Mutations in the cinnamate 4-hydroxylase gene impact metabolism, growth and development in Arabidopsis


  • Anthony L. Schilmiller,

    Search for more papers by this author
    • Co-first authors.

    • Present Addresses: Anthony L. Schilmiller, Department of Biochemistry and Molecular Biology, Michigan State University, East Lansing, MI 48824, USA.
      Jake Stout, National Research Council of Canada, Plant Biotechnology Institute, 110 Gymnasium Place, Saskatoon, Saskatchewan, S7N 0W9, Canada.
      John Humphreys, UT Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX 75390-8816,USA.
      Max O. Ruegger, Dow AgroSciences LLC, 9330 Zionsville Road, Indianapolis, IN 46268, USA.

  • Jake Stout,

    Search for more papers by this author
    • Co-first authors.

    • Present Addresses: Anthony L. Schilmiller, Department of Biochemistry and Molecular Biology, Michigan State University, East Lansing, MI 48824, USA.
      Jake Stout, National Research Council of Canada, Plant Biotechnology Institute, 110 Gymnasium Place, Saskatoon, Saskatchewan, S7N 0W9, Canada.
      John Humphreys, UT Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX 75390-8816,USA.
      Max O. Ruegger, Dow AgroSciences LLC, 9330 Zionsville Road, Indianapolis, IN 46268, USA.

  • Jing-Ke Weng,

    1. Department of Biochemistry, Purdue University, West Lafayette, IN 47907, USA
    Search for more papers by this author
  • John Humphreys,

    Search for more papers by this author
    • Present Addresses: Anthony L. Schilmiller, Department of Biochemistry and Molecular Biology, Michigan State University, East Lansing, MI 48824, USA.
      Jake Stout, National Research Council of Canada, Plant Biotechnology Institute, 110 Gymnasium Place, Saskatoon, Saskatchewan, S7N 0W9, Canada.
      John Humphreys, UT Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX 75390-8816,USA.
      Max O. Ruegger, Dow AgroSciences LLC, 9330 Zionsville Road, Indianapolis, IN 46268, USA.

  • Max O. Ruegger,

    Search for more papers by this author
    • Present Addresses: Anthony L. Schilmiller, Department of Biochemistry and Molecular Biology, Michigan State University, East Lansing, MI 48824, USA.
      Jake Stout, National Research Council of Canada, Plant Biotechnology Institute, 110 Gymnasium Place, Saskatoon, Saskatchewan, S7N 0W9, Canada.
      John Humphreys, UT Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX 75390-8816,USA.
      Max O. Ruegger, Dow AgroSciences LLC, 9330 Zionsville Road, Indianapolis, IN 46268, USA.

  • Clint Chapple

    Corresponding author
      (fax +765 496 7213; e-mail
    Search for more papers by this author

(fax +765 496 7213; e-mail


The initial reactions of the phenylpropanoid pathway convert phenylalanine to p-coumaroyl CoA, a branch point metabolite from which many phenylpropanoids are made. Although the second enzyme of this pathway, cinnamic acid 4-hydroxylase (C4H), is well characterized, a mutant for the gene encoding this enzyme has not yet, to our knowledge, been identified, presumably because knock-out mutations in this gene would have severe phenotypes. This work describes the characterization of an allelic series of Arabidopsis reduced epidermal fluorescence 3 (ref3) mutants, each of which harbor mis-sense mutations in C4H (At2g30490). Heterologous expression of the mutant proteins in Escherichia coli yields enzymes that exhibit P420 spectra, indicative of mis-folded proteins, or have limited ability to bind substrate, indicating that the mutations we have identified affect protein stability and/or enzyme function. In agreement with the early position of C4H in phenylpropanoid metabolism, ref3 mutant plants accumulate decreased levels of several different classes of phenylpropanoid end-products, and exhibit reduced lignin deposition and altered lignin monomer content. Furthermore, these plants accumulate a novel hydroxycinnamic ester, cinnamoylmalate, which is not found in the wild type. The decreased C4H activity in ref3 also causes pleiotropic phenotypes, including dwarfism, male sterility and the development of swellings at branch junctions. Together, these observations indicate that C4H function is critical to the normal biochemistry and development of Arabidopsis.


Plants made the transition from aquatic to terrestrial habitats during the Silurian period, approximately 400 Ma (Kenrick and Crane, 1997). Within this period of plant evolution, plants developed an array of strategies to cope with life on land, most notably the ability to transport water and resist desiccation, to defend themselves against harmful wavelengths of solar radiation, and to elaborate an upright and self-supporting plant body. The phenylpropanoid pathway was key to this evolutionary transition, in that flavonoids and other phenylpropanoid derivates provide protection from UV-induced DNA damage, and lignin provides structural support to both individual tracheary elements and the stem as a whole. In addition, lignin impedes the breakdown of cell wall polysaccharides to simple sugars, thus imparting decay resistance to highly lignified tissues such as wood. As a result, lignin has negative impacts on many agricultural uses of plants, and interferes with forage digestibility, pulping and, more recently, biofuel production (Chen and Dixon, 2007; Li et al., 2008; Weng et al., 2008a).

The first three biosynthetic reactions in phenylpropanoid metabolism are often referred to as the general phenylpropanoid pathway, because they produce p-coumaroyl CoA, a major branch-point metabolite between the production of the flavonoids and the pathway that produces monolignols, lignans and hydroxy-cinnamate conjugates (Figure 1) (Winkel-Shirley, 2001; Boerjan et al., 2003). The first of these reactions is the deamination of phenylalanine by phenylalanine ammonia-lyase (PAL) to generate trans-cinnamic acid. Cinnamic acid is then para-hydroxylated by cinnamate 4-hydroxylase (C4H) to produce p-coumaric acid (Russell and Conn, 1967; Russell, 1971), which is then activated to its corresponding CoA thioester by 4-coumarate CoA ligase (4CL).

