The phenylpropanoid pathway in plants leads to the synthesis of a wide range of soluble secondary metabolites, many of which accumulate as glycosides. In Arabidopsis, a small cluster of three closely related genes, UGT72E1–E3, encode glycosyltransferases shown to glucosylate several phenylpropanoids in vitro, including monolignols, hydroxycinnamic acids and hydroxycinnamic aldehydes. The role of these genes in planta has now been investigated through genetically downregulating the expression of individual genes or silencing the entire cluster. Analysis of these transgenic Arabidopsis plants showed that the levels of coniferyl and sinapyl alcohol 4-O-glucosides that accumulate in light-grown roots were significantly reduced. A 50% reduction in both glucosides was observed in plants in which UGT72E2 was downregulated, whereas silencing the three genes led to a 90% reduction, suggesting some redundancy of function within the cluster. The gene encoding UGT72E2 was constitutively overexpressed in transgenic Arabidopsis to determine whether increased glucosylation of monolignols could influence flux through the soluble phenylpropanoid pathway. Elevated expression of UGT72E2 led to increased accumulation of monolignol glucosides in root tissues and also the appearance of these glucosides in leaves. In particular, coniferyl alcohol 4-O-glucoside accumulated to massive amounts (10 μmol g−1 FW) in root tissues of these plants. Increased glucosylation of other phenylpropanoids also occurred in plants overexpressing this glycosyltransferase. Significantly changing the pattern of glycosides in the leaves also led to a pronounced change in accumulation of the hydroxycinnamic ester sinapoyl malate. The data demonstrate the plasticity of phenylpropanoid metabolism and the important role that glucosylation of secondary metabolites can play in cellular homeostasis.
The first committed enzyme in the phenylpropanoid pathway is phenylalanine ammonia-lyase, which catalyses the deamination of phenylalanine into trans-cinnamic acid. The classical view has been that trans-cinnamic acid is then subjected to a grid-like set of biotransformations involving ring modifications (hydroxylation and O-methylation) followed by side-chain modifications, leading to hydroxycinnamic-CoA esters, aldehydes and, ultimately, alcohols. Thus the three monolignols of the common phenylpropanoid pathway were thought to arise from their respective aldehydes which, in turn, were formed from the corresponding acids (Humphreys and Chapple, 2002).
In this context, the multigene family of enzymes that glycosylates plant secondary metabolites offers a useful tool to disturb metabolic homeostasis in planta (Bowles et al., 2005, 2006; Lim and Bowles, 2004). These glycosyltransferases (GTs) of lipophilic small molecules are found in family 1 of the GT classification system (http://afmb.cnrs-mrs.fr/CAZY/geno/3702.html). In Arabidopsis, 107 family members can be identified by the presence of a motif of 44 amino acids involved in nucleotide sugar binding (Li et al., 2001; Paquette et al., 2003; Ross et al., 2001). Arabidopsis genes encoding enzymes that recognize metabolites of the phenylpropanoid pathway have been identified through screening the catalytic activity of their recombinant proteins in vitro (Lim et al., 2001, 2003, 2005). Of particular significance is the UGT72E clade consisting of three proteins, UGT72E1, UGT72E2, UGT72E3 (gene locus numbers At3g50740, At5g66690, At5g26310 respectively), which recognized a variety of common phenylpropanoids in vitro, including acids, aldehydes and alcohols (Lim et al., 2001, 2003, 2005). The structures of the 4-O glucosides that have been reported are highlighted in Figure 1.
This study focuses on the UGT72E clade of Arabidopsis GTs, and explores the metabolic consequences of downregulating individual GT genes and the entire gene cluster. The data demonstrate that the monolignol glucosides formed in light-grown roots of Arabidopsis (Hemm et al., 2004) are products of these GTs. This study also provides evidence for the plasticity of the common phenylpropanoid pathway, as increasing the glucosylation of specific metabolites by overexpression of one gene of the cluster has consequences on levels of metabolically distant secondary products.
