The ArabidopsisREF8 gene encodes the 3-hydroxylase of phenylpropanoid metabolism


*For correspondence (fax 765-496-7213; e-mail


The activity of p-coumarate 3-hydroxylase (C3H) is thought to be essential for the biosynthesis of lignin and many other phenylpropanoid pathway products in plants; however, no conditions suitable for the unambiguous assay of the enzyme are known. As a result, all attempts to purify the protein and clone its corresponding gene have failed. By screening for plants that accumulate reduced levels of soluble fluorescent phenylpropanoid secondary metabolites, we have identified a number of Arabidopsis mutants that display a reduced epidermal fluorescence (ref) phenotype. Using radiotracer-feeding experiments, we have determined that the ref8 mutant is unable to synthesize caffeic acid, suggesting that the mutant is defective in a gene required for the activity or expression of C3H. We have isolated the REF8 gene using positional cloning methods, and have verified that it encodes C3H by expression of the wild-type gene in yeast. Although many previous reports in the literature have suggested that C3H is a phenolase, the isolation of the REF8 gene demonstrates that the enzyme is actually a cytochrome P450-dependent monooxygenase. Although the enzyme accepts p-coumarate as a substrate, it also exhibits significant activity towards other p-hydroxylated substrates. These data may explain the previous difficulties in identifying C3H activity in plant extracts and they indicate that the currently accepted version of the lignin biosynthetic pathway is likely to be incorrect.


Most of the genes encoding the enzymes of the phenylpropanoid pathway have been cloned over the last 15 years by standard biochemical approaches, followed by the isolation of an array of orthologues from various species. The two known cytochrome P450-dependent monooxygenases (P450s) in the pathway, cinnamate 4-hydroxylase (C4H) and ferulate 5-hydroxylase (F5H) were more difficult targets because the instability, low abundance, and membrane-bound nature of plant P450s makes conventional purification problematic (Chapple, 1998). Despite these difficulties, the gene encoding C4H was identified following purification of the enzyme (Gabriac et al., 1991; Mizutani et al., 1993a, 1993b; Teutsch et al., 1993). The gene encoding F5H was isolated using a T-DNA tagged allele of the Arabidopsis fah1 mutant (Meyer et al., 1996b).

The biosynthesis of many phenylpropanoids requires at least two hydroxylation steps. C4H introduces the first hydroxyl group at the 4-position of the aromatic ring of cinnamic acid. C4H activity is readily measured in plants, and was one of the first plant enzymes recognized to be a P450 (Russell and Conn, 1967; Russell, 1971). The next hydroxylation occurs at the 3-position of the ring, and the enzyme that catalyses this reaction, p-coumarate 3-hydroxylase (C3H), has not been unambiguously characterized. Over the past 30 years, many researchers have attempted to assay, characterize and purify C3H, and there is considerable disagreement in their published reports concerning the nature of the enzyme. Activity thought to represent C3H has been detected in extracts of spinach beet, sorghum, oak, mung bean, and potato (Alibert et al., 1972; Bartlett et al., 1972; Bolwell and Butt, 1983; Boniwell and Butt, 1986; Duke and Vaughn, 1982; Halliwell, 1975; Kojima and Takeuchi, 1989; Stafford and Dresler, 1972; Vaughan and Butt, 1969; Vaughan and Butt, 1970). C3H has been characterized by various researchers as either an ascorbate-, NADPH- or flavin-dependent, mixed function oxidase (Boniwell and Butt, 1986; Kojima and Takeuchi, 1989; Stafford and Dresler, 1972; Vaughan and Butt, 1970), a plastidic enzyme that uses plastoquinone or ferredoxin as an electron donor (Bartlett et al., 1972), or most commonly, a phenolase that also oxidizes dihydroxyphenols to their corresponding orthoquinones (Stafford and Dresler, 1972). Some research has also suggested that the 3-hydroxylation reaction is catalysed by a P450 that uses p-coumaroyl quinate or p-coumaroyl shikimate as its substrate (Heller and Kühnl, 1985; Kühnl et al., 1987), or by a Zn2+-dependent enzyme that converts p-coumaroyl CoA to caffeoyl CoA (Kneusel et al., 1989). Given the dearth of clear biochemical data on C3H, we concluded that a genetic approach may provide the best opportunity to isolate the C3H gene, thereby enabling the characterization of the encoded protein after heterologous expression.

Leaves of Arabidopsis fluoresce blue-green under UV light due to the presence of sinapoylmalate, a compound that plays a role in the resistance of Arabidopsis to UV-B irradiation (Landry et al., 1995). In contrast, the sinapate ester-deficient fah1 mutant appears red under UV light (Chapple et al., 1992). By screening for new mutants that exhibit similar phenotypes, we recently identified mutations at a number of new loci (Ruegger and Chapple, 2001). Eight of these mutants have been named reduced epidermal fluorescence (ref1-ref8), all of which segregate as single, recessive, nuclear genes. When viewed under UV light, ref mutants display reductions in the blue-green fluorescence of their leaves, an indication that these mutants fail to accumulate wild-type levels of sinapoylmalate. Under these conditions, plants homozygous for ref8 look similar to the sinapoylmalate-deficient fah1–2 mutant. In this paper we demonstrate that the ref8 mutant cannot synthesize sinapate esters due to a lesion in the gene encoding C3H.