Figure 1.

 The general phenylpropanoid pathway.
The first three steps of the phenylpropanoid pathway are catalyzed by the enzymes phenylalanine ammonia-lyase (PAL), cinnamate 4-hydroxylase (C4H) and 4-coumaroyl CoA ligase (4CL). Together, these steps are often referred to as the ‘general phenylpropanoid pathway’ because the steps they catalyze, and the products they produce, are shared between pathways leading to flavonoids, monolignols, sinapate esters and lignans.

Phenylpropanoid metabolism has been the target of intense investigation, and most, if not all, of the biosynthetic genes of the pathway have been cloned and characterized. In particular, the gene encoding C4H was one of the first phenylpropanoid pathway genes to be cloned (Fahrendorf and Dixon, 1993; Mizutani et al., 1993; Teutsch et al., 1993). Following its initial isolation, C4H has been cloned and characterized from many plant species (Hotze et al., 1995; Frank et al., 1996; Akashi et al., 1997; Mizutani et al., 1997; Koopmann et al., 1999; Nedelkina et al., 1999; Betz et al., 2001; Luo et al., 2001; Ro et al., 2001; Hubner et al., 2003; Gravot et al., 2004), and has been studied in detail at the enzymatic level following expression in yeast (Urban et al., 1994; Ro and Douglas, 2004). There is one copy of C4H in Arabidopsis (Bell-Lelong et al., 1997); thus, any carbon flux through phenylpropanoid metabolism is mediated by the activity of the protein encoded by this single gene. The expression of C4H in Arabidopsis is apparent in seedlings soon after seed imbibition, and is observable in most organs during all stages of growth (Bell-Lelong et al., 1997). This expression pattern has been corroborated by microarray analysis (Schmid et al., 2005).

In recent years, Arabidopsis has rapidly become a choice model species in which to study phenylpropanoid metabolism, in part because of the genetic screens that have been made possible by the production of sinapoylmalate, a UV-protectant in Arabidopsis (Landry et al., 1995), which accumulates in the leaf adaxial epidermis (Chapple et al., 1992). Sinapoylmalate serves as an excellent genetic marker for mutations in genes that are involved in phenylpropanoid metabolism because of its blue-green fluorescence under UV light (Ruegger and Chapple, 2001). Lesions that decrease the accumulation of sinapoylmalate lead to a reduced epidermal fluorescence (ref) phenotype that results from the red fluorescence of chlorophyll in the underlying mesophyll. The ref mutants derived from this screen have been instrumental in the isolation of a number of genes involved in phenylpropanoid metabolism (Humphreys and Chapple, 2002; Nair et al., 2004; Stout and Chapple, 2004; Stout et al., 2008).

The observation that mis-sense mutations in p-coumaroyl shikimate 3′-hydroxylase (C3′H), and the downregulation of hydroxycinnamoyl CoA shikimate/quinate hydroxycinnamoyltransferase (HCT), severely impacts plant growth suggests that phenylpropanoid metabolism is necessary for plant survival, even under controlled laboratory conditions (Franke et al., 2002a,b; Hoffmann et al., 2004). Thus, it is reasonable to suppose that C4H mutants have not been identified previously because loss-of-function mutations are lethal, or severely affect plant development. Here, we report the characterization of the Arabidopsis ref3 mutants that constitute an allelic series bearing mis-sense mutations in the gene encoding C4H. These mutants, to varying degrees, show phenotypes that include male sterility, dwarfism, decreased lignin deposition and vascular collapse. Furthermore, these mutants accumulate cinnamoylmalate in place of sinapoylmalate, demonstrating the plasticity of phenylpropanoid secondary metabolism.


The ref3 mutation maps near the C4H gene

The initial characterization of the ref3 mutant revealed that the mutant exhibited several additional biochemical defects associated with phenylpropanoid metabolism, including reductions in the levels of sinapoylmalate (Figure 2), as well as reduced flavonoids and decreased levels of lignin (Ruegger and Chapple, 2001). In order to determine what role REF3 plays in these processes, we set out to isolate the ref3 gene using map-based cloning procedures. To do so, a mapping population was constructed between Columbia and ref3-1 (Landsberg erecta background). Fortuitously, in the F2 generation, almost all of the homozygous ref3 individuals were also homozygous for erecta, indicating that the REF3 locus is linked to ERECTA. Considering that ERECTA is 10 cM above C4H on chromosome 2, and that a defect in C4H would be consistent with the observed phenylpropanoid phenotypes, we took a candidate gene approach to determine if the REF3 locus is C4H. As an initial test of this hypothesis, a StyI polymorphism in C4H was used as a cleaved amplified polymorphic sequence (CAPS) marker. Twenty-four homozygous ref3-1 individuals from the mapping population were analyzed with this marker, and all were homozygous for Landsberg erecta-derived mutant alleles of C4H, suggesting that REF3 is below ERECTA on chromosome 2, which is relatively close to the C4H locus.

Figure 2.