Pattern of expression of the UGT72E cluster in Arabidopsis
The accumulation of mRNA of UGT72E1, UGT72E2 and UGT72E3 was analysed at different developmental stages using real-time quantitative PCR (RT-qPCR); the results are presented in Figure 2, where levels of transcripts can be compared at different developmental stages for each GT. The data in Figure 2(a–c) show the relative expression of each gene in comparison with the reference gene transcript β-actin; these data are presented as expression ratio relative to the expression levels found in 14-day-old root tissue. The expression of the three genes was found to differ. Whereas transcripts of UGT72E1 were found in seedlings and leaves, those of UGT72E2 were limited to 2- and 4-day-old seedlings and 14-day-old root tissues, and those of UGT72E3 were present in the same tissues as UGT72E2 and in flowers and siliques. Figure 2(d) compares the expression of each GT with the expression of β-actin in 14-day root tissue, illustrating the relative abundance of transcripts in this tissue. It is clear that UGT72E1 has the highest expression of the three GTs (70% of the β-actin expression in 14-day-old roots) relative to UGT72E2 and UGT72E3 (2.7 and 0.7%, respectively, in the same tissue).
Downregulation of the UGT72E cluster in Arabidopsis
An Arabidopsis T-DNA insertion line was identified for UGT72E1 from the Salk collection (NASC N578702.56.00.x) from which a homozygous knockout line (72E1KO) was selected by PCR. No insertional homozygous line could be identified for either UGT72E2 or UGT72E3, therefore RNA interference was used as the strategy for knockdown of their expression (lines 72E2KD and 72E3KD). As the coding regions of UGT72E2 and UGT72E3 are highly homologous (>98.5%), the silencing was targeted to the less conserved 3′ untranslated regions. In addition, the entire UGT72E cluster was also silenced by targeting three different regions in their open reading frames (lines 72E123KD).
The presence of small interfering RNAs was confirmed in several independent homozygous knockdown (KD) lines (data not shown). The accumulation of transcripts in the KO and KD lines was analysed by quantitative RT-qPCR. The results are illustrated in Figure 3, in which transcript levels in 14-day-old root tissue are compared in wild-type and the transgenic lines. Figure 3(a) demonstrates that it was possible to massively downregulate individual GTs with little effect on the other two GTs. Thus accumulation of transcripts of the target genes was specifically reduced compared with wild type in 72E1KO (12% of WT levels), 72E2KD (15%) and 72E3KD (5%). The data shown in Figure 3(b) indicate that a triple knockdown was also achieved (expression of the three genes was reduced to <10% of wild-type levels in all lines). Enzyme activities in homogenates from wild-type rosette leaves towards known substrates of UGT72E1, E2 and E3 were found to be very low (data not shown), therefore this approach could not be used to investigate the potential existence of changed activities in the KO and KD lines.
The UGT72E clade is involved in monolignol glucosylation in Arabidopsis roots
The soluble phenolic chemotype of wild-type root tissue and those of the KO/KD lines were compared following methanol extraction and analysis using HPLC/photodiode array in the presence of an external standard. The results are shown in Figure 4. A representative HPLC profile of metabolites in wild-type roots under light conditions is illustrated in Figure 4(a) and confirms the data described by Hemm et al. (2004) in that significant quantities of coniferyl alcohol 4-O-glucoside (peak 1) and sinapyl alcohol 4-O-glucoside (peak 2) were present. When the UGT72E cluster was silenced in entirety, the levels of these glucosides were massively reduced. In contrast, the quantities of other phenolics, scopolin (peak 3), quercetin triglycoside (peak 4) and kaempferol triglycoside (peak 5), were unaffected.
Figure 4(b,c) compares the presence of coniferyl alcohol 4-O-glucoside and sinapyl alcohol 4-O-glucoside in wild-type and KO/KD lines. The external standard, scopoletin (peak 6), was used to standardize the data. The results clearly demonstrate that UGT72E2 was the member of the clade principally involved in the glucosylation of both coniferyl alcohol and sinapyl alcohol, as a twofold reduction in their glucosides was observed in the 72E2KD lines. In contrast, no significant changes were observed when expression of either UGT72E1 or UGT7E3 was downregulated. Interestingly, a more pronounced effect was observed in the 72E123KD lines (up to 10-fold reduction) than in the 72E2KD lines, suggesting possible redundancy of function. When levels of the two flavonol glycosides (Figure 4a, peaks 4 and 6) were determined, no changes were observed between the individual transgenic lines and wild type (data not shown).