The ref8 mutant appears red under UV light

To identify mutants defective in sinapoylmalate biosynthesis, approximately 100 000 M2 seedlings (Columbia ecotype) were screened for individuals that exhibited altered cotyledon and/or leaf fluorescence when exposed to UV light. This mutant screen identified representatives of two known mutations that are affected in sinapate ester biosynthesis, fah1 and sng1, as well as a number of additional mutant lines that define several other loci. Five of these have been phenotypically characterized (Ruegger and Chapple, 2001), whereas the others have not, largely due to issues associated with vigour and fertility of the M2 plants and their progeny. The ref8 mutant belonged to this latter group.

When observed under UV light, the ref8 mutant is strongly red fluorescent, consistent with the absence of sinapoylmalate, and the accentuation of chlorophyll fluorescence that accompanies the lack of this UV-absorbing secondary metabolite (Figure 1). When F2 seedlings from crosses of ref8 mutants to wild type were examined, the mutant phenotype segregated as a recessive, nuclear, single gene mutation (405 REF8/–, 138 ref8/ref8; χ2 = 0.04, P > 0.7).

Figure 1.

UV phenotype of the ref8 mutant.

Rosette leaves of 3-week-old plants were photographed under 365 nm UV light using a yellow barrier filter.

The blue-green colour of the wild-type rosettes is due to the fluorescence of sinapoylmalate.

The red fluorescence of the ref8 plant is due to chlorophyll fluorescence that is revealed in the absence of sinapate ester fluorescence.

The ref8 mutant is blocked in the synthesis of caffeic acid

In previous studies, we have found that radiotracer feeding experiments are useful in elucidating defects in phenylpropanoid metabolism in sinapate ester-deficient mutants (Chapple et al., 1992). Thus, to investigate phenylpropanoid metabolism in the ref8 mutant, we evaluated the fate of 14C-l-phenylalanine administered to wild-type and ref8 leaves via the transpiration stream. These experiments revealed that radiolabelled pools of cinnamic, p-coumaric, ferulic and sinapic acids can be detected in the extracts of wild-type plants (Figure 2a). In extracts of the ref8 mutant, radiolabel was detected in cinnamic acid and p-coumaric acid, but not in any subsequent metabolites. In neither wild type nor ref8 was radiolabel found to be associated with caffeic acid or 5-hydroxyferulic acid, possibly indicating that the pools of these intermediates were below the detectable limits of this method. Since hydroxycinnamic acids are often found in esterified forms in plants, and hydroxy-cinnamoyl CoA thioesters are thought to comprise an important group of molecules in plant metabolism, samples of the previous extracts were saponified and again separated by two-dimensional thin layer chromatography (TLC). These analyses provided qualitatively similar data: the ref8 extract hydrolysates contained no radiolabelled ferulic acid, but wild-type levels of labelled p-coumaric acid (Figure 2b).

Figure 2.

Analysis of phenylpropanoid synthesis in wild-type and ref8 leaves.

(a) Hydroxycinnamic acids were extracted from leaves following uptake of 14C-l-phenylalanine via the transpiration stream and were visualized on a Packard Instant Imager (C, cinnamic acid; pC, p-coumaric acid; F, ferulic acid; S, sinapic acid).

(b) Similar analysis as described in (a) performed after saponification of the methanolic extracts from the previous experiment to release ester-bound hydroxycinnamic acids.

The inability of the ref8 mutant to convert 14C-phenylalanine into ferulic acid is consistent with the hypothesis that the mutant is blocked in either the hydroxylation of p-coumaric acid, the O-methylation of caffeic acid to ferulic acid, or the analogous reactions occurring at the level of the corresponding CoA thioesters. To distinguish between these possibilities, we compared the levels of caffeic acid/5-hydroxyferulic acid O-methyl transferase (COMT) and caffeoyl CoA O-methyltransferase (CCoAOMT) in wild-type and ref8 stem extracts. These experiments revealed that the two O-methyltransferase activities were present at near wild-type levels in ref8 (Table 1). Although these data suggested that O-methylation of caffeic acid and caffeoyl-CoA are not affected in the ref8 mutant, we wanted to determine whether the supply of the cosubstrate of the O-methyltransferase reaction, S-adenosylmethionine (SAM), might be limiting phenylpropanoid metabolism in the ref8 mutant. As an indirect measure of SAM levels, we quantified the production of ethylene in wild-type and ref8 plants (Table 1). These measurements indicated that ref8 rosettes are competent to synthesize ethylene, and that a block in SAM biosynthesis is not likely to be the cause of the perturbation in phenylpropanoid metabolism in the mutant.

Table 1.  Phenylpropanoid O-methyltransferase activities and ethylene production in wild-type and ref8 Arabidopsis. Data represent the results of triplicate assays
(pkat mg protein−1± se)
(pkat mg protein−1± se)
Ethylene production
(p.p.b. mg FW−1 min−1± se)
Wild type152 ± 9.41453 ± 2911.3 ± 2.7
ref8203 ± 6.11150 ± 646.3 ± 1.3

To directly evaluate the ability of the mutant to hydroxylate p-coumaric acid and/or p-coumaroyl-CoA we performed a modified version of the previous radiotracer feeding experiments, again administering 14C-l-phenylalanine to excised wild-type and ref8 leaves. In this experiment, the leaf extracts were treated with methanolic HCl to convert hydroxycinnamic acid esters and thioesters as well as free hydroxycinnamic acids to their corresponding methyl esters, while simultaneously preventing the destruction of alkali-labile dihydroxy-substituted compounds such as caffeic acid. Following two successive rounds of preparative TLC, we analysed the semi-purified methylcaffeate by HPLC and liquid scintillation counting (Figure 3). Using UV detection, we readily identified the methylcaffeate derived from the internal standard of unlabeled caffeic acid that had been added to the wild-type and ref8 samples at the time of extraction. In contrast, whereas radioactivity co-chromatographing with methylcaffeate was readily detected in the fractions collected from the wild-type samples, no radioactivity was detected in these fractions when this procedure was repeated on the ref8 samples. These experiments, as well as the parallel phenotypic characterization of the ref8 mutant (Franke et al. 2002), provided extremely strong evidence that the REF8 gene encodes a protein required for the activity or expression of C3H.

