23 flavonol structures and their distribution in the mutants show that at least four flavonol glycosyltransferases, an additional flavonol 3-O-glucosyltransferase, a flavonol 3-O-pentosyltransferase, a flavonol 3-O-glycoside:2′′-O-rhamnosyltransferase, and a flavonol 3-O-glycoside:6′′-O-glucosyltransferase remain to be identified.

New Flavonoid Biosynthetic Genes Deduced by Transcriptome Coexpression Analysis

Transcriptome coexpression analysis (Saito et al., 2008) was used to identify all of the flavonoid-related genes and was executed with the 24 genes encoding flavonoid biosynthetic enzymes or transcription factors listed in Table 2 as query. The lowest correlation coefficient (r) between query genes, except for the flavonol 7-O-glucosyltransferase gene (UGT73C6) and OMT1 is 0.527. UGT73C6 and OMT1 were excluded because UGT73C6 may cross-hybridize with UGT73C5, a brassinosteroid-related UGT, and the OMT1 coexpression profile suggests that it is under the regulation of lignin biosynthesis but not of flavonoid biosynthesis. The coexpression relationships of all Arabidopsis genes exhibiting correlation coefficients > 0.525 (r > 0.525, all data sets version 3) with the query genes are shown in Supplemental Figure 2 online and Table 3. The cutoff value was set according to Aoki et al. (2007). The clusters of coexpressed flavonoid genes were classified into two major groups. One cluster contains the general flavonoid/flavonol subgroup; another cluster represents the anthocyanin subgroup. UGT73C6 and OMT1 did not belong to either of these clusters. The 139 genes were highly correlated with the general flavonoid-related or anthocyanin-related genes (see Supplemental Figure 2 online). Twenty-four genes had more than two correlations with flavonoid-related genes. One-third of the genes listed in Table 3 were coregulated with flavonoid-related genes by flavonoid-specific MYBs (PAP1, MYB11, MYB12, and MYB111; Tohge et al., 2005, Stracke et al., 2007). The genes encoding embryo-specific protein (At5g62210), 3-keto-acyl-CoA thiolase (At5g48880), cryptochrome 3 (At5g24850), UGT84A1 (At4g15480), cinnamoyl-CoA reductase-like protein (At2g23910), early light-induced protein (At3g22840), TT8 (At4g09820), and glutathione S-transferase (GST; At1g10370) were regulated in a manner similar to the flavonoid biosynthetic genes by high irradiance and blue light in a CRYPTOCHROME1 (CRY1) photoreceptor and/or LONG HYPOCOTYL5 (HY5) transcription factor-dependent manner (Kleine et al., 2007). Other genes were also regulated by HY5 under UV-B radiation, much like the flavonoid biosynthetic genes (Table 3; Brown et al., 2005). For further analysis, we focused on UGT78D3 as a putative flavonol glycosyltransferase and RHM1, which was selected from the genes highly correlated with the flavonol biosynthetic genes.

Table 2.

Query Genes Used for Transcriptome Coexpression Analysis

Name	AGI No.	Gene Annotation
4CL3	At1g65060	4-Coumarate:CoA ligase
CHS, TT4	At5g13930	Chalcone synthase
CHI, TT5	At3g55120	Chalcone isomerase
F3H, TT6	At3g51240	Flavanone 3-hydroxylase
F3′H, TT7	At5g07990	Flavonoid 3′-hydroxylase
FLS1	At5g08640	Flavonol synthase
DFR, TT3	At5g42800	Dihydroflavonol reductase
ANS, TT18	At4g22880	Anthocyanidin synthase
GST, TT19	At5g17220	GST
UGT73C6	At2g36790	Flavonol 7-O-glucosyltransferase
UGT75C1	At4g14090	Anthocyanin 5-O-glucosyltransferase
UGT78D1	At1g30530	Flavonol 3-O-rhamnosyltransferase
UGT78D2	At5g17050	Flavonoid 3-O-glucosyltransferase
UGT89C1	At1g06000	Flavonol 7-O-rhamnosyltransferase
A5G6′′′MaT	At3g29590	Anthocyanin 5-O-glucoside:6′′′-O-malonyltransferase
A3G6′′p-CouT	At1g03940	Anthocyanin 3-O-glucoside:6′′-O-p-coumaroyltransferase
A3G6′′p-CouT	At1g03495	Anthocyanin 3-O-glucoside:6′′-O-p-coumaroyltransferase
SCPL10	At2g23000	Sinapoylglucose:anthocyanin acyltransferase
OMT1	At5g54160	Caffeic acid/5-hydroxyferulic acid O-methyltransferase, flavonol 3′-O-methyltransferase
MYB11	At3g62610	R2R3 MYB protein
MYB12	At2g47460	R2R3 MYB protein
MYB111	At5g49330	R2R3 MYB protein
PAP1, MYB75	At1g56650	R2R3 MYB protein
PAP2, MYB90	At1g66390	R2R3 MYB protein

Table 3.