 Analysis of sinapoylmalate (λ = 335 nm) and cinnamate esters (λ = 280 nm) in leaves of wild type, ref3-3, ref3-1 and ref3-2 by HPLC.
Sinapoylmalate can be readily detected in extracts of wild-type plants using UV detection at 335 nm. Extracts of members of the ref3 allelic series contain progressively less sinapoylmalate, but contain increasing quantities of cinnamoylmalate, which only absorbs UV light at shorter wavelengths, such as 280 nm. The ref3-1 mutant is in the Landsberg erecta background. CM, cinnamoylmalate; SM, sinapoylmalate.

ref3 accumulates cinnamoylmalate in place of sinapoylmalate

When analyzed at 280 nm, extracts of all ref3 plants were found to contain a pair of late-eluting compounds that were not detected in wild-type extracts when analyzed by High performance liquid chromatography (HPLC) using UV detection at 335 nm (Figure 2). To identify these compounds, ref3-2 extracts were analyzed by HPLC-MS. Each compound generated a molecular ion with an m/z of 263, and a fragment ion of m/z 147, suggesting that these compounds are both cinnamoylmalate, likely the cis- and trans- isomers (Figure S2), again consistent with the hypothesis that REF3 encodes C4H. When summed with the residual sinapoylmalate in the mutants ref3-3 and ref3-2, rosettes accumulated 0.9 ± 0.04 and 0.6 ± 0.01 nmol g−1 of the combined hydroxycinnamate esters, respectively, whereas wild-type rosettes contained 1.7 ± 0.03 nmol g−1 of sinapoylmalate, demonstrating that despite the accumulation of novel pathway derivatives, the total accumulation of hydroxycinnamic esters is reduced in the mutants.

To further test the hypothesis that the ref3 mutation was in C4H, ref3-2 plants were sprayed with p-coumaric acid in an attempt to chemically complement the mutant’s ref phenotype. We speculated that if the ref3 mutation inactivates C4H, then spraying the plants with p-coumaric acid, the product of the C4H reaction, might restore the synthesis of sinapoylmalate. In these experiments, 3-week-old ref3-2 plants that that had been sprayed daily for 1 week with 1 mmp-coumarate were substantially more blue-fluorescent under UV light (data not shown). In contrast, ref3-2 plants that were sprayed with a control solution lacking p-coumarate showed no increase in fluorescence. HPLC analysis of leaf extracts from the plants sprayed with p-coumaric acid showed a significant increase in sinapoylmalate levels compared with control plants (Figure 3), consistent with the hypothesis that the ref3 mutant is defective in C4H.

Figure 3.

 Chemical complementation of the ref3 phenotype.
Spraying ref3-2 with an aqueous solution of p-coumaric acid containing 1% DMSO and 0.02% Silwet leads to a partial restoration of sinapoylmalate accumulation. HPLC analysis (λ = 335 nm) was performed on extracts of leaves of wild type (a), untreated (b) and p-coumaric acid treated (c) ref3-2.

REF3 encodes C4H

The ability of p-coumarate to rescue the ref3 phenotype, the accumulation of cinnamic acid esters in ref3 leaves, and the map position of the REF3 gene were all consistent with the hypothesis that the ref3 mutation is within the gene encoding C4H. To provide molecular evidence for this hypothesis, we sequenced C4H (At2g30490) in the three ethyl methanesulfonate-induced ref3 alleles. Each mutant allele contained a single G → A transition, which results in a mis-sense mutation: GCG → ACG for ref3-1 (A306T), AGA → AAA for ref3-2 (R249K) and GGA → GAA for ref3-3 (G99E) (Figures 4a and S1).

Figure 4.

 Characterization of the proteins encoded by mutant C4H alleles.
(a) The position of the ref3 mutations and their subsequent amino acid changes.
(b) CO difference spectra. The wild-type protein is found mostly in the native P450 form, whereas each mutant protein is predominantly in the inactive, P420 form.
(c) Substrate binding difference spectroscopy of the protein encoded by the ref3-3 allele. Double reciprocal plot analysis yielded Kd values for cinnamate of approximately 4 μm, comparable with that of the wild-type protein.

To definitively determine whether the mutations identified in the ref3 alleles are responsible for the sinapoylmalate-deficient phenotype of the mutant, ref3-3 was transformed with a 5.4-kb genomic C4H clone carrying 2.9 kb of the C4H promoter and the entire C4H open reading frame (Bell-Lelong et al., 1997). HPLC analysis of extracts of the kanamycin-resistant transformants indicated that these plants accumulated wild-type levels of sinapoylmalate (Figure 5).

Figure 5.

 Complementation of the ref3 mutant phenotype.
Sinapoylmalate content in leaf extracts of the wild type, the ref3-3 mutant and ref3-3 transgenic plants carrying the wild-type C4H transgene was analyzed by HPLC (λ = 335 nm); SM, sinapoylmalate.
(a) Columbia wild-type plants.
(b) ref3-3 plants.
(c) ref3-3 transformed with the C4H genomic construct.

C4H function is compromised in ref3

The ref3 mutants comprise an allelic series, but none contains a non-sense mutation in C4H. To determine the manner and extent to which the mis-sense mutations in each of the mutants affect enzyme function, we expressed native and mutant C4H proteins in E. coli, and used spectroscopy to evaluate protein stability and substrate binding. CO difference spectroscopy of all four proteins showed a difference spectrum with signals at 450 nm, characteristic of properly folded active protein, and at 420 nm, characteristic of an inactive P450, with a disrupted tertiary structure (Figure 4b). For native C4H, the level of P450 relative to the total of P450 and P420 was observed to be about 80%, whereas, for the ref3 mutants, this value ranged from 20 to 40%. Moreover, the CO difference spectrum observed for ref3-1 changed with time, with a half-life of only a few minutes, suggesting that, at least under these experimental conditions, the ref3-1 mutant protein is not stable. Spectroscopic determination of cinnamic acid binding to the proteins encoded by the wild-type and ref3-3 alleles yielded Kd values of approximately 4 μm, indicating that an inability to bind substrate is not the cause of the ref3-3 phenotype. In contrast, under identical conditions, the ref3-2 mutant protein showed no spectroscopic shift indicative of P450 substrate binding, suggesting that in addition to being improperly folded or unstable during manipulation in vitro, the catalytic function of the enzyme may be severely affected. The instability of ref3-1 precluded the evaluation of cinnamic acid binding.