Overexpression of UGT72E2 in Arabidopsis
As UGT72E2 was the principal enzyme involved in the glucosylation of monolignols in roots, it was of interest to determine the consequences of constitutively overexpressing the gene encoding for the same enzyme. Thus the CaMV35S promoter was used to drive the overexpression of UGT72E2 in transgenic Arabidopsis plants (72E2OE). Independent homozygous lines were subjected to Northern analyses to assess the steady-state levels of UGT72E2 transcripts. As shown in Figure 5(a), gene expression in the individual lines varied. To determine whether the elevation in expression corresponded to increased GT activity, leaf homogenates were assayed using substrates known to be recognized in vitro by recombinant UGT72E2 produced by Escherichia coli (coniferyl and sinapyl; acid, alcohol and aldehyde; the structures of the 4-O-glucosides formed are illustrated in Figure 1). Results from a representative experiment are shown in Figure 5(b). In homogenates from wild-type plants, only low levels of sinapyl alcohol 4-O-glucoside were detected by HPLC analysis of the reaction products. In contrast, when extracts from 72E2OE lines were assayed, substantial quantities of the glucosides were formed. Thus constitutive overexpression of UGT72E2 leads to increased enzyme activity in the leaves.
As described above, light-grown roots of wild-type Arabidopsis accumulated the glucosides of coniferyl alcohol and sinapyl alcohol, and the accumulation was prevented by downregulation of UGT72E2. Figure 6 illustrates the impact of overexpressing this GT on the monolignol glucoside profile of the roots. Of particular significance was the quantity of coniferyl alcohol 4-O-glucoside, where approximately 10 μmol g−1 FW accumulated compared with 1 μmol g−1 FW in the wild type. An increase in sinapyl alcohol 4-O-glucoside was also observed, but levels were much lower (up to 0.67 μmol g−1 FW compared with 0.27 in the wild type). Levels of the additional phenolics observed in wild-type light-grown roots (Figure 4a, peaks 3–5) were not significantly changed in the 72E2OE extracts (data not shown).
To determine whether phenylpropanoid accumulation in aerial tissues could also be affected by overexpressing UGT72E2, soluble phenolics were extracted from 4-week-old rosette leaves and subjected to HPLC analysis. Figure 7 illustrates the results, compares HPLC analysis of methanol extracts at 264 and 324 nm (Figure 7a,c), and quantifies the metabolites displaying the greatest differences in extracts of a wild-type and a 72E2OE line (line 5-2 in Figure 5a, Figure 7b,d).
As observed in roots, the peak areas of the monolignol glucosides in the profile determined at 264 nm increased substantially in the transgenic lines compared with wild type (Figure 7a, peaks 1 and 3). The quantitative data shown in Figure 7(b) indicate that the extract from 72E2OE had very high levels of coniferyl alcohol 4-O-glucoside. Interestingly, in the leaf extracts, sinapyl alcohol 4-O-glucoside levels were found to be significantly higher in the 72E2OE lines compared with wild-type levels, contrasting with the impact of overexpressing GT in roots. Additional changes in the profiles occurred in peaks 2, 4, 5 and 7, but to a lower extent: the identity of these peaks was not determined, although the profiles of peaks 2, 4 and 5 were similar to coniferyl and sinapyl alcohol. Peak 11 was identified as sinapoyl malate by LC–MS (data not shown).