The Genes Coexpressed with Those in Flavonoid Metabolism

Correlated Gene No.	AGI No.		PAP1	MYB11 MYB12 MYB111	CRY1 HY5
	Flavonoid/phenylpropanoid biosynthesis
9	At5g05270	Chalcone isomerase family	+	+	+
7	At5g54060	UDP-glucosyltransferase, UGT79B1	+
5	At4g15480	UDP-glucosyltransferase, UGT84A1		+	+
3	At2g23910	Cinnamoyl-CoA reductase-like, CCRL9		+	+
	Transporter
6	At5g17010	Sugar transporter family
2	At1g10370	Glutathione S-transferase, GST30			+
2	At5g02270	ABC transporter, AtNAP9			+
2	At5g44110	ABC transporter, AtNAP2			+
	Protein kinase
2	At1g10850	Ser/Thr kinase family
2	At3g56370	Leu-rich repeat transmembrane protein kinase
	Transcription factor
2	At2g37260	WRKY transcription factor, TTG2	+
2	At4g09820	Basic helix-loop-helix family, TT8	+		+
2	At5g60140	Transcriptional factor B3 family
	Others
8	At5g62210	Embryo-specific protein		+	+
7	At1g65560	Allyl alcohol dehydrogenase
7	At5g48880	3-Keto-acyl-CoA thiolase, KAT5		+	+
6	At1g78570	Rhamnose synthase, RHM1
5	At5g24850	Cryptochrome 3, CRY3			+
4	At1g06550	Enoyl-CoA hydratase/isomerase family
3	At5g20070	Nudix hydrolase homolog, ATNUDT19
2	At1g36160	Acetyl-CoA carboxylase
2	At1g72970	HOTHEAD
2	At3g22840	Early light-induced protein1, ELIP1			+
2	At5g39220	Hydrolase family

Only the genes that correlated with at least two genes are shown. The plus sign means the similar regulatory behavior with flavonoid biosynthetic genes in knockout mutants (CRY1/HY5, HY5 or MYB11/MYB12/MYB111) or an overexpressed plant (PAP1).

UGT78D3 Encodes Flavonol 3-O-Arabinosyltransferase

UGT78D3 is correlated with MYB111 (r = 0.572) but not with the other query genes. However, UGT78D3 belongs to the UGT78D subfamily, which also contains flavonoid 3-O-glucosyltransferase (UGT78D2) and flavonol 3-O-rhamnosyltransferase (UGT78D1), suggesting that UGT78D3 encodes a flavonoid 3-O-glycosyltransferase. UGT78D3 has high sequence identity with UGT78D2 (75.7% at the amino acid level) and UGT78D1 (68.3%). UGT78D2 has an amino acid sequence identity with UGT78D1 of 71.7%.

To test our hypothesis, T-DNA insertion mutant SALK_114099, designated as ugt78d3, was used for reverse genetics analysis. T-DNA was inserted between −122 and −100 bp upstream from the start codon of UGT78D3 (Figure 5A). UGT78D3 transcripts were abundant in floral buds but present at low levels in siliques and stems (Figure 5B). Real-time PCR showed that the accumulation level of UGT78D3 transcripts in flowers of the homozygous ugt78d3 mutant decreased to 30% of those in the wild type (Figure 5C). Flavonol target analysis revealed that three flavonol 3-O-pentoside conjugates (f19, f25, and f29) were under the detection limit in flowers of the ugt78d3 mutant (Figure 5D). This flavonol 3-O-pentoside conjugate-deficient phenotype was complemented by the constitutive expression of UGT78D3 cDNA under the control of the cauliflower mosaic virus 35S promoter (Figure 5D), indicating that UGT78D3 encodes a flavonol 3-O-pentosyltransferase.

Figure 5.

A T-DNA Insertion Mutant of UGT78D3.

(A) Schematic representation of UGT78D3 with a T-DNA insertion mutant used in this work. The thick black line indicates coding sequence. Numbers indicate the position of the T-DNA insertion. The gray box adjacent to T-DNA indicates the region containing the UGT78D3 promoter from −77 bp to +4 bp. LB, left border; RB, right border.

(B) Real-time PCR analysis of UGT78D3 transcripts in organs of the Arabidopsis wild type (Col-0). Error bars represent sd of three technical replicates per sample.

(C) Real-time PCR analysis of UGT78D3 transcripts in wild-type (Col-0) and ugt78d3 mutant flowers. Error bars represent sd of three technical replicates per sample.

(D) Extracted fragment mass chromatograms of aqueous methanol extracts from flowers of the wild type (Col-0), the ugt78d3 mutant, and ugt78d3 complemented with p35S:UGT78D3. The extracted fragment mass ions m/z 565, m/z 581, and m/z 595 correspond to compounds f19, f25, and f29 in Figure 3, respectively.

Flavonoid analysis of ugt78d3 mutants provided no additional information on the structure of flavonol 3-pentoside conjugates (f19, f25, and f29). Therefore, a biochemical approach was used to identify the exact structure of the pentose at the C-3 position. UGT78D3 recombinant protein was produced in Escherichia coli as a GST fusion and purified. The sugar donor specificity of UGT78D3 was examined with UDP-arabinose and UDP-xylose as UDP-pentoses. UDP-galactose, UDP-glucose, and UDP-rhamnose were also tested. The GST-UGT78D3 fusion protein catalyzed the conversion of quercetin to a quercetin 3-O-arabinoside, as confirmed by UPLC retention time, UV absorbance, and MS (Figure 6). No UGT activity was observed with UDP-xylose, UDP-galactose, UDP-glucose, or UDP-rhamnose as substrates, indicating that UGT78D3 is specific for UDP-arabinose as a sugar donor (Table 4). The sugar acceptor specificity of UGT78D3 was examined with kaempferol, quercetin, myricetin, isorhamnetin, cyanidin, and their glycosides as substrates. UGT78D3 activity was specific for flavonol aglycones (kaempferol, quercetin, myricetin, and isorhamnetin) and kaempferol 7-O-rhamnoside but had no activity on cyanidin or the flavonoid 3-O-glycosides. These results indicate that UGT78D3 arabinosylates flavonols at the C-3 position.