Reductions in lignin content in ref3 mutants lead to collapsed vasculature

All phenylalanine-derived units destined to be incorporated into the lignin polymer must be hydroxylated by C4H, because the p-hydroxy group is required for the activation of monolignols to their corresponding free radicals, and for polymerization into lignin. Thus, strong reductions in C4H activity would be expected to lead to a decrease in monolignols available for lignin deposition. Consistent with this hypothesis, a preliminary analysis of lignin composition suggested that ref3 stems have decreased levels of total lignin (Ruegger and Chapple, 2001). The chemically-resistant nature of lignin makes unambiguous analysis difficult, so we chose to re-analyze the lignin content of ref3 stems using the more rigorous Klason method. These analyses revealed data comparable with those previously derived from the thioglycolic acid method: ref3-3, which accumulates the highest levels of sinapoylmalate, deposits normal levels of lignin, and the ref3-1 and ref3-2 mutants deposit approximately one half and one quarter, respectively, of the lignin that is found in the wild type (Table 1). These reductions in lignin content in plants carrying the stronger alleles result in a collapsed xylem phenotype like those previously observed in other Arabidopsis mutants affected in lignin and cellulose biosynthesis (Taylor et al., 1999; Jones et al., 2001). Although tracheary elements in ref3-3 vascular bundles look relatively normal, vessel elements in ref3-1 and ref3-2 are almost entirely collapsed, consistent with the low levels of lignin deposited in these lines (Figure 6).

Table 1.   Lignin content of wild type ref3 stems (±represents one standard error; n = 3). The ref3-1 mutant is in the Landsberg erecta background
GenotypeKlason lignin content (% dry weight ± SE)
Columbia wild type16.9 ± 0.8
ref3-19.3 ± 0.2
ref3-23.3 ± 0.3
ref3-316.6 ± 0.3
Figure 6.

 Anatomy of vascular bundles from wild-type and ref3 stems.
Rachis segments were embedded in resin, sectioned and stained with Toluidine Blue O. Vessel elements in wild-type vascular bundles are large and open, whereas those in ref3-3 show some wall irregularities, and those in ref3-1 and ref3-2 are almost completely collapsed. The ref3-1 mutant is in the Landsberg erecta background.

The lignin deposited in ref3 contains fewer guaiacyl-derived subunits

Decreases in C4H expression in tobacco and alfalfa have previously been reported to cause a decrease in the ratio of syringyl- to guaiacyl-derived subunits in lignin (Sewalt et al., 1997; Reddy et al., 2005). To determine whether there are similar changes in the types of lignin deposited in ref3, isolated stem cell walls were analyzed using the derivitization followed by reductive cleavage (DFRC) method. For these experiments, the weakest ref3 allele, ref3-3, was chosen for analysis because it is developmentally the most similar to the wild type, and thus we could avoid confounding our analyses by changes in lignification that might accompany perturbations in plant development in ref3-1 and ref3-2. In contrast with Klason lignin analysis (Table 1), less total DFRC-lignin was detected in the ref3-3 samples, as compared with the wild type (Table 2). More importantly, this decrease in total lignin appears to be caused exclusively by a reduction in guaiacyl-derived subunit content, leading to a substantial increase in the mole percentage of syringyl-subunits in the lignin of the mutant.

Table 2.   Lignin quality of wild type and ref3-3 stems as measured by DFRC (±represents one standard deviation; n = 3)
GenotypeTotal DFRC lignin mg g−1 cell wallμmol G g−1 cell wallμmol S g−1 cell wallLignin mol % S
Col WT18.4 ± 0.772.6 ± 2.025.2 ± 1.825.7 ± 1.0
ref3-311.9 ± 1.534.9 ± 3.726.9 ± 4.043.4 ± 1.2

ref3 mutations perturb normal plant growth and development

As mentioned previously, the ref3 mutants represent a partial loss-of-function allelic series, in which ref3-3 is most like the wild type in terms of sinapoylmalate accumulation, and in which ref3-2 is the most severely affected. Paralleling these changes in phenylpropanoid metabolism are changes in the growth and development of the mutants. The most obvious perturbations are evident in the altered architecture of ref3 plants. Plants homozygous for the ref3-3 allele are relatively normal in appearance; whereas ref3-1 and ref3-2 mutants are dwarfed, and show a loss of apical dominance (Figure 7a). The most surprising developmental abnormality observed in ref3 is the presence of swellings at the base of lateral branches of plants homozygous for the two strongest alleles (Figure 7b). Transverse sections of ref3-1 branch-points reveal that the swellings consist of large, disordered cortical cells (Figure 7c).

Figure 7.

 Developmental abnormalities in the ref3 mutants.
Although the development of ref3-3 mutants is comparable with that of the wild type, plants homozygous for ref3-1 and ref3-2 alleles are dwarfed, and show changes in apical dominance (a). The two most severe ref3 alleles also lead to the development of swellings at inflorescence nodes (b). A cross section of a ref3-1 internode shows disordered cortex cells at the swelling site (c). The ref3-1 mutant is in the Landsberg erecta background.