Figure 7(b) shows the phenylpropanoid profiles of a wild-type and a 72E2OE at 324 nm, with Figure 7(d) providing quantitative data on the metabolites that had changed significantly. The data revealed that levels of sinapoyl malate (peak 11) were massively reduced in the 72E2OE line. In extracts of wild-type Arabidopsis leaves, ferulic acid 4-O-glucoside (peak 12) was not detectable, but was substantially increased in those of 72E2OE line. A small amount of sinapic acid 4-O-glucoside (peak 13) was observed but the level was too close to the detection limit for quantification.
Glucosides of coniferyl alcohol and sinapyl alcohol are known to accumulate in Arabidopsis roots in response to light, with levels 100- to 400-fold higher than in dark grown roots (Hemm et al., 2004). While constant exposure to light was required to maintain the high levels of glucosides, the means through which the metabolites were synthesized were not determined. Interestingly, light was not found to affect the relative transcript levels of two GTs that had previously been shown to glucosylate the monolignols in vitro (Hemm et al., 2004; Lim et al., 2001, 2005). The GTs were members of a small clade in Arabidopsis that had been identified through the presence of a consensus sequence typical of a subset of family 1 GTs (Li et al., 2001; Ross et al., 2001). This study now uses a genetic strategy to demonstrate that the clade of proteins UGT72E1, UGT72E2 and UGT72E3 is involved in the glucosylation of monolignols in light-grown roots of Arabidopsis.
Gene silencing has been used successfully to downregulate the expression of individual genes from the UGT72E cluster, as well to downregulate the entire cluster. The data indicate that UGT72E2 is the principal GT involved in glucosylation of monolignols in roots. There is some redundancy of function, as the cluster knockdown reduced the levels of the monolignol glucosides substantially from a twofold reduction in 72E2KD to up to 10-fold in 72E123KD. These transgenic KD lines will provide a useful tool to determine the consequences of preventing monolignol glucoside production in roots. For example, it has been suggested that the glucosides that accumulate in the light may constitute pools of metabolic precursors (Hemm et al., 2004), or may function directly through exhibiting a suggested antibacterial or antiviral activity (Pan et al., 2003). Recently, a detailed analysis of secondary metabolism in Arabidopsis root liquid cultures in response to Pythium silvaticum infection revealed a twofold decrease in coniferyl and sinapyl alcohol 4-O-glucosides (Bednarek et al., 2005). The authors suggested that the glucosides could represent preformed precursors converted during infection to defence-related compounds.
In recent years it has become increasingly apparent that flux through the common phenylpropanoid pathway may be much more plastic than previously thought. The discovery that the REF1 gene, cloned from A. thaliana, encoded an aldehyde dehydrogenase (Nair et al., 2004) led to the realization that the aldehydes could represent precursors of both monolignols and hydroxycinnamic acid esters. Similarly, the finding that an Arabidopsis mutant deficient in the expression of an O-methyltransferase led to changes in both lignin and sinapoyl malate (Goujon et al., 2003) highlighted the potential role of alcohols and aldehydes in the synthesis of both these products. Glucosylation of small molecules can lead to changes in intracellular compartmentation (Bowles et al., 2006) and thereby changes in the availability of metabolites to act as substrates for cytosolic reactions. In this context, the overexpression of a member of the UGT72E cluster has provided an opportunity to determine the consequences of increased glucosylation of monolignols on levels of other metabolites in the common phenylpropanoid pathway.
Thus, when UGT72E2 was overexpressed in root tissues, the only effects were increased levels of the monolignol glucosides, particularly coniferyl alcohol 4-O-glucoside. In contrast, the impact of overexpressing the GT in leaf tissues was much more wide-ranging. The most surprising feature of the results was the decreased level of sinapoyl malate (up to 10-fold) in leaves of the transgenic lines overexpressing UGT72E2. This loss in sinapoyl malate correlated with the massive accumulation of monolignol glucosides and increased levels of ferulic acid 4-O-glucoside. In order to gain an insight into the effects of overexpressing the GT on the ‘phenylpropanoid budget’ in the leaf tissue, total soluble phenolics were measured, but no overall changes were observed (data not shown), suggesting that the impact of increasing the level of glucosides was compensated by loss of sinapoyl malate. In turn, this may indicate that overexpression of UGT72E2 led to the redirection of flux within the phenylpropanoid pathway without affecting the overall phenylpropanoid budget.