Figure 3.

Radiotracer feeding experiment designed to measure incorporation of label from 14C-l-phenylalanine specifically into caffeic acid, analysed as its methylester.

The continuous line represents the UV-absorption of the HPLC column eluate. The histogram represents radioactivity associated with each fraction (Ca, caffeic acid; pC, p-coumaric acid; Me-Ca, methylcaffeic acid; Me-pC, methyl-p-coumaric acid).

The REF8 gene is CYP98A3

To isolate the REF8 gene using positional cloning, we took advantage of the advanced state of the Arabidopsis genome-sequencing effort. Using a mapping population of 535 F2 plants derived from a ref8/ref8 (Columbia background) × REF8/REF8 (Landsberg erecta) cross, the position of the REF8 gene was initially determined to be between markers nga168 and T8M12. Thirty-nine plants were found to carry chromosomes that were recombinant within this region, and these individuals were studied further to determine a smaller mapping interval for the REF8 gene. These studies showed REF8 to lie between markers T7D17 and SGCSNP169, a region defined by a contig of approximately 10 BACs. The annotations of these clones were inspected for genes encoding putative oxidases and hydroxylases at the Munich Information Center for Protein Sequences website ( Two P450s and one peroxidase were identified within this region that we considered to be candidates for REF8 (Figure 4). Based upon the relative position of these genes within the mapping interval, and the recombination distances determined for the flanking upper and lower markers (2 recombination events between marker T7D17 and REF8; 10 recombination events between REF8 and marker SGCSNP169), we identified a gene on BAC T20B5 encoding a putative P450 belonging to the CYP98 family (CYP98A3) as the most likely candidate for REF8. To provide a preliminary indication whether CYP98A3 was likely to correspond to REF8, we evaluated the tissue specificity of its expression using RNA gel blot hybridization (Figure 5). The expression pattern of CYP98A3 is virtually indistinguishable from that of C4H and F5H (Bell-Lelong et al., 1997; Ruegger et al., 1999), suggesting that C4H, F5H, and CYP98A3 may be involved in the same biochemical pathway and thus co-regulated.

Figure 4.

Map-based cloning of the REF8 gene.

A narrow mapping interval was determined for the REF8 gene containing three genes considered to be candidates for REF8.

The P450 gene (underlined) encoded on BAC T20B5 (T20B5.9, CYP98A3) was selected as the most likely candidate for REF8 based upon its position within the mapping interval relative to the frequency of recombinant chromosomes identified within the mapping population at the flanking markers.

Near the heme-binding motif of T20B5.9 (shown below shaded in grey), the ref8 gene contains a single G to A transition mutation that results in a G444D substitution in the amino acid sequence.

Figure 5.

RNA gel blot analysis of REF8 expression in Arabidopsis.

Ethidium bromide staining of the 28S rRNA band is shown as a loading control.

The experiments described above strongly suggested that REF8 corresponds to T20B5.9, a gene also annotated as CYP98A3, a putative P450. To provide further supporting evidence for this hypothesis, we sequenced the putative ref8 cDNA. The mutant sequence contained a single G to A nucleotide substitution that leads to a non-conservative amino acid substitution that is likely to be situated at the beginning of the l-helix near the conserved heme-binding region of the protein (Figure 4), and introduces an EcoRV polymorphism into the mutant allele. Next, using an EST corresponding to the putative REF8 cDNA, we probed a transformation-competent Landsberg erecta cosmid library (Meyer et al., 1996a) and identified three independent clones carrying the corresponding genomic DNA (Figure 6a). Since the ref8 mutant is female sterile, it cannot be transformed directly to test whether the putative REF8 gene complements the ref8 mutant phenotype. Instead, F3 seeds from a REF8/ref8 F2 plant were sown, and ref8 plants were removed. The remaining plants (REF8/REF8 and REF8/ref8 individuals in a 1 : 2 ratio) were transformed with the putative REF8 cosmids using the floral dip method (Clough and Bent, 1998). After selection of kanamycin-resistant progeny, the genotype of all transformants was determined by PCR using the EcoRV polymorphism in the ref8 gene, and a HindIII polymorphism in the Landsberg erecta derived transgene (Figure 6b). Four independent transformed homozygous ref8 plants and the subsequent progeny of four additional independent REF8/ref8 individuals were analysed further.

Figure 6.

Complementation of the ref8 mutant.

(a) The CYP98A3 gene is encoded on the complementary strand of BAC T20B5 as deposited in GenBank. The three cosmids pCC554, pCC555, and pCC556 were identified using the CYP98A3 cDNA as a probe. The inserts of these cosmids carry two to four HindIII fragments that span the CYP98A3 locus. The HindIII restriction sites in these cosmids are designated with vertical lines. One of these sites is polymorphic between the Columbia ecotype and the Landsberg erecta-derived cosmid inserts, and the ref8 allele contains an additional EcoRV polymorphism caused by the mutation that inactivates the gene product. These polymorphisms were used to design oligonucleotide primers to produce a PCR amplicon for genotyping transgenic ref8 plants.