Strong ref3 alleles affect pollen development

Strong ref3 alleles are male sterile (Ruegger and Chapple, 2001). To further explore the male sterility in ref3-2, fully opened flowers and mature anthers were examined for alterations in morphology. The general appearance of the floral organs in ref3-2 was similar to the wild type, indicating that the sterility of the mutant is unlikely to be mechanical, such as being caused by reduced filament length. In contrast, closer examination of the stamens of open, fully mature flowers by scanning electron microscopy revealed that by the time that wild-type anthers were shedding pollen, ref3-2 anthers had failed to dehisce (Figure 8a,b).

Figure 8.

 Pollen development is defective in ref3-2 plants.
Scanning electron microscopy shows pollen is clearly visible on wild-type anthers (a), whereas the anthers of ref3-2 plants are indehiscent (b). Fluorescence microscopy reveals that fluorescent pollen can be visualized within wild-type anthers (c), whereas no pollen is visible in mutant anthers (d). Toluidine Blue O staining almost mature anthers similarly reveals normal pollen develops within wild-type anthers (e), whereas none is present in mutant anthers (f). Scale bars: 100 μm.

Although an indehiscent anther would be sufficient to explain the male sterility phenotype of the mutant, we also considered the possibility that this defect was only an indirect result of defects in pollen development resulting more directly from changes in C4H activity. To examine this possibility, wild-type and ref3-2 anthers were examined by fluorescence and bright-field microscopy (Figure 8c,d). The auto-fluorescence of pollen grains was readily observable in the wild-type anthers prior to dehiscence. In contrast, the ref3-2 anthers exhibited reduced auto-fluorescence, and no pollen grains were discernable. This observation suggests that mature pollen is not found in the ref3-2 anthers; however, it was also possible that pollen develops in the mutant, but is not auto-fluorescent. To test this possibility, anthers were examined for the presence of pollen using toluidine blue O. As expected, large numbers of pollen grains were clearly observed in wild-type anthers (Figure 8e). No mature pollen was observed in ref3-2 anthers sampled from open flowers (Figure 8f), nor from anthers sampled from earlier developmental stages (not shown). Furthermore, no pollen grains were observed when mature ref3-2 anthers were mechanically opened (data not shown). These observations demonstrate that a functional C4H is necessary for pollen development in Arabidopsis.


Phenylpropanoid metabolism provides for the plant a suite of compounds that are key to growth and survival. The work presented in this paper provides direct forward-genetic evidence that reduction in C4H activity results in both a strong reduction in phenylpropanoid content and perturbations in growth and development. Previous reverse-genetic antisense and RNAi approaches have also indicated that strong reductions in the activity of early enzymes of the phenylpropanoid pathway lead to altered metabolism and growth (Sewalt et al., 1997; Blee et al., 2001; Hoffmann et al., 2004; Reddy et al., 2005), and this research has opened up new perspectives on the roles of phenylpropanoid metabolites in plants, and the potential impacts of their mis-accumulation. The identification of the ref3 mutant provides a new opportunity to assess these findings in another plant species, and the ability to do so in a situation that avoids issues associated with transgene promoter specificity and relative tissue efficacy of antisense or RNAi constructs, both of which can add ambiguity to experiments of this type (Blount et al., 2000; Reddy et al., 2005).

Defects in C4H lead to perturbations in phenylpropanoid metabolism

The ref3 phenotypes, including decreased lignin, decreased sinapoylmalate and the reduction in condensed tannin content of seeds (Ruegger and Chapple, 2001), are strongly indicative of a defect early in the phenylpropanoid pathway, and are entirely consistent with mutations in C4H. The Arabidopsis genome contains one gene that encodes C4H (Bell-Lelong et al., 1997), and therefore a plant homozygous for a null C4H allele would presumably be severely affected. The finding that only ‘leaky’C4H alleles were identified in our original mutant screen supports this hypothesis. Indeed, identification of a T-DNA insertional C4H line (GABI line #753B06) reveals that C4H-null mutants display a seedling lethal phenotype in which the seedlings germinate, but primary leaves fail to expand (data not shown). Furthermore, each of the ref3 alleles displays a different severity of mutant phenotypes, forming an allelic series, with ref3-2 showing the most severe phenotypes, followed by ref3-1 then ref3-3.

In the leaves of all ref3 mutants, HPLC analysis revealed the accumulation of cinnamoylmalate (Figure 2). The presence of these compounds provides further evidence that plant secondary metabolism is highly plastic, in that the suite of enzymes involved in sinapoylmalate biosynthesis, or homologs thereof, can support the production of cinnamoylmalate when sinapic acid production is blocked. In the biosynthesis of sinapoylmalate, the activation of sinapic acid is catalyzed by the BRT1 glucosyltransferase (UGT84A2) (Sinlapadech et al., 2007). In the context of the ref3 mutant, it is important to note that this enzyme is unable to utilize cinnamic acid as a substrate in vitro (Lim et al., 2001). Thus, in the ref3 mutant, it seems likely that other UDPG-glucosyltransferases, possibly UGT84A1 or UGT84A3, may be involved in cinnamic acid glucosylation (Lim et al., 2001).

Finally, the observation that the combined content of lignin, flavonoids and hydroxycinnamic acid esters in the mutants is substantially less than in the wild type suggests that phenylpropanoid homeostasis is perturbed in the ref3 mutants. This observation may indicate that there is either decreased synthesis of cinnamic acid in ref3 plants, or enhanced turnover. Although we know virtually nothing about phenylpropanoid catabolism, it seems more likely that reduced C4H activity may cause an increase in free cinnamic acid content in vivo, causing inhibition of PAL and reduced entry of carbon into phenylpropanoid metabolism (Sato et al., 1982; Blount et al., 2000).