As shown in Figure 8, sinapic acid is a precursor in sinapoyl malate synthesis (Chapple et al., 1992), therefore any reactions that disturb the homeostasis of sinapic acid could have an impact on levels of sinapoyl malate. In the 72E2OE lines, the increased production of glucosides and their likely accumulation in the vacuole may have led to the depletion of cytosolic metabolites that would otherwise contribute to sinapoyl malate production. Changes in subcellular compartmentation may therefore prove to be as influential on metabolic flux as changes in levels of biosynthetic enzymes.
Plant material and growth conditions
Arabidopsis thaliana. ecotype Columbia was used in this study. For aerial tissues, plants used were grown in Levington's seed and modular compost with a controlled environment of 16/8-h light/dark cycle (light, 100 μmol photons m−2 sec−1, 22°C; dark, 17°C). Plants were grown for 4 weeks, and leaves were harvested and stored at −80°C until further analysis.
For analysis of metabolite and gene expression in axenic roots of transgenic lines, seedlings were grown vertically on MS medium (Murashige and Skoog, 1962) for 14 days in constant light (180–200 mol m−2 sec−1, 21°C) as described by Hemm et al. (2004). After 2 weeks the roots were harvested and stored at −80°C until further analysis.
Transgenic plants and binary constructs
Arabidopsis plants were transformed by standard Agrobacterium tumefaciens methods (Clough and Bent, 1998) and primary transformants were selected with the appropriate antibiotic according to the binary vector used. The homozygous UGT72E1 knockout line (72E1KO) was selected from a T-DNA insertion line obtained at the Salk Institute (NASC N578702.56.00.x) using PCR with one insertion-specific primer (LBa1) and two gene-specific primers (LP and RP; for primer sequences see Table S1) as described by the Salk Institute Genomic Analysis Laboratory (http://signal.salk.edu/tdnaprimers.2.html). Both the 72E2KD and 72E3KD RNA-silencing vectors were obtained by cloning the 300- or 166-bp fragment, respectively, downstream of the stop codon, into the pHannibal vector (Wesley et al., 2001). These fragments were PCR-amplified with pair of primers RNAi72E2F/RNAi72E2R and RNAi72E3F/RNAi72E3R and cloned into BamHI/HindIII and XhoI/KpnI sites. The resulting inverted-repeat cassettes were transferred by cutting with NotI into the binary vector pART27 for plant transformation. The triple 72E123KD vector was constructed as follows: approximately 250-bp fragments of each ORF gene were PCR-amplified with primers E1F/E1R, E2F/E2R and E3F/E3R respectively. Then the three fragments were PCR-linked taking advantage of the overlapping sequences of E1R/E3F and E3R/E2F. The 72E123 linked fragment was cloned into pGEM-T easy (Promega, Southampton, UK), excised sequentially by XbaI/XmaI and XhoI/SwaI, and cloned into the same sites of pFGC5941 (http://www.chromdb.org/info/plasmids/pFGC5941.html) for plant transformation. The 72E2OE construct was obtained by cloning the full ORF between the CaMV35S promoter and NOS terminator into the binary vector pJR1Ri (primers 72E2ORF F/R; Table S1).
Total RNA was extracted from plant tissue using the TRI Reagent (T9424, Sigma, Gillingham, Dorset, UK) and standard procedures. The analysis of RNA accumulation for the developmental study and in KO/KD lines was carried out by RT-qPCR using Syber Green labelling in an ABI prism 7000 machine following standard manufacturers’ protocols. Extracted RNA was treated with RQ1 DNase (M6101, Promega) to remove any potential DNA contamination before reverse transcription was carried out using InvitrogenII (180064-014; Invitrogen, Paisley, UK). The primers used were qPCR actin, 72E1, E2 and E3, F or R (for forward and reverse; Table S1). The results were normalized to the reference gene β-actin and always expressed as expression ratios relative to a ‘control’ value using the ΔΔCt method as described in the Applied Biosystems guide (Applied Biosystems, Foster City, CA, USA). The control sample was 14-day root for the development study and wild type for the transgenic analysis. The standard deviations correspond to half the range of expression between 2−(ΔΔCt+s) and 2−(ΔΔCt−s) where s is the SD of ΔΔCt (n = 2). RNA accumulation in 72E2OE lines was assessed by Northern blotting using standard procedures; 10 μg RNA was used for each sample. 32P-dCTP was used for radiolabelling the probes corresponding to the ORFs.