(b) Agarose gel electrophoresis of the PCR amplicon shown in (a) from various genotypes after digestion with HindIII and EcoRV. Lanes 1 and 10, DNA ladder; lane 2, Columbia wild type; lane 3, ref8 homozygote; lane 4, REF8/ref8 heterozygote; lane 5, Landsberg erecta; lane 6, ref8 homozygote transformed with pCC554; lane7, ref8 homozygote transformed with pCC555; lane 8, ref8 homozygote transformed with pCC556; lane9, REF8/ref8 heterozygote transformed with pCC556. The Landsberg erecta transgene-derived products seen in lanes 7–9 are absent from lane 6 because pCC554 lacks the primer-binding site downstream of the CYP98A3 gene.

(c) Appearance of wild type, ref8, and a representative ref8 homozygote transformed with pCC556.

Inspection of transgenic ref8 plants under UV light indicated that they contained wild-type levels of sinapoylmalate, indicating that the transgenes complemented the ref8 sinapate ester phenotype (data not shown). When transformed REF8/ref8 transgenics were allowed to set seed, and their T2 progeny were analysed in the next generation, T2 plants segregated for the ref8 fluorescence phenotype when grown on soil, and segregated for kanamycin resistance when grown on selective medium. Transformed ref8 homozygotes from these populations were also sinapoylmalate proficient, and like the ref8 tranformants originally identified, showed none of the other phenotypes associated with the ref8 mutant, including dwarfing and reduced fertility (Figure 6c). These data indicate that the REF8 gene is found in a region common to these three cosmids. The only annotated genes within this common region encode a putative cysteine proteinase, a nodulin-like protein, and CYP98A3.

REF8 is a cytochrome P450-dependent monooxygenase

To evaluate whether the G to A nucleotide substitution found in the mutant gene impairs enzymatic function, we expressed the wild-type and mutant genes in E. coli and yeast. For expression of the protein in E. coli, we used the pBOV vector, a version of pCWOri + modified for the high-level expression of eukaryotic P450s (Barnes et al., 1991). In this vector, the coding sequence of a portion of the N-terminal domain of the eukaryotic P450 is replaced by the first eight codons of the bovine P450 CYP17α. For the yeast experiments, we expressed the native CYP98A3 protein using the vector YeDP60 (Urban et al., 1990). The CYP98A3 expression constructs were then transformed into WAT11 yeast in which the endogenous yeast P450 reductase gene has been replaced with the ATR1 Arabidopsis P450 reductase gene under the control of the yeast GAL10-CYC1 promoter to provide the expressed P450 with high levels of its normal reductase partner (Pompon et al., 1996).

First, membrane preparations were isolated for spectroscopic and SDS-PAGE analysis from E. coli carrying either the pBOV control vector, the REF8 expression vector pBOV-REF8, or the pBOV-ref8 construct carrying the mutant gene sequence. SDS-PAGE analysis indicated that membranes from bacteria carrying pBOV-REF8 and pBOV-ref8 contained an abundant protein with a molecular mass of approximately 58 kDa, consistent with the expected mass of 57 926 Da for the inferred translation product of the putative C3H cDNA (Figure 7a). This protein was absent in samples prepared from bacteria carrying the control pBOV vector. As expected, carbon monoxide difference spectroscopy indicated that samples prepared from control bacteria contained no spectrally active P450. Spectroscopic examination of membranes prepared from bacteria expressing the putative C3H protein revealed a 450-nm absorbance peak characteristic of P450s (Figure 7b). In contrast, the CO difference spectrum of membranes from bacteria transformed with the pBOV-ref8 construct gave a spectrum dominated by a peak at 420 nm, indicating that the protein is capable of binding the heme prosthetic group found in P450s, but that the heme coordination is disrupted in such a way that the protein is likely to be inactive.

Figure 7.

Heterologous expression of REF8.

(a) SDS-PAGE analysis of membrane proteins from E. coli harbouring pBOV (lane 1), pBOV-REF8 (lane 2), and pBOV-ref8 (lane 3).

(b) Carbon monoxide difference spectra of dithionite-reduced E. coli membranes. Dashed line, baseline prior to CO treatment; solid line, CO difference spectrum of REF8-containing membranes; dotted line, CO difference spectrum of membranes from E. coli expressing the protein encoded by the ref8 allele.

(c) SDS-PAGE analysis of the Triton X-114 detergent-rich phase of yeast microsomal proteins harbouring YeDP60 (lane 1), YeDP60-REF8 (lane 2), and YeDP60-ref8 (lane 3).

(d) Carbon monoxide difference spectrum of dithionite-reduced Triton X-114 detergent phase from microsomes prepared from yeast harbouring YeDP60-REF8. Dashed line, baseline prior to CO treatment; solid line, difference spectrum after CO treatment.