Figure 6.

UPLC Analyses of the Reaction Products of UGT78D3 Recombinant Protein.

Elution profiles of reaction products of GST protein (control) or GST-fused UGT78D3 protein (UGT78D3) and the standards (quercetin 3-O-arabinoside) are shown.

Table 4.

Substrate Specificity of Arabidopsis UGT78D3

	Relative Activity (%)
Sugar donora
UDP-arabinose	100.0 ± 10.0
UDP-glucose	ND
UDP-rhamnose	ND
UDP-xylose	ND
UDP-galactose	ND
Sugar acceptorb
Kaempferol (Kae)	100.0 ± 8.5
Kae 3-O-glucoside	ND
Kae 7-O-rhamnoside	140.4 ± 13.9
Quercetin (Que)	282.2 ± 9.6
Que 3-O-glucoside	ND
Que 3-O-rhamnoside	ND
Myricetin	148.2 ± 2.0
Isorhamnetin	164.9 ± 34.4
Cyanidin (Cya)	ND
Cya 3-O-glucoside	ND
Cya 3-O-rhamnoside	ND
Cya 3,5-O-diglucoside	ND

ND, not detected.

^aThe reactions were performed with kaempferol as the sugar acceptor.

^bThe reactions were performed with UDP-arabinose as the sugar donor.

To confirm UGT78D3 function in planta, the structures of f19 and f25 were confirmed by direct comparison with the enzymatically synthesized standard compounds. Kaempferol 3-O-arabinoside-7-O-rhamnoside was prepared by the reaction catalyzed by UGT89C1 with kaempferol 3-O-α-l-arabinoside and UDP-β-l-rhamnose as substrates. An enzymatic product (1.8 mg) was isolated by multistep chromatography, and its structure was determined by ¹H-NMR and rotating frame overhauser enhancement (ROE). The ROE correlation of protons at C-6 (δ 6.43) and C-8 (δ 6.73) by the irradiation of anomeric proton of rhamnose (rhamnose H-1, δ 5.52) indicated that rhamnose was attached at the C-7 position of kaempferol (Mulinacci et al., 1995). The chromatographic behavior and MS data of kaempferol 3-O-l-arabinoside-7-O-l-rhamnoside were identical with those of f19 and enzymatically synthesized products by UGT78D3 with kaempferol 7-O-α-l-rhamnoside and UDP-β-l-arabinose (Figure 7). These results indicated that f19 is kaempferol 3-O-l-arabinoside-7-O-l-rhamnoside. The component of f25 was also identified to be quercetin 3-O-l-arabinoside-7-O-l-rhamnoside by the similar procedure.

Figure 7.

UPLC-MS Analyses of the Reaction Products of UGT78D3 or UGT89C1 Recombinant Protein.

Extracted fragment mass chromatograms of aqueous methanol extracts from flowers of wild type (Col-0) ([A] and [F]), the reaction products of UGT89C1 with kaempferol 3-O-arabinoside (B), (A) coeluted with (B) (C), the reaction products of UGT78D3 with kaempferol 7-O-rhamnoside (D), (A) coeluted with (D) (E), the reaction products of UGT89C1 with quercetin 3-O-arabinoside (G), and (F) coeluted with (G) (H). The extracted fragment mass ions m/z 565 ([A] to [E]) and m/z 581 ([F] to [H]) correspond to compounds f19 and f25 in Figure 3, respectively. Kae 3-Ara, kaempferol 3-O-arabinoside; Kae 7-Rha, kaempferol 7-O-rhamnoside; Que 3-Ara, quercetin 3-O-arabinoside.

The UDP-Rhamnose Synthase 1 Gene Plays a Role in Rhamnosylation of Flavonoids

In Arabidopsis, there are three UDP-rhamnose synthase genes (RHM1, RHM2, and RHM3) (Usadel et al., 2004; Western et al., 2004; Oka et al., 2007). Expression of the UDP-Rhamnose Synthase 1 (RHM1) gene was highly correlated with the general flavonoid pathway genes, for example, 4CL3 (r = 0.736), FLS1 (r = 0.696), UGT89C1 (r = 0.669), CHI (r = 0.647), F3H (r = 0.611), and CHS (r = 0.603), based on data from all experiments. RHM2/MUM4 and RHM3 expression profiles differ from the flavonoid biosynthetic genes (r < 0.4). The accumulation of RHM1-RHM3 transcripts in the target organs was analyzed by real-time PCR (Figure 8A). Major flavonol glycosides were predominantly accumulated in floral buds and flowers (Figure 2; Yonekura-Sakakibara et al., 2007), which is consistent with accumulation pattern of RHM1 transcripts. RHM1 transcripts also made up the bulk of RHM transcript accumulation in all of the tested organs except for siliques. In siliques, the transcripts of RHM2/MUM4, which are involved in seed mucilage pectin synthesis (Usadel et al., 2004; Western et al., 2004), were predominantly accumulated.

Figure 8.

T-DNA Insertion Mutants of RHM1 and RHM2, and UPLC-MS Analyses of the rhm Mutant Lines.

(A) Real-time PCR analysis of RHMs transcripts in Arabidopsis organs. Error bars represent sd of three technical replicates per sample.