The ref3 mutations result in changes in C4H stability and substrate binding

The amino acid changes induced in the mutant REF3 proteins reveals elements of primary structure required for wild-type C4H function and/or stability (Figure S1). The mutation in ref3-1 converts an alanine (A306) to a threonine residue. A306, or its counterpart in homologous sequences, is conserved in all land plant C4Hs, and is embedded in the conserved motif IVENINVAAIETTLWS. Based on the alignment of land plant C4Hs together with those P450s with known structure, this residue belongs to substrate recognition sequence 4 (SRS4), and includes the oxygen binding motif. SRS4 is on the P450 α-helix I (Gotoh, 1992). With respect to other lignin biosynthetic P450s, this A residue is also conserved in C3′Hs, but in a different flanking sequence context. It is not conserved between C4Hs and F5Hs. A previous study has also revealed that substitution of A306 with glycine led to protein instability and improper positioning of the substrate in the active site of the Jerusalem Artichoke C4H (Schalk et al., 1999), consistent with the deleterious phenotype that we observed in the ref3-1 mutant. The ref3-2 mutation alters R249 to a lysine residue. This arginine residue is conserved within a larger conserved motif, CxxxxxxRxxxFxxxFxxxR, found in all land plant C4Hs studied to date. This residue is part of SRS3, and probably faces the active site as part of the P450 G-helix. This R residue is also conserved in C3′Hs, again in a different sequence context, and again is not conserved between C4Hs and F5Hs (Figure S1). The mutation in ref3-3, G99E, alters a residue that is conserved in all land plant C4Hs in the conserved motif VEFGSR. This residue belongs to SRS1, which includes the B′α-helix, and probably is part of a region of coil between helix B and B′. This residue is not conserved in other phenylpropanoid P450s (Figure S1).

Defects in C4H lead to changes in plant development

An unexpected phenotype of plants homozygous for strong ref3 alleles are the stem enlargements that form at the branch points between lateral branches and the main rachis (Figure 7b). How and why these structures form is currently unknown; however, it is interesting to note that cis-cinnamate has been reported to function as an anti-auxin (Vanoverbeek et al., 1951; Wong et al., 2005). Although it is not clear that this activity is relevant in wild-type plants, it is possible that increases in the content of cinnamic acid or cinnamate-derived compounds at the nodes of ref3-1 and ref3-2 plants, possibly synthesized by the cauline leaves that subtend these nodes, may inhibit local auxin transport, leading to the aberrant growth observed.

Many plants defective in, or downregulated for, lignin biosynthesis exhibit alterations in normal development or phenotypes that may not be directly linked to altered lignification (Franke et al., 2002b; Hoffmann et al., 2004). The causes for these aberrations are often unclear, but in some cases include vascular collapse, leading to phenotypes that appear to be associated with drought stress (Vermerris et al., 2002). In another case, it has recently been shown that the reductions in growth in hydroxycinnamoyl CoA shikimate/quinate HCT-silenced plants is due to an over-accumulation of flavonoids, some of which can inhibit auxin transport (Buer and Muday, 2004; Peer et al., 2004; Besseau et al., 2007). It is also interesting that metabolic bottlenecks earlier in the pathway more frequently lead to altered growth than do perturbations later in the pathway. With this in mind, it is probably not surprising that the strongest ref3 alleles are dwarf mutants. Given that the reductions in C4H activity in ref3 decrease the flavonoid content in the mutant (i.e. lead to a transparent testa phenotype), it seems unlikely that a flavonoid-mediated reduction in auxin transport is the cause of dwarfism in this mutant. In addition, considering that sinapate esters are dispensable for normal development, it seems apparent from the ref3 allelic series that plants can tolerate a substantial accumulation of cinnamoylmalate without severe phenotypic consequences (i.e. in ref3-3). In contrast, a threshold is crossed with the ref3-1 and ref3-2 mutants, in terms of soluble esters, lignin deposition, cinnamate-mediated inhibition of auxin transport or some as yet unidentified factor, which leads to dwarfing, and, in the most severe conditions (i.e. ref3-2), male sterility.

ref3-2 is male sterile

In maize and petunia, it has been shown that an absence in flavonoids leads to male sterility (Taylor and Jorgensen, 1992). These mutants are not deficient in pollen development, as seen in ref3-2, but rather are infertile because of the inability of the pollen tubes to penetrate the pistil and extend towards the ovule. In contrast, the Arabidopsis tt4 mutant that is null for chalcone synthase produces pollen that is fully fertile (Burbulis et al., 1996), demonstrating that the requirement for flavonoids in pollen tube growth is specific for only some species. Phenylpropanoid metabolism has also been implicated in playing a role in anther dehiscence. This process is affected in the Arabidopsis ms35 mutant, which is deficient in a transcription factor, AtMyb26, required for the secondary cell wall thickening of the endothecium necessary for dehiscence (Dawson et al., 1999; Steiner-Lange et al., 2003). The authors postulated that this transcription factor controls both the cellulose and lignin deposition that is necessary for the differential thickening of the endothecium, which in turn results in the rupture of the stomium (Manning, 1996). It should, however, be noted that these mutants form fully viable pollen.