Next, microsomal preparations prepared from yeast transformed with YeDP60, YeDP60-REF8 or YeDP60-ref8 were similarly analysed by SDS-PAGE and CO difference spectroscopy. Prior to these analyses, membrane preparations were fractionated using a Triton X-114 phase partition procedure to enrich the sample in integral membrane proteins (Werck-Reichhart et al., 1991). When analysed by SDS-PAGE (Figure 7c), the Triton phase prepared from membranes of control yeast contained a number of bands, whereas similar preparations from yeast expressing the putative wild-type REF8 protein contained an additional protein with a molecular mass of approximately 58 kDa. In contrast, mutant protein does not accumulate in yeast carrying the YeDP60-ref8 vector (Figure 7c). As expected, carbon monoxide difference spectroscopy indicated that samples prepared from control cells contained essentially no spectrally active P450; whereas, under the same conditions membranes prepared from yeast expressing the wild-type version of the putative C3H protein exhibited a strong 450 nm absorbance peak (Figure 7d). Taken together, these data suggest that the putative REF8 protein is probably targeted to the endoplasmic reticulum in yeast cells, where it should be catalytically active. In contrast, the mutant protein does not accumulate, possibly due to enhanced degradation arising from misfolding like that previously seen when the mutant protein was expressed in E. coli.

Finally, WAT11 yeast carrying the control vector and the YeDP60-REF8 vector were assayed for C3H activity in vivo by adding p-coumarate directly to the medium of galactose-induced yeast cultures. This method has previously been used to demonstrate the activity of C4H and F5H heterologously expressed in yeast (Humphreys et al., 1999; Pierrel et al., 1994), and exploits the ability of simple hydroxycinnamic acids to readily cross yeast membranes. When p-coumarate was added to the medium of control yeast, it was the predominant UV-absorbent substance present in the medium after several hours of incubation. In contrast, when a parallel experiment was performed with yeast harbouring the YeDP60-REF8 expression vector, a novel peak was found whose retention time and UV-spectrum matched precisely those of caffeic acid (Figure 8a). Replicate experiments analysed by GC–MS permitted unequivocal identification of the C3H reaction product (Figure 8b). In the context of the phenotypic characterization of the ref8 mutant (Franke et al., 2002), and the complementation of the ref8 mutant phenotypes, these data provide definitive proof that CYP98A3 encodes C3H, and that C3H is a P450.

Figure 8.

In vivo analysis of C3H activity in yeast.

A novel peak which co-chromatographs with authentic caffeic acid in both HPLC (a) and GC (b) was found in the medium of galactose-induced cells carrying the vector YeDP60-REF8 after 24 h incubation of the cells in the presence of 5 mmp-coumaric acid. The UV and mass spectra of this novel peak (HPLC retention time 13.1 min, GC retention time 11.359 min) and the caffeic acid standard is shown at right. The GC–MS chromatograms presented are the extracted ion chromatograms for m/z 219, a fragment common to both p-coumaric acid and caffeic acid.

C3H exhibits activity against multiple substrates

Although the in vivo assays of yeast carrying the YeDP60-REF8 vector demonstrated that C3H is capable of hydroxylating p-coumarate, it was not possible to use this approach to determine kinetic constants for the enzyme, nor to use this system to assay the activity of C3H toward substrates that cannot readily cross the yeast plasma membrane. To experimentally address these issues, C3H-containing microsomes were prepared for use in in vitro assays of enzymatic activity. Consistent with the in vivo results, incubation of C3H in the presence of p-coumarate in vitro resulted in the production of caffeic acid (Figure 9a), although this activity was so low that it precluded detailed kinetic analysis. In addition to p-coumarate, several other compounds have been suggested to be substrates for the 3-hydroxylase(s) of phenylpropanoid metabolism (Heller and Kühnl, 1985; Kneusel et al., 1989; Kühnl et al., 1987; Tanaka and Kojima, 1991). Because the assays using p-coumarate suggested that it might not be the optimal substrate for C3H, we assayed the activity of the enzyme against an array of other possible substrates. In these experiments, no activity of C3H toward p-coumaroyl CoA, p-hydroxycinnamyl alcohol, and 1-O-p-coumaroyl-β-d-glucose was detected. Levels of activity comparable to those seen with p-coumarate were seen when p-coumaraldehyde was used as a substrate. In contrast, much higher levels of activity were seen when p-coumaroyl methyl ester was used as a substrate (Figure 9a), although the apparent Km for this substrate (2.5 ± 0.1 mm; Figure 9b) was still higher than those of other phenylpropanoid pathway P450s, and the turnover number for this substrate was still very low (approximately 3 h−1). No activity was seen when cinnamate, caffeate or ferulate was used as a substrate for C3H.

Figure 9.

C3H activity in yeast microsomes.

(a) Comparison of relative activity of C3H using p-coumarate, p-coumaraldehyde, p-coumaryl alcohol, and p-coumarate methyl ester at a concentration of 1 mm.

(b) Lineweaver-Burk plot of p-coumaroyl methyl ester 3-hydroxylation. For both graphs, error bars represent one standard deviation for triplicate assays.


Phenylpropanoid metabolism has been studied for several decades; however, modern molecular, biochemical, and genetic investigations have led to many recent revisions in our thinking about how the products of the pathway are synthesized. One of the obstacles to building a comprehensive model of the phenylpropanoid pathway has been the dearth of knowledge concerning C3H. There are many reports of C3H activity in the literature, although few of them agree on what kind of enzyme C3H is, and on what its actual substrate might be. Given the difficulties associated with the study of C3H, we concluded that a genetic approach, similar to the one we previously employed to isolate the gene encoding F5H, might provide the best opportunity to identify C3H. The cloning of the REF8 gene demonstrates that, although many previous reports associated C3H with a phenolase activity in plant extracts, C3H is a P450, like the two other hydroxylases of lignin biosynthesis.