(B) Schematic representation of RHM1 and RHM2 with the mutations used in this work. rol1-1, a nonsense mutation at position 318; rol1-2, a missense mutation changing an Arg at position 283 to a Lys; rhm2-1 and rhm2-3, T-DNA insertion mutants. LB/RB, left/right borders. Numbers indicate the position of T-DNA insertion.

(C) UPLC-PDA-MS analyses of the extracts from leaves and flowers in the wild type (Col-0) and the rhm mutants. Absorbance at 320 nm was used for detection. Labels correspond to compounds shown in Figure 3. The labels in parentheses indicate compounds at <5% in the peak.

(D) The ratio of flavonol derivatives to total flavonol in leaves of the wild type (Col-0) and the mutants. Error bars represent sd for five independent experiments.

(E) The ratio of flavonol derivatives to total flavonol in flowers of the wild type (Col-0) and the mutants. Error bars represent sd for five independent experiments. N.D., not determined.

To verify the physiological relationship between RHM1 and the flavonoid biosynthetic genes, the flavonol profiles of the knockout mutants of RHM1 (rol1-1 and rol1-2) and RHM2 (rhm2-1 and rhm2-3) were analyzed (Figure 8). Flavonol profiles in leaves and flowers of the rhm2 mutants showed no significant alteration. However, in leaves of rol-1 and rol1-2, accumulation levels of kaempferol 3-O-rhamnoside-7-O-rhmanoside (f1) were considerably reduced and those of kaempferol 3-O-glucoside-7-O-rhmanoside (f2) were elevated. In flowers of the rol1 mutants, levels of kaempferol/quercetin 3-O-rhamnoside-7-O-rhamnosides (f1 and f5) and kaempferol 3-O-rhamnosyl(1→2)-glucoside-7-O-rhamnoside (f3) were reduced and those of flavonol 3-O-glucosides (f16, f18, and f28) were elevated considerably. These results imply that UDP-rhamnose produced by RHM1 is used for flavonol rhamnosylation and RHM1 plays a major role in supplying UDP-rhamnose for modification of flavonols.

DISCUSSION

The Power of Detailed Targeted Flavonol Analysis and Transcriptome Coexpression Analysis for Functional Genomics

Here, we used a novel strategy for functional genomics in Arabidopsis using flavonol as a case study (Figure 9). Coupling the detailed analysis of secondary metabolites with coexpression can be used as a more general method for identifying the biosynthetic pathways of plant compounds associated with specific organs and developmental stages as well as their genetic components. The use of T-DNA mutations in genes encoding flavonoid biosynthetic enzymes to provide comprehensive annotation of the flavonols in Arabidopsis is effective and can provide an efficient method for elucidating flavonol structures. The integration of metabolic profiling and transcriptome coexpression could efficiently narrow the pool of candidate genes. Using these methods, a flavonol 3-O-arabinosyltransferase gene was found, and the relationship between flavonols and UDP-rhamnose synthesis has emerged. After listing the candidates potentially involved in the reactions or regulation of a metabolic pathway, the specific identification of a gene function could be accomplished using a traditional reverse genetics approach with knockout mutants and/or by biochemical characterization with the recombinant proteins, as has been exemplified in this study. The above strategy will be much improved as large metabolome data sets follow transcriptome data sets into to the public domain.

Figure 9.

Proposed Comprehensive Flavonol Pathway in Arabidopsis.

Each number in blue corresponds to the compounds shown in Figure 3. The compounds detected in Arabidopsis are shown in a blue background. Dotted lines indicate proposed but unidentified pathways. F3AraT, flavonol 3-O-arabinosyltransferase; Fd3GlcT, flavonoid 3-O-glucosyltransferase; F3RhaT, flavonol 3-O-rhamnosyltransferase; F7GlcT, flavonol 7-O-glucosyltransferase; F7RhaT, flavonol 7-O-rhamnosyltransferase; Ara, arabinose; DeoxyHex, deoxyhexose; Glc, glucose; Hex, hexose; Rha, rhamnose.

This combined strategy can also be useful for the detailed characterization of relatively minor flavonoid pathways that are evident only in mutants. The detailed flavonoid analyses of ugt78d2 mutants suggest the existence of another flavonoid 3-O-glucosyltransferase because a small amount of flavonoid 3-O-glucoside is present in the knockout mutant of UGT78D2. The ugt89c1 mutants accumulate flavonol 3-O-glucosides but not 3-O-rhamnosyl(1→2)glucosides, suggesting that 7-O-rhamnosylation likely occurs prior to 2′′-O-rhamnosylation. Accumulation of f19, f25, and f29 may also suggest that 7-O-rhamnosylation of kaempferol 3-O-arabinoside likely occurs prior to further glycosylation with deoxyhexose. The presence of f21, f26, and f31 suggests that a novel flavonol 3-O-glucoside:hexosyltransferase using glucose, galactose, or mannose may be involved in flavonol modification because those sugars have been so far identified as a sugar moiety of flavonol glycosides and have the molecular weight corresponding to unidentified hexose of f21, f26, and f30 (Andersen and Markham, 2006).