In contrast, the male sterility of ref3-2 appears to arise because of perturbations in pollen development. Sporopollenin is a major structural component of the pollen wall that provides both resistance against dehydration and protection against UV radiation (Edlund et al., 2004). This compound is a heteropolymer of p-coumaric acid ester linked to long-chain and medium-chain modified fatty acids (Wehling et al., 1989; Morant et al., 2007). Given that C4H activity is required for the production of the p-coumaric acid moieties of the polymer, the defective C4H alleles in ref3 may lead to a deficiency in sporopollenin deposition, to the extent that the pollen development is completely aborted in the mutant anthers, as observed in ref3-2.

ref3 mutations lead to changes in lignin deposition and its G/S ratio

Its importance in paper pulping (Huntley et al., 2003; Reddy et al., 2005) and biofuel production (Chen and Dixon, 2007) makes lignin a highly studied phenylpropanoid end-product. The effect of reducing C4H activity on lignin composition has been previously assessed by characterizing transgenic plants with downregulated C4H transcript levels. Many of these experiments have, for unknown reasons, detected a reduction in the S/G ratio of lignin monomers in tobacco and alfalfa plants, when analyzed by various methods (Sewalt et al., 1997; Blee et al., 2001; Reddy et al., 2005). Our data shows that lignin of the ref3-3 mutant has a higher S/G ratio than the wild type, with the S content essentially unchanged, whereas the G content is reduced by half. Although this result is very similar to the lignin phenotype of the Arabidopsis pal1/pal2 double mutant previously reported (Rohde et al., 2004), it is at odds with most of the literature on transgenic plants with downregulated C4H. This discrepancy may result from differences in the techniques employed to degrade and analyze the lignin polymer (Lapierre et al., 1985; Anterola and Lewis, 2002), species-specific variations in phenylpropanoid metabolism, and/or differences between reducing C4H transcript abundance through antisense or RNAi strategies and decreasing C4H activity through genetic lesions in C4H.

The availability of Arabidopsis mutants for each of the P450-catalyzed enzymatic steps in sinapate ester and monolignol biosynthesis will be of great utility in determining how lesions at various steps in the pathway impact flux of metabolites and gene expression, and how these changes lead to the spectrum of other phenotypes observed in these mutants.

Experimental procedures

Plant material

Arabidopsis plants were grown in a potting mix (Redi-Earth; Scotts, at 22°C under a photoperiod of 16-h light/8-h dark. Mutants homozygous for ref3 alleles were identified by red fluorescence under long-wave UV light as previously described (Ruegger and Chapple, 2001). The ref3-1 allele was isolated from ethyl methanesulfonate-mutagenized populations of Landsberg erecta M2 seed. Alleles ref3-2 and ref3-3 are from similarly mutagenized populations of Columbia M2 seed. All ref3 lines had been backcrossed to wild type to remove unlinked background mutations.

Metabolite analysis

HPLC analysis of 50% methanolic extracts were performed as previously described (Franke et al., 2002b). For HPLC-MS analysis, these extracts (1 ml) were first extracted with an equal volume of chloroform, and after separation of the phases, the resulting aqueous solution was acidified with 5 μl 5 m HCl, and extracted with an equal volume of ethylacetate. The ethylacetate fraction was then dried in vacuo, redissolved in 50 μl of water, separated by HPLC with a Shimadzu Shim-pack XR-ODS (3.0 × 75 mm × 2.2 μm) column (Shimadzu, and analyzed using an Agilent 6210 MSD time-of-flight mass spectrometer. Lignin quantity was analyzed by the Klason method (Kaar and Brink, 1991), whereas lignin quality was analyzed by DFRC (Lu and Ralph, 1997; Weng et al., 2008b). Hydroxycinnamic acid esters were quantified by HPLC using the extinction coefficients of cinnamic acid and sinapic acid.

Mapping of the REF3 locus

To determine the chromosomal position of the ref3 gene, ref3-1 (Landsberg erecta background) was crossed with the wild-type Columbia ecotype, and the F1 generation was allowed to self-fertilize. DNA was isolated using the CTAB method (Rogers and Bendich, 1985) from 24 homozygous ref3-1 F2 individuals. To determine whether ref3/ref3 plants from among this mapping population were also homozygous for Landsberg DNA at the C4H locus, a 2.3-kb fragment of C4H, including a polymorphic StyI site, was amplified by 40 cycles of polymerase chain reaction (PCR) using the primers 5′-ACTCCTAAGAGTGAGAACG-3′ and 5′-CTCTGGTCTAAACTCTTCAG-3′, with an annealing temperature of 55°C and an extension time of 2 min. PCR products were digested with StyI and analyzed by agarose gel electrophoresis to identify products expected for the Columbia (2.3 kb) and Landsberg (1.0 and 1.3 kb) ecotypes.

Sequencing of ref3 alleles

To search for mutations within each of the ref3 alleles, the coding sequence of C4H plus some upstream and downstream sequence was amplified by PCR using genomic DNA isolated from each ref3 allele. The 5′ portion was amplified using the primers 5′-AACACCAAGCCCCACTCACAC-3′ and 5′-TTCCCCACTCGATAGACCACAAT-3′, and the 3′ half was amplified using the primers 5′-GGGAGAAATCAACGAGGAC-3′ and 5′-TATTCCCGCTGCGTAATCTC-3′. In each case, the PCR products were subcloned into standard vectors and submitted for automated sequencing.

Complementation of ref3

A 5.4-kb HindIII fragment, previously isolated from an Arabidopsis genomic library (Bell-Lelong et al., 1997), carrying the C4H coding and regulatory sequences was ligated into the binary vector pGA482, and then introduced into Agrobacterium tumefaciens by electroporation. ref3-3 plants were transformed with Agrobacterium harboring the construct by the floral-dip method (Clough and Bent, 1998). Transformed seedlings were identified by selection on MS medium containing 50 mg l−1 kanamycin and 200 mg l−1 timentin, and were then transferred to soil. The sinapate ester content in T1 and T2 transgenics was analyzed as described above.