Although we could readily demonstrate that C3H can catalyse the conversion of p-coumarate to caffeate in living yeast cells and in vitro, accurate determination of the enzyme's kinetic constants proved to be problematic due to the low activity of the enzyme toward p-coumarate. Nevertheless, these experiments suggested the Km of C3H for p-coumarate is very high, approaching or possibly exceeding its solubility in water. The magnitude of this value is inconsistent with those determined for many other plant P450s, including C4H (Urban et al., 1994), and provides a plausible explanation for previous difficulties associated with demonstrating C3H activity in plant extracts using p-coumarate as a substrate. These data are also similar to those that prompted investigations into the substrate specificity of F5H. These experiments demonstrated that ferulate is not the preferred substrate for F5H (Humphreys et al., 1999; Osakabe et al., 1999), and that the enzyme instead has a much lower Km for the corresponding aldehyde and alcohol. Even more recently, COMT was demonstrated to show high activity toward caffeylaldehyde and caffeyl alcohol (Parvathi et al. 2001). These data led us to postulate that p-coumaraldehyde and p-coumaryl alcohol might be the preferred substrates for C3H in vivo. Instead, C3H was inactive towards p-coumaryl alcohol, and showed only low levels of activity towards p-coumaraldehyde. Although it is not currently feasible to determine the concentrations of p-coumarate and p-coumaraldehyde in the vicinity of C3H in vivo, these findings do not lend support to a model of monolignol biosynthesis in which the 3- and 5-hydroxylation and O-methylation steps occur at the aldehyde and/or alcohol level. On the other hand, the enzyme's comparatively high activity towards p-coumarate methyl ester is difficult to reconcile since this compound is not known to be a phenylpropanoid intermediate in plants. It seems more likely that p-coumarate methyl ester is acting as a surrogate for another substrate that we have not assayed in our study. In fact, while our manuscript was under review, another paper describing the characterization of CYP98A3 was published (Schoch et al., 2001). The authors of this work employed a bioinformatics approach to identify the gene encoding C3H, and showed that the enzyme is highly active towards p-coumaroyl quinate and p-coumaroyl shikimate (turnover numbers of approximately 400 and 600 min−1, respectively, compared to approximately 3 h−1 for p-coumaroyl methyl ester in our study). Their work thus supports previous research that demonstrated this P450-catalysed reaction in microsomal extracts of parsley and carrot suspension cultures (Heller and Kühnl, 1985; Kühnl et al., 1987). Although p-coumaroyl quinate and p-coumaroyl shikimate have not previously been considered to be lignin biosynthetic intermediates, immunolocalization of CYP98A3 revealed that the protein is expressed in xylem, providing circumstantial evidence that the protein is the 3-hydroxylase involved in lignin biosynthesis (Schoch et al. 2001). The loss-of-function phenotypes of the ref8 mutant (Franke et al., 2002) and the demonstration that the REF8 gene encodes CYP98A3 provides unambiguous proof that this P450 is C3H and is required for the synthesis of the precursors of wild-type lignin and sinapate esters in Arabidopsis.

The high apparent Km of C3H toward p-coumarate raises questions of how hydroxycinnamic acid conjugates are synthesized in plants. For C3H to be an effective catalyst with p-coumarate in vivo, the bulk concentration of this substrate within the cytoplasm would have to be very high. Alternatively, metabolic channelling between C4H and C3H could provide high local concentrations of p-coumarate in the vicinity of the C3H active site. The results of our work and the study by Schoch et al. (2001) would suggest that it is more likely that the synthesis of derivatives of caffeic acid instead requires enzymes that release caffeic acid from its shikimate, quinate or CoA esters for subsequent metabolism, or transferases that catalyse the direct transfer of caffeic acid from these pathway intermediates onto other acceptor molecules.

Experimental procedures

Plant material

Arabidopsis thaliana L. Heynh. ecotype Columbia were cultivated at a light intensity of 100 mE m−2 sec−1 at 23°C under a photoperiod of 16 h light/8 h dark in Redi-Earth potting mix (Scotts-Sierra Horticultural Products; Marysville, OH, USA).

O-methyltransferase enzyme assays

Protein extracts were prepared and COMT and CCoAOMT activity were measured as described previously (Inoue et al., 1998). Total protein content was measured using the Pierce BCA assay using bovine serum albumin as a standard.

Ethylene measurements

Rosettes of 4-week-old plants were harvested, weighed, and incubated under ambient laboratory lighting in a sealed 5 ml scintillation vial for 90 min ethylene content of the gas phase was measured as described previously (Jones and Woodson, 1999).

Radiotracer feeding experiments

U-14C-l-phenylalanine was administered to individual illuminated leaves of wild-type and mutant plants for 3 h before extraction in methanol at 60°C in the presence of unlabeled hydroxycinnamic acids, as described previously (Chapple et al., 1992). Extracts were analysed directly or after saponification for 30 min in 1 m NaOH followed by acidification and extraction of the radiolabelled products into diethyl ether. Extract components were resolved by a two-dimensional silica gel TLC system that separates all of the natural hydroxycinnamic acids (solvent 1, petroleum ether/ethyl acetate/methanol/acetic acid 10 : 10 : 1 : 0.2; solvent 2, toluene/acetic acid/water 2 : 1:sat.). Incorporation of label into pathway intermediates was determined using a Packard Instant Imager. To measure the incorporation of label from U-14C-l-phenylalanine specifically into caffeic acid, an identical feeding experiment was conducted; however, prior to analysis, the extract was incubated in methanolic HCl (3 m HCl in anhydrous methanol) (80°C, 1 h) to convert caffeic acid and its ester conjugates to caffeic acid methylester. Methylcaffeate was purified by semipreparative silica gel TLC using the two solvents described above. Following each round of TLC, the band of methylcaffeate was identified under UV light, scraped from the TLC plate, eluted in methanol, and after the second round of TLC, analysed by reversed phase HPLC (solvent A, 5% acetic acid in water; solvent B, 20% acetic acid, 25% acetonitrile in water; 5–55% B in 20 min, 55–100% B in 10 min, 100% B for 5 min; flow rate 1 ml min−1) using diode array UV detection. One ml fractions were collected and analysed for radioactivity by liquid scintillation counting.