Transcriptome Coexpression Analysis Delimits Genes Related to Flavonoid Metabolism

Transcriptome coexpression analyses suggest other candidate genes that might be related to flavonoid metabolism. Eight of 24 candidate genes were regulated in PAP1-overexpressing or in MYB11/MYB12/MYB111 triple knockout plants. CHI_put (At5g05270) may play a supplementary role to the primary CHI (At3g55120). The coordinated expression of CHI_put with CHI tends to support this hypothesis. UGT79B1 belongs to a cluster containing the anthocyanin biosynthetic group. Phylogenetic trees of flavonoid UGTs suggest that UGT79B1 encodes anthocyanin 3-O-glucoside 2′′-xylosyltransferase (Tohge et al., 2005). These data suggest that CHI_put and UGT79B1 are involved in the flavonoid pathway. UGT84A1 (At4g15480) has 1-O-glucosyltransferase activity for phenylpropanoids such as p-coumaric acid and sinapic acid (Lim et al., 2001). Sinapoylmalate converted from 1-O-sinapoylglucose by malate sinapoyltransferase (SNG1), is a UV-B protective compound much as the flavonoids are (Ruegger and Chapple, 2001). Under UV stress conditions, accumulation of sinapoylmalate was higher in tt4 mutants than in the wild type (Kusano et al., 2007), suggesting that sinapoylmalate biosynthesis acts as a backup system for flavonoid metabolism under stress conditions. The positive correlation between MYB12 and UGT84A1 (r = 0.781) and the negative correlation of MYB12/UGT84A1 and SNG1 (r = −0.357/−0.323) tends to support this hypothesis. 1-O-Sinapoylglucose is also a substrate for anthocyanin sinapoyltransferase (Fraser et al., 2007), and anthocyanins accumulate with UV exposure and high sucrose stress. Arabidopsis may thus coregulate 1-O-sinapoylglucose with the flavonoids for prompt responses to stress. Cinnamoyl-CoA reductase–related protein (At2g23910) also may be involved in phenylpropanoid biosynthesis. Although coexpression does not necessarily indicate a functional relationship (Stuart et al., 2003), we believe that the significance of transcriptome coexpression analysis for functional gene identification has been established, at least for secondary metabolism.

A Flavonol Glycosyltransferase:Flavonol 3-O-Arabinosyltransferase

We demonstrated that UGT78D3 encodes a flavonol 3-O-arabinosyltransferase. The UGT78D3 function in planta was established by the identification of f19 and f25 as kaempferol/quercetin 3-O-arabinoside-7-O-rhamnoside, respectively. The conformation of substrates (kaempferol/quercetin 3-O-α-l-arabinosides and UDP-β-l-rhamnose) and previous data are consistent with the hypothesis that f19 and f25 are kaempferol/quercetin 3-O-α-l-arabinoside-7-O-α-l-rhamnosides, respectively (Mulinacci et al., 1995; Sakar et al., 2005). Compared with UGT78D1 and UGT78D2, UGT78D3 is weakly correlated with genes in the flavonoid metabolic pathway, which is consistent with the anatomical distribution of flavonol glycosides. The flavonol 3-O-arabinosides make up only a minor fraction of flavonols or are under the detection limit in organs except for flowers, whereas the flavonol 3-O-glucosides and 3-O-rhamnosides are distributed in all tested organs as major flavonols. Among the 6850 flavonoids registered in the Flavonoid Viewer database (http://www.metabolome.jp/software/FlavonoidViewer/viewer), 2.9% are arabinosylated flavonoids (M. Arita, personal communication). To find genes involved in such minor metabolite biosynthesis, an in-depth analysis of gene-to-metabolite correlation is essential.

Functional identification of three Arabidopsis flavonol 3-O-glycosyltransferases (3GlyTs) in the UGT78D subfamily with distinct UDP-sugar specificity allowed us to deduce the amino acid residues involved in substrate recognition. The three-dimensional structure of grape (Vitis vinifera) flavonoid 3-O-glucosyltransferase (3GlcT) was determined, and a structural analysis suggests the presence of several key residues that interact with UDP-sugar and the flavonoid backbone (Offen et al., 2006). The amino acids Gln-375, Asp-374, and Thr-141 have been proposed to interact with hydroxyl groups at the C-2 and C-3, C-3, and C-4, and C-6 positions of the glucose moiety of UDP-glucose, respectively, but Gln-375 and Thr-141 are not conserved in the Arabidopsis 3GlyTs (see Supplemental Figure 3 online). His residues corresponding to Gln-375 are conserved among the galactosyltransferases (Kubo et al., 2004). A His residue was also found as the corresponding residue in UGT78D3. The configurations of hydroxyl groups at C-2, C-3, and C-4 positions are the same in arabinose and galactose. However, UGT78D3 could not use UDP-galactose. Asp-374 in grape flavonoid 3GlcT is conserved in the three Arabidopsis 3GlyTs, although the configurations of hydroxyl groups at the C-3 and C-4 positions are different (see Supplemental Figure 3 online). Given this disparity, the residues involved in sugar donor specificity cannot be ascribed to a single amino acid residue.