Chemical complementation of ref3

To determine whether the sinapoylmalate-deficient phenotype of the ref3 mutant could be chemically complemented, 3-week-old wild-type Columbia and ref3-2 plants were sprayed daily for 1 week with a 1 mmp-coumaric acid solution in 1% DMSO, containing 0.02% Silwet (OSi Specialties, now part of Momentive, as a surfactant. After 1 week, leaf extracts were prepared and analyzed by HPLC as described above.


To look for perturbations in xylem development that may have occurred as a result of decreased lignification, stems from wild-type Columbia plants and ref3 plants were fixed, dehydrated and embedded in Spurr’s resin (Peterson et al., 1978). Sections (1-μm thick) were stained with Toluidine blue O and visualized by light microscopy. To evaluate the impact of the ref3 mutations on pollen development, whole anthers were either visualized under UV-light, for autofluorescence microscopy, or stained with Toluidine blue O, and observed under a dissecting microscope.

P450 spectroscopy

C4H was expressed in E. coli using a modified form of the pCWori+ (Barnes et al., 1991) expression vector, pBOV (Franke et al., 2002b). A version of the wild-type C4H open reading frame lacking the first 20 codons was amplified by PCR using the primer pair cc56 (5′-CACGGTGATTCAAAGCT-3′) and cc58 (5′-ACAGTTCCTTGGTTTCAT-3′), and subcloned into StuI-digested pBOV to create pBOV-C4H. To create expression vectors for two of the ref3 alleles, SacII/ScaI cassettes carrying each of the ref3 mutations were excised from clones used to sequence C4H in each of the ref3 plants, and were swapped into pBOV-C4H to generate pBOV-ref3-2 and pBOV-ref3-3. To create pBOV-ref3-1, the C4H cDNA was used in two separate PCR reactions using the SP6 primer with cc267 (5′-CATCAATGTCGCCACGATTGAGACAAC-3′) and the T7 primer with cc268 (5′-GTTGTCTCAATCGTGGCGACATTGATG-3′), which introduced the ref3-1 mutation (set in bold in oligonucleotide sequences) into each of two overlapping fragments of the cDNA. The PCR products were purified by agarose gel electrophoresis, and combined in a final PCR reaction with the SP6 and T7 primers. The resulting product was digested with BamHI and AatII, and subcloned into pGEM-7Zf(+), followed by sequencing to confirm the presence of the ref3-1 mutation. A ScaI/AatII fragment was then swapped into pBOV-C4H to create pBOV-ref3-1. Restriction digests were used to confirm the presence of the mutation in the expression vectors, as each ref3 mutation abolished a recognition site for a particular restriction enzyme (ref3-1, AciI; ref3-2, HinfI; ref3-3, BamHI).

The C4H expression vectors were then transformed into E. coli DH5-α. Overnight cultures were grown and used to inoculate 200 ml of terrific broth (TB). Cultures were grown at 30°C to an OD600 of approximately 0.5, then induced by adding isopropyl-β-d-thiogalactopyranoside (IPTG) to 1 mm and δ-aminolevulinic acid to 0.5 mm. After growing for another 24 h, cells were harvested by centrifugation at 4°C for 15 min at 2500 g, resuspended in 20 ml of 50 mm potassium phosphate buffer, pH 7.25, containing 20% glycerol and 1 mm DTT, and lysed using a French Press (20 000 psi). Cell debris was removed by centrifugation at 10 000 g for 15 min, and the membrane fraction was collected from the resulting supernatant by ultracentrifugation at 140 000 g for 1 h. Membranes were resuspended in 4 ml of 50 mm potassium phosphate buffer, pH 7.3, containing 20% glycerol and 1 mm DTT, frozen in liquid nitrogen, and stored at −80°C until ready for use.

Carbon monoxide difference spectra were obtained using microsomes prepared from E. coli expressing the wild-type or mutant C4H proteins. Immediately after thawing on ice, microsomes were solubilized in 4% Triton X-100 and 0.1% CHAPS (3-((3-cholamidopropyl) dimethylammonium)-1-propanesulfonate), 50 mm sodium PIPES (piperazine-N,N′-bis(2-ethanesulfonate)), pH 7.0, 20% glycerol and 4 mm sodium Ethylenediaminetetra-acetate (EDTA), reduced by the addition of two or three grains of sodium dithionite, and divided into two cuvettes. CO difference spectroscopy was conducted essentially as previously described (Omura and Sato, 1964) using a Shimadzu UV-2100 spectrophotometer. First, a baseline spectrum was taken, then CO was gently bubbled through the test cuvette for 1 min, and the difference spectra were recorded.

Spectroscopic measurement of substrate binding was performed essentially as previously described (Estabrook and Werringloer, 1978). For each data point, an aliquot of microsomes was thawed on ice and immediately diluted in 50 mm sodium PIPES, pH 7.0, 20% glycerol and 4 mm sodium EDTA, divided into two cuvetes. A baseline difference spectrum was taken, then a 2-μl aliquot of an appropriate concentration of cinnamic acid dissolved in dimethyl sulphoxide (DMSO) was added, and a difference spectrum was recorded. All experiments were conducted in triplicate, with the Kd determined by double-reciprocal plot.

Scanning electron microscopy

Anthers were dissected from wild-type and ref3-2 mutant flowers, fixed to sample stubs using carbon tape and carbon/cryo adhesive, before being plunged into liquid nitrogen slush. After sputter-coating with gold, the specimen holder was then transferred into the chamber of a JEOL JSM-840 scanning electron microscope ( Samples were imaged at −140°C using 2 or 4 kV, aperture 3, with a probe current of 6 × 10−11 A.


This work was supported by grant DE-FG02-07ER15905 from the Division of Energy Biosciences, United States Department of Energy, and is also based upon work supported by the National Science Foundation under grant no. IOB-0450289.