Map-based cloning

The ref8 mutant (Columbia background) was used as the male parent in a cross to the Landsberg erecta ecotype. F1 individuals were allowed to self-pollinate, and F2 plants were screened for the ref8 phenotype. Because ref8 plants are small in stature and are female sterile, seeds from phenotypically wild-type plants (REF8/ref8 and REF8/REF8 individuals) were collected and the F3 progeny were scored for segregation of the ref8 phenotype. DNA was extracted (Doyle and Doyle, 1990) from homozygous wild-type lines for ARMS mapping (Schaffner, 1996) to determine an initial map position for the REF8 gene. Subsequently, DNA was extracted from additional F2 plants and F3 families for use in PCR-based genotyping experiments. Individuals carrying recombinant chromosomes in the region of the REF8 locus were used to determine a mapping interval for the gene, and were analysed further.

RNA gel blot analysis

For the isolation of RNA, plant tissues were harvested, frozen in liquid nitrogen, and stored at −70°C until ready for extraction. Total RNA was isolated as previously described (Goldsbrough and Cullis, 1981). Samples were electrophoretically separated, transferred to Hybond N+ membranes (Amersham), hybridized at 65°C with a DNA probe (DECAprime II system, Ambion) using a CYP98A3 EST ordered from the Arabidopsis Biological Resource Center (209A1T7; GenBank accession number N37715), washed, and exposed to film.

Isolation of the ref8 cDNA

The ref8 cDNA was isolated by reverse transcriptase-PCR from total RNA using the Promega Access RT–PCR system (Madison, WI, USA) using primer 1 (5′-gcaaggatccatgtcgtggtttctaatagcg-3′) and primer 2 (5′-tcaggaattcatttacatatcgtaaggcacg-3′). These primers correspond to the 5′ and 3′ ends of the open reading frame and introduce a BamHI site upstream of the start codon and an EcoRI site downstream of the stop codon, respectively, for use in subsequent yeast expression studies. Two independent reaction products were subcloned and sequenced to identify the mutation in the ref8 allele.

Complementation of the ref8 mutant

The REF8 genomic clones were identified in an Arabidopsis thaliana (ecotype Landsberg erecta) library generated in the binary cosmid vector pBIC20 (Meyer et al., 1996a) using the REF8 cDNA as a probe. Constructs for plant transformation were introduced into Agrobacterium tumefaciens C58 pGV3850 (Zambrisky et al., 1983) by electroporation. Plant transformation was performed by the floral dip method (Clough and Bent, 1998). Transformed seedlings (T1) identified by selection on MS medium containing 50 mg l−1 kanamycin and 200 mg l−1 timentin were transferred to soil.

The genotyping of the transformed plants was performed using two cleavable amplified polymorphic sequence (CAPS) markers that exploit the EcoRV polymorphism caused by the ref8 mutation, and a downstream HindIII polymorphism that was identified between the Columbia and Landsberg ecotypes. Using these two polymorphisms in concert, it was possible to identify ref8/ref8 plants (amplicons from the endogenous gene are cleaved by EcoRV but not HindIII), carrying either of two of the putative REF8 cosmids (plants are kanamycin-resistant and amplicons from the REF8 cosmid transgene are cleaved by HindIII but not EcoRV). The genotype of plants carrying the cosmid pCC554 was simpler to determine since the 3′ primer used in PCR lies downstream of the cosmid vector/insert border and thus does not anneal to the integrated transgene (Figure 6). As a result, only the endogenous gene is amplified and homozygous ref8 plants can be readily identified using only the EcoRV polymorphism.

Heterologous expression

To generate the pBOV E. coli P450 expression plasmid, pCWOri+ was first digested with HindIII and NdeI and purified by gel electrophoresis. The overlapping complementary primers 3 (5′-tatggctctgttattagcagtttttata-3′), 4 (5′-caggcctataaaaactgctaataacagagcca-3′), 5 (5′-ggcctgcatgccatcatcatcatcatcattag-3′), and 6 (5′-agctctaatgatgatgatgatgatggcatg-3′) were then phosphorylated using polynucleotide kinase and ligated into the pCWOri+ vector backbone. The resulting cloning site in the plasmid contains the first eight codons of the bovine CYP17α gene followed by a StuI site for the blunt-ended cloning of PCR-amplified P450 cDNAs, a downstream SphI site, a sequence coding for a 6x-His tag, and a stop codon.