RHM1 Is Involved in Flavonol Biosynthesis

Rhamnosylation is a major glycosylation of flavonols in Arabidopsis. Twenty-one of 32 detected flavonol derivatives in wild-type Arabidopsis are rhamnosides (Figure 3). A comparison of flavonol profile indicates that RHM1 plays a major role in supplying UDP-rhamnose for flavonol biosynthesis. RHM1 transcripts are accumulated as the major RHM species in most of the tested Arabidopsis organs. The expression pattern of RHM1 was well coordinated with the known flavonoid biosynthetic genes. The RHM1 correlation coefficients for flavonol rhamnosyltransferase (UGT89C1, r = 0.67; UGT78D1, r = 0.51) are higher than those for flavonol glucosyltransferase (UGT78D2, r = 0.45; UGT78D3, r = 0.32). In leaves, glycosylation at the C-3 position was strongly affected by the restriction of UDP-rhamnose supply because subsequent glycosylations at the C-7 or C-2′′ positions for major flavonols are rhamnosylation only. In flowers of rol1s, additional flavonol 3-O-glucosides were detected, suggesting that more UDP-rhamnose may be necessary in flowers and glycosylation at the C-7 position is also affected in the severe limitation of UDP-rhamnose. In fact, major six flavonol contents in flowers are ∼10- to 70-fold higher than in leaves (Yonekura-Sakakibara et al., 2007). The role of RHM1 for UDP-rhamnose supply to flavonols was also supported by a recent report with young shoots of the rol1 mutants (Ringli et al., 2008) and our data with RHM1 knockdown mutants (SALK_027926 and SALK_143589; see Supplemental Figure 4 online).

The extent of the RMH3 contributions is unclear, and it may be committed to other pathways. RHM2/MUM4 is involved in the synthesis of pectinaceous rhamnogalacturonan I in seed mucilage (Usadel et al., 2004; Western et al., 2004). Interestingly, RHM3 has a low correlation coefficient with RHM1 (r = 0.46, all data sets version 3). The molecular evolution of RHMs may explain this correlation. The Arabidopsis genome contains a number of large duplicated chromosomal segments. Blanc et al. (2003) proposed that the Arabidopsis lineage underwent at least two distinct episodes (old and recent) of duplication. According to Blanc et al. on their website (http://wolfe.gen.tcd.ie/athal/dup) for exploring paralogous blocks identified within the Arabidopsis genome, RHM1 is the ancestral gene form. RHM3 was duplicated and diverged from RHM1 during the old duplication, and RHM2/MUM4 emerged from RHM3 by some recent duplication (see Supplemental Figure 5 online). Comprehensive analysis of rhamnose-containing metabolites using RHM knockout mutants will be helpful to clarify the additional functions of RHM1, RHM2/MUM4, and RHM3.

METHODS

Plant Materials

Arabidopsis thaliana (accession Columbia-0; Lehle Seeds) was used as the wild type in this study. The mutants ugt75c1, ugt78d1, ugt78d2, ugt89c1, rol1-1, rol1-2, and rhm2-1 were described previously (Jones et al., 2003; Usadel et al., 2004; Tohge et al., 2005; Diet et al., 2006; Yonekura-Sakakibara et al., 2007). The T-DNA–inserted knockout mutants of omt1 (CS25167) and tt7 (CS6509) were obtained from the ABRC. The tt4 (C1) mutant was obtained from tt mutant lines induced by ion beam irradiation of Arabidopsis (Shikazono et al., 2003). The T-DNA–inserted mutant of Arabidopsis, lines SALK_114099 for UGT78D3 and SALK_085051 (rhm2-3, a homozygous knockout line) for RHM2, were obtained from the Salk Institute. T-DNA insertion lines were screened by PCR using specific primers for UGT78D3, RHM2 ,and T-DNA: UGT78D3f, UGT78D3r, RHM2f, RHM2r, LBa1, and RBa1 (see Supplemental Table 1 online). PCR products were sequenced to determine the exact insertion points.

Plants were cultured on Murashige and Skoog–agar medium containing 1% sucrose (Valvekens et al., 1988) in a growth chamber at 22°C with 16 h/ 8 h light and dark cycles for 18 d or in a greenhouse at 22°C with 16 h/ 8 h light/dark for 4 weeks. Plants were harvested, immediately frozen with liquid nitrogen, and stored at −30°C until use. Five biological replicates were used for flavonoid profiling.

Chemicals

Flavonoid standards, including quercetin 3-O-arabinoside, were purchased from Extrasynthese and Analyticon. Kaempferol 3-O-α-l-rhamnoside-7-O-α-l-rhamnoside, kaempferol 3-O-α-l-rhamnosyl(1→2)-β-d-glucoside-7-O-α-l-rhamnoside, quercetin 3-O-α-l-rhamnoside-7-O-α-l-rhamnoside, and quercetin 3-O-α-l-rhamnosyl(1→2)-β-d-glucoside-7-O-α-l-rhamnoside were from our laboratory collection. UDP-β-l-arabinose and UDP-α-d-xylose were purchased from CarboSource Services (supported in part by National Science Foundation–Plant Cell Wall Biosynthesis Research Network Grant 0090281).

Flavonol Profiling by UPLC-PDA-ESI/Q-TOF/MS

Flavonol analyses were performed in quintuplicate. Frozen leaves were homogenized in 5 μL of extraction solvent (methanol: CH₃COOH:H₂O = 9:1:10, 0.02 mM naringenin-7-O-glucoside) per milligram of fresh weight of tissue in a mixer mill (MM300; Retsch) for 5 min at 30 Hz. After centrifugation at 12,000g, the supernatants were immediately used for flavonoid analysis.