Targeted metabolite profiling
Soluble phenylpropanoid extraction. To extract soluble phenylpropanoids accumulated in leaves from transgenic lines, 1 g frozen leaf tissue was incubated for 1 h on ice with 10 ml 80% methanol supplemented with an external standard (scopoletin 0.33 mm, Sigma). After filtration through glass wool, the supernatant was collected and concentrated to approximately 1 ml on a rotary evaporator (Buchi R114; Buchi, Flawil 1, Switzerland) before loading on an activated Strata-X solid-phase extraction column (8B-S100-TAK, 1 ml, Phenomenex, Macclesfield, Cheshire, UK). Elution was carried out with 1 ml methanol. The flowthrough, washes and eluate were gathered and constituted the methanol extract. An aliquot of 50 μl was then used for reverse-phase HPLC analysis. To extract soluble phenylpropanoid accumulated from axenic root, 100 mg root tissue were incubated in 1 ml 80% methanol supplemented with scopoletin (0.33 mm) for 90 min at room temperature, before loading on a Strata-X column; reverse-phase HPLC analysis was as described above.
Reverse-phase HPLC (SpectraSYSTEM HPLC systems and UV6000LP Photodiode Array Detector; ThermoQuest, ThermoFinnigan, San Jose, CA, USA) analysis was carried out using a Columbus 5-μm C18 column (250 × 4.60 mm; Phenomenex) maintained at 30°C. The data were acquired and analysed using the software ChromQuest ver. 2.51. A linear gradient of 10–30% acetonitrile in water (all solutions contained 0.1% trifluoroacetic acid) with a flow rate of 1 ml min−1 over 20 min was used. Each peak on the chromatogram was scanned between 200 and 400 nm (photodiode array profile) and integrated at 264, 280 and 324 nm.
HPLC–MS was carried out using an Agilent HPLC (Agilent, Wokingham, Berkshire, UK) attached directly to an Applied Biosystems Qstar Pulsar I mass spectrometer with a turbo ion spray source. An HPLC linear gradient of 10–35% acetonitrile in water (all solutions contained 0.1% formic acid) over 20 min was used. The mass spectrometer was operated in negative ion mode with an ion spray voltage of −2500 V and the nebulisor and turbo gases set at 60 and 70 units respectively. Parent ions were fragmented by collision-induced dissociation and the product ions analysed from 50–1000 atomic mass units. The fragmentation experiments used collision energy settings of −10, −30 and −60 U respectively.
The MS ion-fragmentation spectra obtained from fractionated HPLC peaks were compared with published MS spectra for metabolite identification (Hemm et al., 2004 for coniferyl alcohol 4-O-glucoside and sinapyl alcohol 4-O-glucoside; Goujon et al., 2003 for sinapoyl malate; Rohde et al., 2004 for scopolin). The specific identity of the flavonol glycosides has yet to be confirmed.
Glycosyltransferase assay on crude protein extract from homozygous transgenic plants. Four-week-old leaves were harvested and frozen in liquid nitrogen. Crude protein extractions and enzymatic assays were performed as described by Hou et al. (2004). Before enzymatic assay, soluble phenolic metabolites were removed from the protein homogenates by centrifuging and washing the extracts three times in a microcon filter YM-30 (Millipore, Watford, UK).
The authors would like to thank the UK Biotechnology and Biological Science Research Council (BBSRC) and the Garfield Weston Foundation for funding support. Members of D.B.’s group at the Centre for Novel Agricultural Products are also thanked for their support. David Ashford is thanked for LC–MS support.