Using the CYP98A3 cDNA as a template for PCR, the pBOV-REF8 plasmid was generated by first using primer 7 (5′-gacaatcgccgccgtcgtatcctac-3′) and primer 8 (5′-catatcgtaaggcacgcgtttgtac-3′) to produce a truncated version of the open reading frame that lacked the first nine codons of the protein's N-terminal signal peptide. This PCR product was subcloned into StuI-digested pBOV, the orientation of the insert was determined using diagnostic restriction digests, and the fidelity of the PCR process was verified by sequencing. To generate the pBOV-ref8 plasmid, the CYP98A3 EST was used in two separate reactions using primer 7 with primer 9 (5′-cggtgcacaacttgatatcaatttgg-3′) and primer 6 with primer 10 (5′-ccaaattgatatcaagttgtgcaccg-3′) to introduce the ref8 mutation (italicized in primers 9 and 10) into each of two overlapping fragments of the cDNA. The PCR products were purified by agarose gel electrophoresis, combined in a single PCR reaction, and amplified using only primers 7 and 6. The resulting full-length product was subcloned into pBS KS-, sequenced, and a StyI/SphI fragment containing the ref8 mutation was used to replace the corresponding portion of pBOV-REF8 to yield pBOV-ref8.

The construction of the Saccharomyces cerevisiae strain WAT11, a derivative of the W303-B strain (MAT a; ade2–1; his3–11,-15; leu2–3,-112; ura3–1; canR.; cyr+) expressing the ATR1 Arabidopsis NADPH-P450 reductase, was previously described (Pompon et al., 1996; Truan et al., 1993). For the construction of the YeDP60-REF8 expression construct, the CYP98A3 EST was used as the template for PCR as described above with primer 1 and primer 2. The resulting 1.5 kB PCR product was subcloned, sequenced, and ligated into BamHI/EcoRI digested YeDP60 (Urban et al., 1990) to yield the plasmid YeDP60-REF8. To generate the plasmid YeDP60-ref8, PCR using the CYP98A3 EST was conducted in two separate reactions using primer 1 with primer 9 and primer 2 with primer 10 to introduce the ref8 mutation into each of two overlapping fragments of the cDNA. The PCR products were purified by agarose gel electrophoresis, combined in a single PCR reaction, and amplified using only primers 1 and 2. The resulting full-length product was subcloned into pBS KS-, sequenced and subcloned into YeDP60 as described above.

Measurement of enzymatic activity in vivo

WAT11 cells were transformed with YeDP60, YeDP60-REF8, and YeDP60-ref8 (Gietz et al., 1992), cultured and induced with galactose as described previously (Urban et al., 1994). For in vivo measurements of enzyme activity, cells were grown in 30 ml media supplemented with 5 mmp-coumaric acid. At the end of the 24 h incubation period the medium was extracted with ethylacetate, and analysed by HPLC and GC–MS. SDS-PAGE analysis of heterologusly expressed C3H and CO difference spectroscopy were performed as described previously (Humphreys et al., 1999).

Measurement of enzymatic activity in vitro

For C3H assays, an NADPH regenerating system consisting of 1 mm NADP+, 10 mm glucose-6-phosphate, and 1 unit glucose-6-phosphate dehydrogenase was preincubated at 30°C for 5 min to permit the generation of NADPH in the presence of one of the putative C3H substrates in a final volume of 400 µl of assay buffer. Assays were initiated by the addition of 100 µl of microsomes and were allowed to incubate for 60 min at 30°C before being terminated by boiling. Assays were clarified by centrifugation for 20 min at 13 000 g, and were analysed directly by HPLC on a Microsorb-MV C-18 column (Rainin, Woburn MA, USA) using a gradient of solvent A (1.0% acetic acid in water) and solvent B (acetonitrile; 0–10% B in 5 min, 10–25% B in 25 min; flow rate 1 ml min−1). C3H reaction products were quantified using UV detection (caffeic acid, 322 nm; caffeyl aldehyde, 340 nm; caffeic acid methyl ester, 324 nm). Assays conducted using microsomes isolated from yeast transformed with YeDP60 served as negative controls. The apparent Km for p-coumarate methyl ester was determined using triplicate assays analysed by the Eadie-Hofstee method (Cornish-Bowden, 1995), and P450 content determined on crude yeast microsomes.


Analysis of C3H reaction products was performed by GC–MS (GCMS-QP5050A, Shimadzu Corp. Kyoto, Japan). Standards were prepared from methanolic stock solutions of the substrates and products. Aliquots of each stock solution were taken in order to obtain a solution having approximately 1000 pmol µl−1 of each standard. The standards solution and the assay samples were dried by centrifugal evaporation (RC 10.10, Jouan, Winchester VA, USA) resuspended in 100 and 50 µl of pyridine, respectively, and derivatized using 10 and 5 µl N-methyl-N-trimethylsilyltrifluoro-acetamide (MSTFA), respectively, at 37°C for 30 min. One microliter aliquots of each sample were injected with the split ratio set to 1 : 25. Helium was used as the carrier gas with a flow rate of 1 ml min−1. The injector temperature was maintained at 220°C. Gas chromatography was performed using a 30-m DB-5MS column (0.25 mm I.D., 0.25 µm film thickness) (J & W Scientific, Folsom CA, USA). The column temperature was initially maintained at 125°C for 2 min and then ramped at 15°C min−1 to 250°C and held for 4 min. The interface temperature was held at 280°C. After a 4.95-min solvent cut time, mass analysis of the column eluent was performed using a single quadrupole mass filter set to scan from mass-to-charge 50–410 in 0.17 sec.


The authors are grateful to John Ralph and Hoon Kim, as well as Stephen Fry for the gifts of p-coumaraldehyde and p-coumaryl alcohol and to Richard Dixon for authentic samples of the caffeyl aldehyde and caffeyl alcohol. This work was supported by a grant from the Division of Energy Biosciences, United States Department of Energy, and the National Science Foundation. This is journal paper number 16716 of the Purdue University Agricultural Experiment Station.