For flavonol profiling, a Waters Acquity UPLC system (Waters) fitted with a Q-TOF Premier mass spectrometer (Micromass MS Technologies) was used. UPLC was performed on a UPLC phenyl C18 column (Φ2.1 mm × 100 mm; Waters) at a flow rate of 0.5 mL/min at 35°C. Compounds were separated with a linear elution gradient with solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in acetonitrile) from 0 min, 100% A to 10 min, 40% B. PDA was used for the detection of UV-visible absorption in the range of 210 to 500 nm. The TOF mass analyzer was used for the detection of flavonol glycosides [M + H]⁺ and the peak of fragment ions in a positive ion scanning mode with the following setting: desolvation temperature, 400°C with a nitrogen gas flow of 600 L/h; capillary spray, 3.0 kV; source temperature, 150°C; and cone voltage of 10 V for flavonoid glycosides [M + H]⁺ or 30 V for fragment ions. For comprehensive flavonol profiling, nonprocessed MS data were converted to NetCDF format by MassLynx software (Micromass MS Technologies). Data analyses including principal component analysis were performed by the Phenomenome Profiler (Phenomenome Discoveries). Flavonol glycoside standards were used for the identification of the peaks in the plant extracts based on retention times, UV-visible absorption spectra, and mass fragmentation by tandem MS analysis. Other flavonol peaks were annotated by comparing their UV-visible absorption spectra, elution times, m/z values, and MS² fragmentation patterns with 85 reference flavonoid compounds and the reported data (Tohge et al., 2005, 2007; Routaboul et al., 2006; Yonekura-Sakakibara et al., 2007). The mass spectrum data of standard compounds (see Supplemental Data Set 2 online) were recorded in the MASSBANK Database (http://www.massbank.jp/index-e.html).

Transcriptome Coexpression Analysis

Coexpression analyses were performed using a Coexpression Gene Search algorithm on the RIKEN PRIMe website (http://prime.psc.riken.jp/; Akiyama et al., 2008) as described previously (Yonekura-Sakakibara et al., 2007; Saito et al., 2008). Twenty-four genes (listed in Table 2) were used as core query genes. The genes exhibiting positive correlations (r > 0.525) against the flavonoid biosynthetic genes were extracted using ATTED-II (Obayashi et al., 2007) based on all data sets version 3 in AtGenExpress. The coexpression graph was depicted by the Pajek program.

Evaluation of T-DNA Insertion Mutants

For complementation tests, full-length UGT78D3 cDNA was amplified by PCR using Arabidopsis flower cDNA as a template with primers UGT78D3-GWf and UGT78D3-GWr (see Supplemental Table 1 online). The Gateway system was used for construction of the binary vectors for Arabidopsis transformation. cDNA was cloned into the pCR8/GW/TOPO vector (Invitrogen) as an entry vector and sequenced to confirm the absence of PCR errors. pGWB2 was used as a destination vector. Transformation into Agrobacterium tumefaciens and Arabidopsis and the selection of transformants were performed as described previously (Yonekura-Sakakibara et al., 2007).

Arabidopsis plants were grown at 22°C under 16 h/8 h light and dark cycles in a plant growth room for ∼5 weeks. The leaves and flowers of plants were harvested, immediately frozen with liquid nitrogen, and stored at −30°C until use for metabolite profiles.

Quantitative Real-Time PCR

RNA Extraction and cDNA synthesis were performed as described previously (Yonekura-Sakakibara et al., 2004). The developmental stage of each organ used for analysis was as described previously (Yonekura-Sakakibara et al., 2007). Accumulation levels of UGT78D3, RHM1, RHM2, and RHM3 transcripts were analyzed by real-time PCR with an ABI PRISM 7500 real-time PCR system (Applied Biosystems), monitoring amplification with SYBR-Green I dye (Applied Biosystems). The primers UGT78D3-RTf, UGT78D3-RTr, RHM1-RTf, RHM1-RTr, RHM2-RTf, RHM2-RTr, RHM3-RTf, and RHM3-RTr (see Supplemental Table 1 online) were designed using Primer Express software (Applied Biosystems) and checked for specific product formation by a dissociation program. In each case, plasmid DNA containing the corresponding gene was used as a template to generate a calibration curve. Real-time PCR was performed in triplicate on a single biological sample for each genotype.

Production of Recombinant UGT78D3 Protein and Assay of Glycosyltransferase

Full-length UGT78D3 cDNA was amplified and cloned into pCR8/GW/TOPO (Invitrogen) as described above. To construct the protein expression vector, the resultant plasmid was amplified by PCR with the primers UGT78D3-pET41bf and UGT78D3-pET41br (see Supplemental Table 1 online), the PCR product was digested with StuI and XhoI, and ligated with StuI-XhoI-digested pET-41b(+) vector (Novagen). The sequence of the resultant plasmid, pKYS342, was confirmed. Escherichia coli strain BL21star(DE3) was used as a host for expression. Transformed cells were cultivated at 37°C until A₆₀₀ reached 0.5. After the addition of isopropyl-β-d-thiogalactopyranoside to a final concentration of 1 mM, cells were cultured at 25°C for 4 h. The cells were collected, and the protein was purified as a GST fusion according to the manufacturer's instructions (Qiagen).

The standard enzyme assay reaction mixture (final volume of 50 μL) consisted of 50 mM HEPES-KOH, pH 7.5, 150 μM flavonoid substrate, and 500 μM UDP-sugar. The mixture was preincubated at 30°C for 2 min, and the reaction was started by the addition of enzyme. Reactions were stopped after 0, 4, 8, 12, or 30 min of incubation at 30°C by the addition of 50 μL ice-cold 0.5% (v/v) trifluoroacetic acid/methanol for flavonols or 50 μL ice-cold 1.0% (v/v) HCl/methanol for anthocyanidins and anthocyanins, and the supernatant was recovered by centrifugation at 12,000g for 3 min. Flavonoids in the resultant solution were analyzed as described above.

Preparation and NMR Analysis of Kaempferol 3-O-Arabinoside-7-O-Rhamnoside and Quercetin 3-O-Arabinoside-7-O-Rhamnoside

A reaction mixture of UGT89C1 with kaempferol 3-O-α-l-arabinoside and UDP-β-l-rhamnose as substrates (total 300 mL) was concentrated and chromatographed on COSMOSIL 75C₁₈-OPN (Nacalai Tesque) and stepwise eluted with water, 10% aqueous CH₃CN, 30% aqueous CH₃CN, 50% aqueous CH₃CN, 70% aqueous CH₃CN, and 100% methanol. Kaempferol glycosides in the fractions of 30% aqueous CH₃CN, 50% aqueous CH₃CN, 70% aqueous CH₃CN, and 100% methanol were further separated by preparative HPLC using a C18 reversed-phase column (10 mm ×150 mm I.D., particle size 5 μm, Inertsil ODS-EP; GL Science) using 30% aqueous CH₃CN with flow rates of 2.0 mL/min at 30°C. Kaempferol 3-O-l-arabinoside-7-O-l-rhamnoside (1.8 mg) was obtained and the structure was supported by ¹H-NMR (600 MHz, methanol-d₄): δ 5.16 (1H, d, J = 6.6 Hz, arabinose H-1), δ 5.52 (1H, br. s, rhamnose H-1), δ 6.43 (1H, d, J = 1.8 Hz, H-6), δ 6.73 (1H, d, J = 1.2 Hz, H-8), δ 6.86 (2H, d, J = 9.0 Hz, H-3′and H-5′), δ 8.06 (2H, d, J = 8.4 Hz, H-2′ and H-6′), ROE correlation between the anomeric proton of rhamnose (rhamnose H-1, δ 5.52), and the protons at C-6 (H-6, δ 6.43) and C-8 position (H-8, δ 6.73), ESI-TOF-MS (positive ion mode): m/z 565.1554 [M+H]⁺, calculated for C₂₆H₂₉O₁₄, 565.1557, ESI-TOF-MS/MS (positive ion mode, collision energy 30 eV): m/z (relative intensity) 433.1145 [M+H-arabinose]⁺ (39), 287.0605 [M+H-arabinose-rhamnose]⁺ (100), and UV λ_max nm: 345 (on flow).

A large-scale enzymatic product of UGT89C1 with quercetin 3-O-α-l-arabinoside and UDP-β-l-rhamnose was isolated by a similar procedure except for the mobile phase (20% aqueous CH₃CN) of preparative HPLC step. Quercetin 3-O-l-arabinoside-7-O-l-rhamnoside (2.1 mg) was obtained and the structure was supported by ¹H-NMR (600 MHz, methanol-d₄): δ 5.22 (1H, d, J = 6.0 Hz, arabinose H-1), δ 5.56 (1H, br. s, rhamnose H-1), δ 6.46 (1H, br. s, H-6), δ 6.75 (1H, br. s, H-8), δ 6.88 (1H, d, J = 7.8 Hz, H-5′), δ 7.61 (1H, br. d, J = 8.4 Hz, H-6′), δ 7.76 (1H, br. s, H-2′). ROE (irradiation δ 5.56 [rhamnose H-1]): H-6 (δ 6.46) and H-8 (δ 6.75), ROE correlation between the anomeric proton of rhamnose (rhamnose H-1, δ 5.52) and the protons at C-6 (H-6, δ 6.43) and C-8 position (H-8, δ 6.73), ESI-TOF-MS (positive ion mode): m/z 581.1496 [M+H]⁺, calculated for C₂₆H₂₉O₁₅, 581.1506. ESI-TOF-MS/MS (positive ion mode, collision energy 30 eV): m/z (relative intensity) 449.1086 [M+H-arabinose]⁺ (41), 303.0554 [M+H-arabinose-rhamnose]⁺ (100), and UV λ_max nm:340 (on flow). Monitoring was accomplished by PDA. ¹H-NMR spectra were recorded on a JNM ECP-600 spectrometer (JEOL).

Accession Numbers

Sequence data from this article can be found in the Arabidopsis Genome Initiative database under accession numbers TT4 (At5g13930), TT7 (At5g07990), UGT75C1 (At4g14090), UGT78D1 (At1g30530), UGT78D2 (At5g17050), UGT78D3 (At5g17030), UGT89C1 (At1g06000), OMT1 (At5g54160), RHM1 (At1g78570), RHM2 (At1g53500), and RHM3 (At3g14790).

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Figure 1. Venn Diagrams of MS Peaks Detected in Various Organs of Col-0 and tt4.

• Supplemental Figure 2. Coexpression Relationships of Genes Involved in the Flavonoid Pathway in Arabidopsis.

• Supplemental Figure 3. Multiple Alignment of the Deduced Amino Acid Sequences of Flavonoid 3-O-Glucosyltransferases from Arabidopsis and Grape (Vitis vinifera).

• Supplemental Figure 4. T-DNA Insertion Mutants of RHM1 and UPLC/MS Analyses of the rhm1 Mutant Lines.

• Supplemental Figure 5. Schematic Representation of Duplicated Regions Containing RHMs in the Arabidopsis Genome.

• Supplemental Table 1. Primers Used in This Study.

• Supplemental Data Set 1. MS-Detectable Peaks in the Wild Type and tt4.

• Supplemental Data Set 2. Characteristics of Standard Compounds.